Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
Research ArticleShip-Borne Phased Array Radar Using GA Based Adaptive 120572-120573-120574Filter for Beamforming Compensation and Air Target Tracking
J Mar12 Chen-Chih Liu1 and M B Basnet1
1Department of Communications Engineering Yuan-Ze University 135 Yuan-Tung Road Jungli Taoyuan 320 Taiwan2Communication Research Center Yuan-Ze University 135 Yuan-Tung Road Jungli Taoyuan 320 Taiwan
Correspondence should be addressed to J Mar eejmarsaturnyzuedutw
Received 7 March 2014 Revised 27 August 2014 Accepted 11 September 2014
Academic Editor Hang Hu
Copyright copy 2015 J Mar et al This is an open access article distributed under the Creative Commons Attribution License whichpermits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
Beam pointing error caused by ship motion over the ocean affects the tracking performance of the ship-borne phased array radarDue to the dynamic nature of the sea environments the ship-borne phased array radar must be able to compensate for the shiprsquosmotion adaptively In this paper the adaptive 120572-120573-120574 filter is proposed for the ship-borne phased array radar to compensate for thebeam pointing error and to track the air target The genetic algorithm (GA) and the particle swarm optimization (PSO) methodsare applied to estimate the gain parameters of adaptive 120572-120573-120574 filters while achieving the optimum objective of minimum rootmean square error (RMSE) The roll and pitch data measured from a gyroscope of the sea vehicle and generated from ship motionmathematical model are used in the experiments The tracking accuracy of adaptive 120572-120573-120574 filter using the GAmethod is comparedwith PSO method under different ship motion conditions The convergent time and tracking accuracy of ship-borne phased arrayradar using the proposed GA based adaptive 120572-120573-120574 filter are also compared with the adaptive extended Kalman filter (AEKF)Finally it is proved that the proposed GA based adaptive 120572-120573-120574 filter is a real time applicable algorithm for ship-borne phased arrayradar
1 Introduction
Sea wave causes the effect of roll and pitch motions onships These ship rotational motions result in measurementerror in phased array radar aboard the ship The antennastabilization to achieve the beam pointing accuracy over thelong dwell time is an important issue for ship-borne phasedarray radar [1] There are two ship motion compensationscompensation for rotational motion (ie pitch roll andheading angle) and compensation for translational motion(ie radial speed relative to the earth) Gyroscope providespitch roll heading angles of the ship speed course andvertical velocity of antenna installed on the ship at a datarate of 10Hz To compensate for the translational motionthe speed course and vertical velocity acquired from thegyroscope are averaged for the duration of radar-dwelltime Then the Doppler shift introduced by the translationalmotion is estimated in the digital signal processor (DSP)of the radar to compensate the radial velocity The radar
control computer (RCC) provides the target locations relativeto earth schedules beam directions and predicts beampointing error to the beam steering controller (BSC) whichcompensates for the beam pointing error and controls thephased array antenna to point the beam at the target directionrelative to the ship coordinates
The motion compensation method based on coordinateconversion has been described in [2 3] which requires themeasurement of a devicersquos coordinates the shiprsquos coordinatesand the earthrsquos coordinates systems to compensate the deviceerrors due to motion disturbances In [1] Kalman filteringalong with coordinates conversion is applied to reducethe beam pointing error and to stabilize a tracking beamThe 120572-120573-120574 filter which is more easily implemented than ageneralized Kalman filter is used for motion compensationunder different sea states in [1]The sea environments are verydynamic hence there is need of an adaptive system for con-trolling and compensating devices regardless of ship motionMost recent works have used Kalman filtering (KF) [4]
Hindawi Publishing CorporationInternational Journal of Antennas and PropagationVolume 2015 Article ID 563726 16 pageshttpdxdoiorg1011552015563726
2 International Journal of Antennas and Propagation
Adaptive
filter for rollpitchprediction
Shipearth coordinates conversion
Radar detection target data in
earth coordination
Adaptive
filter for target
tracking
Performanceverifications for
1 ship motion compensation2 target tracking accuracy
Earthshipcoordinates conversion
120601R(t) 120579P(t)
Δ120579E Δ120601E
120579E 120601E
120579EC 120601EC 120579A 120601A+
minus
1205721-1205731-1205741
1205722-1205732-1205742
Figure 1 High level structure of ship rotational motion compensation for phased array radar
and extended Kalman filtering (EKF) [5] to estimate theshiprsquos attitude Estimation accuracy of KF and EKF dependson the values of different parameters such as error covari-ance matrices thus the KF and EKF require knowledgeof covariance In [6] the automatic beam pointing errorcompensation mechanism employs the parallel fuzzy basisfunction network (FBFN) architecture to estimate the beampointing error caused by roll and pitch of the ship The effectof automatic beam pointing error compensation mechanismon the tracking performance of adaptive extended Kalmanfilter (AEKF) implemented in ship-borne phased array radaris investigated It shows that the tracking error of AEKF con-verges to less than about 20m at 550 iterations (seciteration)and the estimation errorwill remainwithin the range of about20m when beam pointing error is compensated by FBFNcontroller
The 120572-120573-120574 filter has a similar structure with the KF butdepends on the gain value of 120572 120573 and 120574 which are limitedand interdependent [7ndash9] The uses of various methods toadjust parameter values for Kalman filter EKF and 120572-120573-120574 filter have been proposed and implemented over the yearsFuzzy membership function is used in [10] to estimate andcorrect the target position for the signal being trackedGenetic algorithm (GA) is used in [9] to search the suitableparameter values for the 120572-120573-120574 filter which can providethe approximate position velocity and acceleration signaland simultaneously decrease themeasurement noise Particleswarm optimization (PSO) is used in [11] to tune the noisecovariance of a Kalman filter In [12] the PSO is used to findthe optimal gain parameter values for an 120572-120573-120574 filter PSOand GA methods are suitable for the real time applicationbecause they do not require differentiation and are lesscomplicated than other methods
In this paper GA based adaptive 120572-120573-120574 filter is proposedfor automatic beam pointing error correction and air targettracking in three-dimensional space Continuous monitor-ing of the environment and adapting filter gain parameterwith less computational burden is needed for real timeapplication The adaptive 120572-120573-120574 filter requires knowledgeof the coefficients for which GA algorithm is used only incertain time intervals GA is used to find the gain values byminimizing the objective function that is root mean squareerror (RMSE) The proposed adaptive 120572-120573-120574 filter algorithm
is implemented and compared with using PSO method Theroll and pitch data were recorded by a gyroscope of thesea vehicle to simulate the tracking performance of ship-borne phased array radar using the proposed adaptive 120572-120573-120574 filter In addition the roll and pitch data generated fromship motionmathematical model at sea states 2 and 3 are alsoapplied for the proposed adaptive 120572-120573-120574 filter to verify thecorrectness of the test results
The rest of this paper is organized as follows the modelof ship rotational motion coordinates transform and planararray antenna of ship-borne phased array radar are describedin detail in Section 2The algorithm of the proposed adaptive120572-120573-120574 filter based on the GA is presented in Section 3 wherethe 120572-120573-120574 filter GA gain estimator PSO gain estimatorand optimization problem formulation are described InSection 4 six different experimental cases are performed thevariations of roll and pitch angles in sea states 2 and 3 areanalyzed the GA and PSO methods are used to determinethe optimal filter gain of adaptive 120572-120573-120574 filter that minimizesRMSE the tracking performance of ship-borne phased arrayradar using the proposed adaptive 120572-120573-120574 filter is simulatedFinally conclusions are made in Section 5
2 Ship Rotational Motion Compensation
The ship-borne phased array radar must be able to com-pensate the shiprsquos motion and track the maneuvering tar-gets automatically The adaptive 120572-120573-120574 filtering algorithmis designed to real time compensate the errors caused bythe shiprsquos motion The block diagram of rotational motioncompensation system for ship-borne phased array radar isshown in Figure 1 which consists of adaptive 120572-120573-120574 filterfor beam pointing error prediction ship coordinatesearthcoordinates conversion earth coordinatesship coordinatesconversion and adaptive 120572-120573-120574 filter for target trackingThe roll angle 120601
119877(119905) and the pitch angle 120579
119875(119905) of the ship
angularmotion aremeasured by the gyroscope (120579119860 120601119860) is the
antenna beam pointing angle relative to the ship body where120579119860is the angle off antenna boresight and 120601
119860is the azimuth
angle counterclockwise from the bowof the ship Assume thatthe beam steering angle at a certain time instant is (120579
0 1206010) the
antenna point angle offset caused by the ship motion (pitch
International Journal of Antennas and Propagation 3
Table 1 Sea state parameters
Roll Pitch119860119877(∘) 119879
119877(sec) 119860
119875(∘) 119879
119875(sec)
Sea state 2 1 25 02 12Sea state 3 1 20 02 06
roll) is (120579119890 120601119890) and then the current antenna pointing angle
is (120579119864 120601119864) with respect to the earth coordinates
120579119864= 1205790+ 120579119890 120601
119864= 1206010+ 120601119890 (1)
If the beampointing error is predicted as (Δ120579119864 Δ120601119864) then the
beam pointing angle is corrected as
120579119864119862
= 1205790+ 120579119890minus Δ120579119864 120601
119864119862= 1206010+ 120601119890minus Δ120601119864 (2)
The value of (120579119864119862 120601119864119862) is approximate to (120579
0 1206010)
21 Ship Rotational Motion Model Ship is affected by thewaves in the ocean resulting in six degrees of freedomof movement In this paper assuming zero yaw angle thesimplified roll and pitch model [1 13] is adopted to describethe ship rotational motion in the earth coordinates Shiprsquosrotational motion is modeled with sinusoidal signal The rollangle is
120601119877 (119905) = 119860
119877sin (120596
119877119905) + 119899
119877 (119905) (3)
The pitch angle is
120579119875 (119905) = 119860
119875sin (120596
119875119905) + 119899
119875 (119905) (4)
where 119860119877and 119860
119875are the amplitude of shiprsquos roll and pitch
angles 119899119877(119905) and 119899
119875(119905) are assumed to be zeromeanGaussian
noise 119879119877and 119879
119875are the roll and pitch periods and 120596
119877=
2120587119879119877and120596
119875= 2120587119879
119875are the roll and pitch angular frequen-
ciesThe random roll and pitch angle errors caused by other
unknown factors including the change of shiprsquos travelingdirection and weather in the actual ship navigation envi-ronment will be considered into the standard deviation ofthe noise terms The amplitude and period parameters of seastates 2 and 3 are listed in Table 1 [1]
22 Coordinates Transform Since the phased array antennais installed on the ship the beam steering control employsthe ship body coordinates But the target tracking of adaptive120572-120573-120574 filter employs the Earth coordinates The Euler coor-dinates transform formula is used to convert antenna beampointing angle relative to the earth (120579
119864 120601119864) where 120579
119864is the
angle from vertical and 120601119864is the angle clockwise from true
north into antenna beam pointing angle relative to the shipbody (120579
119860 120601119860) [13]
[
[
sin 120579119860cos120601119860
sin 120579119860sin120601119860
cos 120579119860
]
]
= 119879 (120601) 119879 (120579) 119879 (120595)[
[
sin 120579119864cos120601119864
minus sin 120579119864sin120601119864
cos 120579119864
]
]
(5)
where the coordinates transform matrices are defined as
119879 (120601) =[
[
1 0 0
0 cos120601 sin1206010 minus sin120601 cos120601
]
]
119879 (120579) =[
[
cos 120579 0 sin 1205790 1 0
minus sin 120579 0 cos 120579]
]
119879 (120595) =[
[
cos120595 minus sin120595 0
sin120595 cos120595 0
0 0 1
]
]
(6)
where the ship roll angle 120601 is positive with downward roll ofstarboard the ship pitch angle 120579 is positive with bow pitchupward and the ship yaw angle 120595 is positive clockwise fromthe north
23 Planar Array Antenna An 119872 times 119873 element planararray [6] is designed for the ship-borne phased array radarsystem which includes the beamforming (BF) mode anddirection of arrival (DOA) mode The planar array canproduce multibeams in the azimuth and elevation by usingthe beamformer network (BFN) which consists of a set ofpower dividers and phase shifters The planar array usingamplitude comparison method [14] generates the differencesignal patterns for theDOAestimationThedifference signalsobtained from two neighboring beams canmeasure the DOAof the target signal of the ship-borne phased array radarsystem The beam pattern is expressed as [14 15]
119860119865 (120579 120601) =
119873
sum
119899 = 1
1198681119899[
119872
sum
119898=1
1198681198981119890119895(119898minus 1)(119896119889
119909sin 120579 cos120601+120573
119909)]
times 119890119895(119899 minus 1)(119896119889
119910sin 120579 sin120601+120573
119910)= 119878119909119898119878119910119899
(7)
where
119878119909119898
=
119872
sum
119898=1
1198681198981119890119895(119898minus 1)(119896119889
119909sin 120579 cos120601+120573
119909)
119878119910119898
=
119873
sum
119899 = 1
1198681119899119890119895(119899 minus 1)(119896119889
119910sin 120579 sin120601+120573
119910)
120573119909= minus 119896119889
119909sin 120579119905cos120601119905
120573119910= minus 119896119889
119910sin 120579119905sin120601119905
(8)
where 1198681198981
= 1198681119899= Chebyshev weighting [13]119872 = the number
of array elements in the 119909-axis 119873 = the number of arrayelements in 119910-axis 120601
119905= the azimuth steering angle of the
main beam 120579119905= the elevation steering angle of the main
beam 119889119909= the interelement spacing in 119909-axis 119889
119910= the
interelement spacing in 119910-axis 119896 = 2120587120582 = the phasepropagation constant and 120582 is the wavelength of the carrier
The phase of the RF signal at each array element isadjusted to steer the beam to the coordinates (120579
119905 120601119905)
4 International Journal of Antennas and Propagation
Estimation Prediction
Delay
Pitch gain estimator
Roll gain estimator
Estimation Prediction
Delay
Roll pitchs anglemeasurement
1205721-1205731-1205741 filter(g1r)
1205721-1205731-1205741 filter(g1p)
120579mr(k)
120579mp(k)
r(k)
r(k)
ar(k)
p(k)
p(k)
ap(k)
r(k + 1)r(k + 1)ar(k + 1)
p(k + 1)
p(k + 1)
ap(k + 1)
120579er(k)
120596er(k)
aer(k)
120579ep(k)
120596ep(k)
aep(k)
Figure 2 Parallel adaptive 120572-120573-120574 filter for beam pointing error prediction
3 Adaptive 120572-120573-120574 Filter
As shown in Figure 1 the adaptive 120572-120573-120574 filters are applied tobeam pointing error prediction and target tracking respec-tively
31 Parallel Adaptive 120572-120573-120574 Filter for Beam Pointing ErrorPrediction Figure 2 shows the beam pointing error predic-tion system which consists of separate adaptive 120572
1-1205731-1205741
filters to predict the roll and pitch angles of ship motionand their outputs are fed into the beam steering controller(BSC)The BSC compensates the beam pointing errors of thephased array antenna onboard the ship to reduce the impactof the roll and pitch motion on the phased array radar Rolland pitch data are random but they are similar in natureand they can be characterized with different amplitudes andfrequencies [1]The 120572
1-1205731-1205741filter predicts the next positions
using current error also known as innovation [8] Thisinnovation process has two steps prediction and estimation
The pitch prediction equations are expressed as
120579119901 (119896 + 1) = 120579
119890119901 (119896) + 1198791
120596119890119901 (
119896) +
1198792
1
2
119886119890119901 (
119896)
119901 (119896 + 1) = 120596
119890119901 (119896) + 1198791
119886119890119901 (
119896)
119886119901 (119896 + 1) = 119886
119890119901 (119896)
(9)
The pitch estimation equations are expressed as
120579119890119901 (
119896) =120579119901 (119896) + 1205721119901
[120579119898119901 (
119896) minus120579119901 (119896)] + 1198991119901 (
119896)
120596119890119901 (
119896) = 119901 (119896) +
1205731119901
1198791
[120579119898119901 (
119896) minus120579119901 (119896)]
119886119890119901 (
119896) = 119886119901 (119896) +
1205741119901
21198792
1
[120579119898119901 (
119896) minus120579119901 (119896)]
(10)
where 120579119901(119896 + 1) 119908
119901(119896 + 1) and 119886
119901(119896 + 1) are pitch angu-
lar position pitch angular velocity and pitch angular accel-eration values predicted at (119896 + 1)th interval respectivelyThe sampling time 119879
1= 01 sec 120579
119890119901(119896) 119908
119890119901(119896) and 119886
119890119901(119896)
are pitch angular position pitch angular velocity and pitchangular acceleration estimated at 119896th interval 120572
1119901 1205731119901 and
1205741119901
are the smoothing parameters whose region of stabilityand parameter constraint have been studied in [8] Therelationship among 120572
1119901 1205731119901 and 120574
1119901parameters can be
related to pitch gain 1198921119901 where 0 lt 119892
1119901lt 1 The formula
is described as follows
1205721119901
= 1 minus 1198923
1119901
1205731119901
= 15 sdot (1 + 1198921119901) (1 minus 119892
1119901)
2
1205741119901
= (1 minus 1198921119901)
3
(11)
The roll prediction equations roll estimation equations androll smoothing parameters have similar forms as the pitchprediction equations pitch estimation equations and pitchsmoothing parameters respectively
32 Adaptive 120572-120573-120574 Filter for Target Tracking in Three-Dimensional Space The GA based 120572
2-1205732-1205742filter algorithm
for target tracking is used by ship-borne phased array radarto predict the next target positions using current error andfurther improves its tracking accuracy The tracking processof 1205722-1205732-1205742filter is shown in Figure 3 where the prediction
equations are
X (119896 + 1) = X119890 (119896) + 1198792
V119890 (119896) +
1198792
2
2
A119890 (119896)
V (119896 + 1) = V119890 (119896) + 1198792
A119890 (119896)
A (119896 + 1) = A119890 (119896)
(12)
International Journal of Antennas and Propagation 5
Tracking gain estimator
Estimation Prediction
Positionmeasurement
Delay
1205722-1205732-1205742 filter(g2)
Xm(k)
X(k)
V(k)
A(k)
Xe(k)
Ve(k)
Ae(k)
X(k + 1)
V(k + 1)
A(k + 1)
Figure 3 Adaptive 120572-120573-120574 filter for target tracking in three-dimensional space
The estimation equations are
X119890 (119896) =
X (119896) + 1205722[X119898 (
119896) minusX (119896)] + n
2 (119896)
V119890 (119896) = V (119896) +
1205732
1198792
[X119898 (
119896) minus X (119896)]
A119890 (119896) = A (119896) +
1205742
21198792
2
[X119898 (
119896) minus X (119896)]
(13)
where X(119896) = [119909(119896) 119910(119896) 119911(119896)]119879 V(119896) = [V
119909(119896) V119910(119896)
V119911(119896)]119879 and A(119896) = [119886
119909(119896) 119886119910(119896) 119886119911(119896)]119879 are the position
vector velocity vector and acceleration vector respectivelyat time instant 119896 n
2(119896) is the zero mean white Gaussian noise
and the sampling time 1198792= 1 sec The relationship among
1205722 1205732 and 120574
2parameters can be related to gain factor vector
g2= [1198922 1198922 1198922]119879 The formula is described as follows
1205722=[
[
1205722119909
0 0
0 1205722119910
0
0 0 1205722119911
]
]
=[
[
1 minus 1198923
20 0
0 1 minus 1198923
20
0 0 1 minus 1198923
2
]
]
1205732=[
[
1205732119909
0 0
0 1205732119910
0
0 0 1205732119911
]
]
=[
[
[
15 (1 + 1198922) (1 minus 119892
2)2
0 0
0 15 (1 + 1198922) (1 minus 119892
2)2
0
0 0 15 (1 + 1198922) (1 minus 119892
2)2
]
]
]
1205742=[
[
1205742119909
0 0
0 1205742119910
0
0 0 1205742119911
]
]
=[
[
[
(1 minus 1198922)3
0 0
0 (1 minus 1198922)3
0
0 0 (1 minus 1198922)3
]
]
]
(14)
33 GA Gain Estimator The flow chart of GA rollpitch gainestimators of adaptive 120572-120573-120574 filter is shown in Figure 4 TheGA rollpitch gain estimators are used to estimate the roll andpitch angles gain (119892
11199031198921119901) of adaptive 120572
1-1205731-1205741filters The
GA tracking gain estimator is used to estimate the optimalgain g
2of adaptive 120572
2-1205732-1205742filter The flow chart of GA
tracking gain estimator is similar to Figure 4TheGAmethoduses selection crossover andmutation [16] techniques to findthe solutions At first the initial population which containsrandomly generated chromosomes is made Chromosomesare chains of 0 and 1 bits whose length is defined as stringlength Chromosomes number is defined as the populationsizeThese chromosomes from themating pool are shuffled to
create more randomness and then applied to the problem tofind their outputs and fitness scores accordingly Fitness valuedefines how well the chromosome solves the problem Twochromosomes are selected from the mating pool dependingon selection method Depending on the crossover rate allbits between two randomly chosen points selected from twodifferent chromosomes are interchanged which is known astwo-point crossover Chromosomes may undergo mutationwhere random bits of chromosomes are flipped dependingon the mutation rate Termination criteria are defined by thenumber of generations
Selection is done usually by using two methods asfollows roulette wheel selection and tournament selection
6 International Journal of Antennas and Propagation
Get optimal filter gain
Evaluate
YesNo
End
Roulette wheelselection
Crossover
Mutation
Generate the next generation
Is termination criteria met
Generate the initial population randomly
between range of 0 and 1
gr1gp1
Input the rollpitch sea state data
Figure 4 Flow chart of GA rollpitch gain estimator
In roulette wheel selection the chromosomes selected frompopulation are put into a mating pool The chromosomeselection probability is proportion to the fitness value Intournament selection chromosomes have to go throughseveral tournaments The chromosome which has the fittestvalue is selected and put into the mating pool Here theroulette wheel selection method is chosen to determine thesolution
Two-point crossover calls for two random points selectedfrom two different parent chromosome strings all bitsbetween two points are swapped for two parent chromo-somes Crossover rate is commonly chosen in a range of060 to 080 [16] Mutation rate describes the probability thata bit in a chromosome will be flipped (0 flips to 1 and 1flips to 0) Mutation brings diversity to the population asmutation of a chromosome is a random process which willbring randomness to present chromosome group Generationdefines iteration the GA method runs for a defined numberof generations and the final best solution is considered as theresult
34 PSO Gain Estimator [9 12 17] PSO method was firstproposed in [18] PSO is based on the number of particlesflying around the solution space to find the best solutionthat minimizes the cost function Particles move around thesolution space as soon as a particle detects a better solutionthe information is passed to other particles and then particleschange their position with respect to their best positionobserved among all the particles Figure 5 describes the PSOmethod in detail The PSO changes the velocity and positionof the particle according to (15) and (17) respectively toachieve the fitness [16]
V119894119889(119899 + 1)
= 119870 lowast [V119894119889(119899)
+ 1198881sdot rand ( ) sdot (119901119894119889(119899) minus 119909119894119889(119899))
+ 1198882sdot Rand ( ) sdot (119901119892119889(119899) minus 119909119894119889(119899))]
(15)
119870 =
2
1003816100381610038161003816100381610038161003816
2 minus 120593 minus radic1205932minus 4120593
1003816100381610038161003816100381610038161003816
where 120593 = 1198881+ 1198882 120593 gt 4 (16)
119909119894119889(119899 + 1)
= 119909119894119889(119899)
+ V119894119889(119899)
(17)
International Journal of Antennas and Propagation 7
Initialize particles with random positions and random velocities
For each particle calculate the fitness values
Is fitness value ltbest fitness value
Particle exhausted
Set global best fitness value as best fitness value
Update particle positions and velocities using (18) and (20)
Maximum iterationreached
Give global best fitnessvalue and position
Set global best as local best fitness value
Yes
No
No
No
Yes
Yes
Figure 5 Flow chart of PSO method
where 119909119894119889(119899)
and V119894119889(119899)
are position and velocity of (119894119889)thparticle at (119899)th iteration respectively 119901
119894119889(119899)and 119901
119892119889(119899)are
the best position and global best position of (119894119889)th particle at(119899)th iteration respectively Rand( ) and rand( ) are randomnumber between 0 and 1 respectively Equation (15) updatesthe velocity of the particles and (17) updates the position ofparticles 119870 means inertia 119888
1and 1198882are correction factors
and swarm size means number of particles for an iterationas described in [16]
Particles in the PSO method sometimes go out of con-straints while changing their position Reinitialization of theparticlersquos position randomly to fall within the constraints isdone as soon as a particle moves out of constraints Thisensures full use of particles and also increases the chance ofdiscovering a better solution To reduce the chance of gettingtrapped in a local minimum the advantages of using PSOmethod are that it is derivative-free easy to implement andeasy to comprehend and the solution is independent of theinitial point [17] One of the drawbacks of the PSOmethod isthat it does not guarantee success that is the solution foundmay not be the best solution
35 GAPSO Based Adaptive 120572-120573-120574 Filter The flow chart ofthe adaptive 120572-120573-120574 filter algorithm is shown in Figure 6where 119873 is the number of sampling data to determine thefilter gain In this algorithm the parameters120572120573 and 120574 whichare related to the filter gain 119892 are optimized by minimizingthe RMSE using GA or PSO method The proposed adaptive120572-120573-120574 filter processes the current sampled data 119909
119894which
predicts the next sample data 119909119894+1
as iteration continuesThe adaptive GA or PSO process only occurs at certain timeintervalThe previous119873 data that is 119909
119901(119894minus119873) sdot sdot sdot 119909
119901(119894minus1) is
sent to the adaptiveGAor PSO algorithmwhere the optimumvalue of filter gain (119892) is determined for the 120572-120573-120574 filterThen 119909
119894+1is predicted and the process continues The value
of119873 determines running time and accuracy of the algorithmso 119873 must have small enough value to support real timeimplementation and large enough value to provide acceptableaccuracy
Since the roll and pitch angles of ship motion changeover time randomly in order to speed up the processingtime and to reach the objectives of minimum estimationerror the filter gain values are real time updated by using thepipelined architectureThe recorded roll and pitch signals are
8 International Journal of Antennas and Propagation
Collect previous N data using PSO or GA algorithm
Predicted value for120572-120573-120574 filterxi
xi+1
ximinus1 ximinusFind g for Nximinus1 ximinusN
that minimizes RMSE
Figure 6 Block diagram of adaptive 120572-120573-120574 filter
segmented into block of 1000 sampled data (100 sec) for lineartrajectory target and 420 sampled data (42 sec) for circulartrajectory target respectivelyThe gain value of 120572
2-1205732-1205742filter
in 101ndash200 sec is estimated by GA and PSO methods usingthe collected data during 1ndash100 sec the gain value of 120572
2-
1205732-1205742filter in 201ndash300 sec is estimated by the GA and PSO
methods using the collected data in 101ndash200 sec and so onSimilarly the measured flight target signals are segmentedinto block of 100 sampled data (100 sec) for linear trajectorytarget and 42 sampled data (42 sec) for circular trajectorytarget respectively The gain value of 120572
2-1205732-1205742filter in 101ndash
200 sec is estimated by the GA and PSO methods using thecollected data during 1ndash100 sec the gain value of 120572
2-1205732-1205742
filter in the interval of 201ndash300 sec is estimated by the GA andPSO methods using the collected data in 101ndash200 sec and soon
For the PSO process it was found that gain parameterconverged within 100 iterations so the number of iterationsfor an adaptive 120572-120573-120574 filter was kept at constant 100 iterationsper processing cycle swarm size = 30 inertia = 07298 andcorrection factors 119888
1= 21 119888
2= 2 For the GA process we
have considered population size = 8 string length = 12crossover rate = 08 and mutation rate = 005
If 120579119894119901 119894 = 1 2 119873 and
120579119894119901 119894 = 1 2 119873 are real
measurement pitchrsquos values and estimated pitchrsquos values ofadaptive 120572
1-1205731-1205741filter respectively then 120579
119894119903 119894 = 1 2 119873
and 120579119894119903 119894 = 1 2 119873 are realmeasurement rollrsquos values and
estimated rollrsquos values of adaptive 1205721-1205731-1205741filter respectively
where 119894 is the current iteration and 119873 is the total number ofsampling data The RMSEs of pitch and roll estimations aredefined as follows
RMSE119901= radic
1
119873
119873
sum
119894 = 1
(120579119894119901minus 120579119894119901)
2
(18)
RMSE119903= radic
1
119873
119873
sum
119894 = 1
(120579119894119903minus 120579119894119903)
2
(19)
where the main purpose of GA and PSO methods is to findthe optimum values of 119892
1119901and 119892
1119903which minimize the
RMSE119901and RMSE
119903 respectively
If X119894and X
119894 119894 = 1 2 119873 are real measurement target
position vector values and estimated target position vector
values of adaptive 1205722-1205732-1205742filter the RMSE of target tracking
is defined as
RMSE = radic1
119873
119873
sum
119894 = 1
((119909119894minus 119909119894)2+ (119910119894minus 119910119894)2+ (119911119894minus 119894)2) (20)
where 119909119894 119910119894 and 119911
119894are the components of target position
vector X119894in 119909 119910 and 119911 axes The main purpose of GA and
PSO algorithms is to find the optimum gain value of g2which
minimizes the RMSE
4 Experimental Results
Six scenarios are used to simulate the tracking performanceof ship-borne phased array radar using the proposed adaptive120572-120573-120574 filter for beam pointing error compensation and targettracking in three-dimensional space
Case 1 (roll and pitch estimations using GA based adaptive120572-120573-120574 filter and measured data) The roll and pitch datarecorded by a gyroscope of the ship were used in the exper-iments The total number of data is 1000 and the samplingtime is 01 sec The angle variation of roll and pitch signalsmeasured by ship gyroscope are shown in Figure 7 whichwillresult in the beam pointing errors of phased array antennaTheGA based adaptive 120572-120573-120574 filter is used to estimate the rolland pitch angles Figures 8(a) and 8(b) show that the conver-gent time of roll and pitch angles are 3 sec and 5 sec respec-tively under simulated sea states 2 and 3 and measured dataThe shiprsquos roll and pitch signals are simulated according to (3)and (4) with roll and pitch angle parameters of sea states 2and 3 listed in Table 1 and standard deviation of 001 Itconcludes that the proposed GA based adaptive 120572-120573-120574 filter issuitable to compensate the ship motion As shown in Table 1the period of sea state 2 is longer than sea state 3 Thereforethe convergent RMSE (about 019∘ for roll about 01∘ forpitch) of sea state 2 is less than sea state 3 (about 029∘ for rollabout 022∘ for pitch) because the roll and pitch signals withshorter period in sea state 3 have rapid oscillations which aremore difficult to be predicted precisely
Case 2 (tracking linear moving target in three-dimensionalspace using GA based adaptive 120572-120573-120574 filter (without beampointing error)) Assuming the ship is not affected by thesea waves (antenna beam pointing error is zero) the trackingperformance of GA based adaptive 120572-120573-120574 filter for a straightflight target trajectory is simulated The initial position ofradar is (0 0 0) The ship moves with a speed of 10msec in
International Journal of Antennas and Propagation 9
0 100 200 300 400 500 600 700 800 900 1000
0
02
04
06
08
Pitc
h an
gle (
deg)
Time (10minus1 s)
minus12
minus08
minus06
minus04
minus02
minus1
(a)
0
1
2
3
Roll
angl
e (de
g)
0 100 200 300 400 500 600 700 800 900 1000Time (10minus1 s)
minus1
minus2
minus3
minus4
(b)
Figure 7 Demonstrating angle variation of (a) pitch and (b) roll signals measured by gyroscope
0
005
01
015
02
025
03
035
04
RMSE
(deg
)
Sea 3Sea 2Measurement
0 100 200 300 400 500 600 700 800 900 1000Time (10minus1 s)
(a)
0
005
01
015
02
025
03
035
04RM
SE (d
eg)
Sea 3Sea 2Measurement
Time (10minus1 s)0 100 200 300 400 500 600 700 800 900 1000
(b)
Figure 8 RMSE of (a) roll and (b) pitch under simulated data of sea states 2 and 3 and measured data
the 119884 axial direction The linear path equation of the ship isgiven by
X119904(119909119905 119910119905 119911119905) = X0119904+ V119904119905 (21)
where 119905 = 1 2 3 1000 sec V119904is ship velocity vector
and X0119904is initial ship position vector in three-dimensional
spaceThe radar position is updated every secondThe initialposition of flying target is (minus74840 minus129620 9100)m whichis about 150 km from the radar The flight speed of target is300msec (119883 axial velocity of 150msec and 119884 axial velocityof 260msec) The target location is updated every secondThe linear target trajectory equation is
X119898(119909119905 119910119905 119911119905) = X0119898
+ V119898119905 (22)
where V119898
is the target velocity vector and X0119898
is theinitial target position vector in three-dimensional space Thetracking accuracy of GA based adaptive 120572-120573-120574 filter for astraight flight target is shown in Figure 9 where the trajectoryestimation error is calculated by
119864119905= radic(119909
119905minus 119909119905)2+ (119910119905minus 119910119905)2+ (119911119905minus 119905)2 (23)
where (119909119905 119910119905 119911119905) and (119909
119905 119910119905 119905) denote the real target location
and estimated target location respectively It shows that theGA based adaptive 120572-120573-120574 filter converges to less than about2m after 5 sec
10 International Journal of Antennas and Propagation
Table 2 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0574 06399 0652 06904 06537 0674 06286 06313 06044 053921198921199011
0629 06664 0622 07756 0664 07399 0663 06652 06816 064981198922
0285 05827 03282 03851 00054 03480 02916 03736 0547 04864
Table 3 Estimated gain parameters for PSO based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0557 05986 0357 05345 06142 05774 05544 0529 04592 056261198921199011
0612 05898 04129 05675 05286 03953 05565 06199 05214 062851198922
0326 007 01136 0146 0098 02365 021 02547 02474 01245
0 20 40 60 80 100 1200
20
40
60
80
100
120
140
160
180
Time (s)
Et
(m)
Average estimation error = 15176m
Figure 9 Estimation error for linear target trajectory without beampointing error
Case 3 (tracking circular moving target in three-dimensionalspace using GA based adaptive 120572-120573-120574 filter (without beampointing error)) Figures 10(a) and 10(b) show the simulationscenario of beam pointing error compensation for circularmaneuvering air target and linear moving ship The shiphas the same linear moving speed and path equation asCase 2The flying target is parallel to the119883119884 plane the initialposition of flying target is (8885 4590 9100)m the velocity is300msec and the angle (120579
0) between the initial position and
the119883-axis is 30∘The trajectory of the flight vehicle is a circlewhich has a radius of 10 km The center of circular trajectoryis 119884-axis The rotational cycle time of the flight target is 209seconds and the total observation time is 419 sec
The circular trajectory equation is
X119898(119909119905 119910119905 119911119905) = (119903 cos (120579
0minus 120596119905) 119903 sin (120579
0minus 120596119905) 9100) (24)
where 119905 = 1 2 3 418 419 sec 119903 is the circular flight radiusin meter 120579
0is the initial angle between the flight vehicle
and the 119883-axis 120596 is the angular speed of flight vehicle Theaccuracy of GA based adaptive 120572-120573-120574 filter for tracking acircular flight target is shown in Figure 11 where theGAbasedadaptive 120572-120573-120574 filter converges to less than 3m after 8 secTherefore using the GA based adaptive 120572-120573-120574 filter to trackthe circular trajectory target has the larger average estimationerror and longer convergent time than linear trajectory target
Case 4 (estimating roll and pitch angles and tracking lineartrajectory target using GAPSO based adaptive 120572-120573-120574 filterand measured data (with beam pointing error)) The mea-sured roll and pitch signals shown in Figure 7 are used toevaluate the performance of GAPSO based adaptive 120572-120573-120574filter for estimating the roll and pitch angles and tracking thelinear trajectory targetThe beamof 6times6 planar array antennais steered to track the linear trajectory target The samplingfrequency is set as 10Hz Therefore the shiprsquos roll and pitchsignals are sampled 10000 points within 1000 seconds Asthe scenario described in Case 2 the ship is along the 119884
axis of 119883119884 plane in the earth coordinates the ship speedis 10msec and the flying target speed is 300msec Theflying target is parallel to the 119883119884 plane 91 km above theground and the angle between the flight direction and the119884-axis is 120∘ The largest reconnaissance distance of ship-borne radar is 150 km When the air target is flying withinthe radar detection range the beampointing error estimationand linear trajectory target tracking of ship-borne phasedarray radar are simulated
Figure 12 is a simulation flow chart used to verifythe effect of beam pointing error compensation on thetracking accuracy of adaptive 120572-120573-120574 filter Tables 2 and 3demonstrate the optimal gain parameters of adaptive 120572-120573-120574 filter for roll 119892
1199031 pitch 119892
1199011 and target tracking 119892
2
obtained by the GA and PSO methods respectively thatminimizes the RMSE Figures 13ndash17 show the estimationerrors for five different gain values which are summarizedin Table 4 It demonstrates that the greater gain value(119892 = 01 075 and 09) will generate the greater error andneed the longer convergence time The estimation errors
International Journal of Antennas and Propagation 11
Table 4 Estimation errors for different gains
Calculated by GA Calculated by PSO Fixed gain1198921199031
05392sim06904 0357sim06142 01 075 091198921199011
0622sim07756 03953sim06285 01 075 091198922
0054sim05827 007sim03259 01 075 09Average 119864
119905(m) 14781 163224 169663 534804 3991188
Table 5 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0543 05653 05424 05455 05665 056 058 0532 05414 053851198921199011
0464 0484 04459 0422 04667 04654 04509 04459 04308 046511198922
0161 00929 01937 00865 00241 00609 01174 01897 01713 01727
Y
X
Z
(a)
Y
X10 km
2730∘
(b)
Figure 10 Simulation scenarios of Case 3 for (a) 3D and (b) top view diagrams
0 10 20 30 40 50 60 70 80 90 1000
50
100
150
200
250
Time (s)
Et
(m)
Average estimation error = 24905m
Figure 11 Estimation error for circular target trajectory without beam pointing error
12 International Journal of Antennas and Propagation
Generate the target detection data
Using GA or PSO
to track the target
Generate beam steering angle
Measure or simulate roll and pitch angles
Using GA or PSO
roll and pitch angles
Simulate beam pointing error compensation
Verify tracking accuracy ofGAPSO adaptive 120572-120573-120574 filter
120572-120573-120574 filter 120572-120573-120574 filter to estimate
Figure 12 Flow chart of tracking performance simulation forCase 4experiment
0 100 200 300 400 500 600 700 800 900 10000
50
100
150
200
250
300
Time (s)
Et
(m)
Average estimation error = 14781m
Figure 13 Estimation error of GA based adaptive 120572-120573-120574 filter
presented in Figures 16 and 17 are too large to be appli-cable when the gain values are equal to 075 and 09The GA and PSO have smaller errors compared with thefixed gain values and the GA obtains the best estimationaccuracy The estimation errors of GA and PSO meth-ods were found to be close From Figures 13 14 and 15we can observe that there is a peak value that occurredat the time duration about 510 to 515 seconds becausethe horizontal beam direction of ship-borne phased array
0
50
100
150
200
250
300
Et
(m)
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 163224 m
Figure 14 Estimation error of PSO based adaptive 120572-120573-120574 filter
0
20
40
60
80
100
120
140
160
180
200Et
(m)
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 169663 m
Figure 15 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
radar is steered according to the linear trajectory targetwhich is flying through sky just above the ship at thetime instant of 510 second As shown in Figure 18 thehorizontal beam steering angle of ship-borne phased arrayradar at about the time of 510 second is changing fromdescending into ascending When the GA based adaptive120572-120573-120574 filter tracks the linear trajectory target the esti-mation error values are less than 30 meters or less ifthe peak estimation errors during these 5 seconds areignored
Case 5 (estimating roll and pitch angles and tracking lineartrajectory target using GAPSO based adaptive 120572-120573-120574 filterand simulated data (with beam pointing error)) The shiprsquosroll and pitch signals are simulated according to (3) and (4)with parameters of sea state 2 listed in Table 1 and standard
International Journal of Antennas and Propagation 13
0 200 400 600 800 10000
500
1000
1500
Time (s)
Et
(m)
Average estimation error = 534804 m
Figure 16 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0 200 400 600 800 10000
1000
2000
3000
4000
5000
6000
7000
8000
9000
Time (s)
Et
(m)
Average estimation error = 3991188 m
Figure 17 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
deviation of 001 The simulated shiprsquos roll and pitch signalsare applied for five different methods to estimate the gain ofGAPSO based adaptive 120572-120573-120574 filter The total observationtime is 1000 sec and sampling time is 01 sec The simulationreplicates 10000 times The input samples are updated byone new sample for the next iteration The other simulationscenario and parameters are the same asCase 4 Tables 5 and 6demonstrate the optimal gain parameters of adaptive 120572-120573-120574filter for roll gain 119892
1199031 pitch gain 119892
1199011 and target tracking gain
vector g2obtained by the GA and PSOmethods respectively
that minimizes RMSE Figures 19 20 21 22 and 23 showthe estimation errors for five different gain values which aresummarized in Table 7 The estimation error of GA basedadaptive 120572-120573-120574 filter for tracking a linear trajectory targetis shown in Figure 19 where the convergence time of GA
0 100 200 300 400 500 600 700 800 900 1000Time (s)
0
05
1
15
2
25
Azi
mut
h ste
erin
g an
gle (
rad)
With beam pointing errorWithout beam pointing error
Figure 18 Azimuth beam steering angle of ship-borne phased arrayradar
0
10
20
30
40
50
60
70
80
90
100
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 160756 m
Et
(m)
Figure 19 Estimation error of GA based adaptive 120572-120573-120574 filter
based adaptive 120572-120573-120574 filter is about 3 sec and the averageestimation error is about 160756m It concludes that theexperimental results in Case 5 have the same trend as inCase 4 therefore the simulated data can be used to observethe effect of shipmotion on the tracking performance of ship-borne phased array radar under different sea state conditionsThe convergence time and average estimation error of GAbased adaptive 120572-120573-120574 filter are better than themethod used in[6] where the tracking error of AEKF converges to less thanabout 20m at 550 iterations (seciteration) and the estimationerror will remain within the range of about 20m when thebeam pointing error is compensated by FBFN controller
Case 6 (estimating roll and pitch angles and tracking circularmoving target using GA based adaptive 120572-120573-120574 filter and
14 International Journal of Antennas and Propagation
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 187684 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 20 Estimation error of PSO based adaptive 120572-120573-120574 filter
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 199594 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 21 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
measured data (with beam pointing error)) As the scenariodescribed in Case 3 the measured roll and pitch signalsshown in Figure 7 are used to evaluate the performance ofGA based adaptive 120572-120573-120574 filter for estimating the roll andpitch angles and tracking the circular trajectory target Theestimation errors of 120572-120573-120574 filter with fixed gains of 119892
1199031=
1198921199011
= 1198922= 01 and 119892
1199031= 1198921199011
= 1198922= 075 are shown in
Figures 24 and 25 respectivelyThe average estimation errorsfor fixed filter gains of 01 and 075 are about 41638m and572386m respectively The optimal roll gain 119892
1199031 pitch gain
1198921199011 and target tracking gain 119892
2of adaptive 120572-120573-120574 filter
estimated by the GA method during the observation timeof 419 sec are listed in Table 8 The estimation error of GAbased adaptive 120572-120573-120574 filter for tracking a circular flight targetis shown in Figure 26 where the convergence time of GAbased adaptive 120572-120573-120574 filter is about 5 sec and the average
0
500
1000
1500
0 200 400 600 800 1000Time (s)
Average estimation error = 1080515 m
Et
(m)
Figure 22 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
0 200 400 600 800 1000Time (s)
Average estimation error = 4279449 m
Et
(m)
Figure 23 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
estimation error is about 36667m It concludes that the GAbased adaptive 120572-120573-120574 filter can greatly improve the trackingaccuracy and convergence time of the conventional 120572-120573-120574filter for tracking the circular trajectory target
5 Conclusions
This paper proposed an intelligent beam pointing error com-pensationmechanism for ship-borne phased array radarTheGA based adaptive 120572-120573-120574 filter estimates the roll and pitchangles of the ship moving in the sea and thus compensatesfor the antenna beam pointing error in order to enhance thetracking accuracy of phased array radar system
International Journal of Antennas and Propagation 15
Table 6 Estimated gain parameters for PSO based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0366 03645 0532 04519 05151 05002 04797 05254 04146 053511198921199011
0366 03647 03634 03644 03643 0363 03644 03684 03643 036421198922
0125 01271 02286 02166 01169 01799 01735 01564 00767 01346
Table 7 Estimation errors for different gains
Calculated by GA Calculated by PSO Fixed gain1198921199031
0459sim0583 03645sim05352 01 075 091198921199011
0498sim0667 0363sim03684 01 075 091198922
00586sim01697 00367sim02291 01 075 09Average 119864
119905(m) 160756 187684 199594 1080515 4279449
Table 8 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim42 43sim84 85sim126 127sim168 169sim210 211sim252 253sim294 295sim336 337sim378 379sim4191198921199031
0619 04996 05316 0644 05052 05568 05316 04459 05092 059241198921199011
0582 06727 07201 04913 04168 07169 05983 02036 05858 06211198922
0214 02215 01612 02474 01822 0157 02073 01211 02278 01074
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 41638 m
Figure 24 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
Six experimental cases are conducted to verify the perfor-mance of ship-borne phased array radar using the proposedGA based adaptive 120572-120573-120574 filter In Case 1 the GA basedadaptive 120572-120573-120574 filter is used to estimate the roll and pitchangles with the simulated and measured roll and pitchsignals respectively The simulation results show that themeasurement data has the minimum estimation error theestimation error in the sea state 2 is the second and theestimation error in the sea state 3 is the worst Cases 2 and 3show that when the GA based adaptive 120572-120573-120574 filter is appliedto estimate the beam pointing error and to track the targetin a ship-borne phased array radar the circular trajectory
0 50 100 150 200 250 300 350 400 450Time (s)
0
500
1000
1500
Average estimation error = 572386 m
Et
(m)
Figure 25 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
target will be tracked with the larger average estimation errorand longer convergence time than linear trajectory targetTheexperimental results ofCases 4 and 5demonstrate thatwhen alinear trajectory target is tracked by ship-borne phased arrayradar the average estimation error and convergence time ofship-borne phased array radar using the GA based adaptive120572-120573-120574 filter are better than PSO based adaptive 120572-120573-120574 filterand fixed filter gain 120572-120573-120574 filter The convergence time andtracking accuracy of ship-borne phased array radar usingthe proposed GA based adaptive 120572-120573-120574 filter are superiorto the AEKF significantly In conclusions the proposed GAbased adaptive 120572-120573-120574 filter is a real time applicable algorithm
16 International Journal of Antennas and Propagation
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 36667 m
Figure 26 Estimation error of GA based adaptive 120572-120573-120574 filter
whose accuracy is good with relatively less complexity Inaddition the roll and pitch signalsmeasured in real operationenvironments can be replaced with the simulated roll andpitch signals to test all the relevant properties of practical shipmotion compensation and air target tracking for ship-bornephased array radar
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
This work is supported in part by research grants fromNational Science Council China (NSC 102-2218-E-155-001)
References
[1] Z Fang and Q Rundong Ship-borne Phased Array RadarMotion Compensation System Engineering and ElectronicTechnologies 1998
[2] L J Love J F Jansen and F G Pin ldquoCompensation of waveinduced motion and force phenomenon for ship based highperformance robotic and human amplifying systemsrdquo OakRidge National Laboratory ORNLTM-2003233 2003
[3] R J Hill Motion Compensation for Ship Borne Radars andLidars US Department of Commerce National Oceanic andAtmospheric Administration Office of Oceanic and Atmo-spheric Research Earth System Research Laboratory PhysicalSciences Division 2005
[4] T I Fossen and T Perez ldquoKalman filtering for positioning andheading control of ships and offshore rigsrdquo IEEEControl SystemsMagazine vol 29 no 6 pp 32ndash46 2009
[5] S Kuchler C Pregizer J K Eberharter K Schneider and OSawodny ldquoReal-time estimation of a shiprsquos attituderdquo in Proceed-ings of theAmericanControl Conference (ACC rsquo11) pp 2411ndash2416San Francisco Calif USA July 2011
[6] J Mar K C Tsai Y T Wang and M B Basnet ldquoIntelligentmotion compensation for improving the tracking performanceof shipborne phased array radarrdquo International Journal ofAntennas and Propagation vol 2013 Article ID 384756 14pages 2013
[7] T Dirk and S Tarunraj ldquoOptimal design of 120572-120573-(120574) filtersrdquo inAmerican Control Conference vol 6 pp 4348ndash4352 ChicagoIll USA June 2000
[8] D Tenne and T Singh ldquoCharacterizing performance of 120572-120573-120574filtersrdquo IEEE Transactions on Aerospace and Electronic Systemsvol 38 no 3 pp 1072ndash1087 2002
[9] T E Lee J P Su and K W Yu ldquoParameter optimization fora third-order sampled-data trackerrdquo in Proceedings of the 2ndInternational Conference on Innovative Computing Informationand Control p 336 Kumamoto Japan September 2007
[10] Y H Lho and J H Painter ldquoA fuzzy tuned adaptive Kalmanfilterrdquo in Industrial FuzzyControl and Intelligent System pp 144ndash148 December 1993
[11] N Ramakoti A Vinay and R K Jatoth ldquoParticle swarm opti-mization aided Kalman filter for object trackingrdquo in Proceedingsof the International Conference on Advances in ComputingControl and Telecommunication Technologies (ACT rsquo09) pp 531ndash533 Kerela India December 2009
[12] T-E Lee J-P Su andK-WYu ldquoAparticle swarmoptimization-based state estimation scheme formoving objectsrdquoTransactionsof the Institute of Measurement and Control vol 34 no 2-3 pp236ndash254 2012
[13] D C Law S A McLaughlin M J Post et al ldquoAn electronicallystabilized phased array system for shipborne atmospheric windprofilingrdquo Journal of Atmospheric and Oceanic Technology vol19 no 6 pp 924ndash933 2002
[14] C A Balanis AntennaTheory Analysis and Design JohnWileyamp Sons New York NY USA 3rd edition 2005
[15] J Mar and Y R Lin ldquoImplementation of SDR digital beam-former for microsatellite SARrdquo IEEE Geoscience and RemoteSensing Letters vol 6 no 1 pp 92ndash96 2009
[16] R C Eberhart and Y Shi Computational Intelligence Conceptsto Implementations ElsevierMorgan Kaufmann 2007
[17] K Y Lee and J-B Park ldquoApplication of particle swarm optimi-zation to economic dispatch problem advantages anddisadvan-tagesrdquo in Proceedings of the IEEE PES Power Systems Conferenceand Exposition (PSCE rsquo06) pp 188ndash192 Atlanta Ga USANovember 2006
[18] R C Eberhart and J Kennedy ldquoNew optimizer using particleswarm theoryrdquo in Proceedings of the 6th International Sympo-sium onMicroMachine and Human Science pp 39ndash43 NagoyaJapan October 1995
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
2 International Journal of Antennas and Propagation
Adaptive
filter for rollpitchprediction
Shipearth coordinates conversion
Radar detection target data in
earth coordination
Adaptive
filter for target
tracking
Performanceverifications for
1 ship motion compensation2 target tracking accuracy
Earthshipcoordinates conversion
120601R(t) 120579P(t)
Δ120579E Δ120601E
120579E 120601E
120579EC 120601EC 120579A 120601A+
minus
1205721-1205731-1205741
1205722-1205732-1205742
Figure 1 High level structure of ship rotational motion compensation for phased array radar
and extended Kalman filtering (EKF) [5] to estimate theshiprsquos attitude Estimation accuracy of KF and EKF dependson the values of different parameters such as error covari-ance matrices thus the KF and EKF require knowledgeof covariance In [6] the automatic beam pointing errorcompensation mechanism employs the parallel fuzzy basisfunction network (FBFN) architecture to estimate the beampointing error caused by roll and pitch of the ship The effectof automatic beam pointing error compensation mechanismon the tracking performance of adaptive extended Kalmanfilter (AEKF) implemented in ship-borne phased array radaris investigated It shows that the tracking error of AEKF con-verges to less than about 20m at 550 iterations (seciteration)and the estimation errorwill remainwithin the range of about20m when beam pointing error is compensated by FBFNcontroller
The 120572-120573-120574 filter has a similar structure with the KF butdepends on the gain value of 120572 120573 and 120574 which are limitedand interdependent [7ndash9] The uses of various methods toadjust parameter values for Kalman filter EKF and 120572-120573-120574 filter have been proposed and implemented over the yearsFuzzy membership function is used in [10] to estimate andcorrect the target position for the signal being trackedGenetic algorithm (GA) is used in [9] to search the suitableparameter values for the 120572-120573-120574 filter which can providethe approximate position velocity and acceleration signaland simultaneously decrease themeasurement noise Particleswarm optimization (PSO) is used in [11] to tune the noisecovariance of a Kalman filter In [12] the PSO is used to findthe optimal gain parameter values for an 120572-120573-120574 filter PSOand GA methods are suitable for the real time applicationbecause they do not require differentiation and are lesscomplicated than other methods
In this paper GA based adaptive 120572-120573-120574 filter is proposedfor automatic beam pointing error correction and air targettracking in three-dimensional space Continuous monitor-ing of the environment and adapting filter gain parameterwith less computational burden is needed for real timeapplication The adaptive 120572-120573-120574 filter requires knowledgeof the coefficients for which GA algorithm is used only incertain time intervals GA is used to find the gain values byminimizing the objective function that is root mean squareerror (RMSE) The proposed adaptive 120572-120573-120574 filter algorithm
is implemented and compared with using PSO method Theroll and pitch data were recorded by a gyroscope of thesea vehicle to simulate the tracking performance of ship-borne phased array radar using the proposed adaptive 120572-120573-120574 filter In addition the roll and pitch data generated fromship motionmathematical model at sea states 2 and 3 are alsoapplied for the proposed adaptive 120572-120573-120574 filter to verify thecorrectness of the test results
The rest of this paper is organized as follows the modelof ship rotational motion coordinates transform and planararray antenna of ship-borne phased array radar are describedin detail in Section 2The algorithm of the proposed adaptive120572-120573-120574 filter based on the GA is presented in Section 3 wherethe 120572-120573-120574 filter GA gain estimator PSO gain estimatorand optimization problem formulation are described InSection 4 six different experimental cases are performed thevariations of roll and pitch angles in sea states 2 and 3 areanalyzed the GA and PSO methods are used to determinethe optimal filter gain of adaptive 120572-120573-120574 filter that minimizesRMSE the tracking performance of ship-borne phased arrayradar using the proposed adaptive 120572-120573-120574 filter is simulatedFinally conclusions are made in Section 5
2 Ship Rotational Motion Compensation
The ship-borne phased array radar must be able to com-pensate the shiprsquos motion and track the maneuvering tar-gets automatically The adaptive 120572-120573-120574 filtering algorithmis designed to real time compensate the errors caused bythe shiprsquos motion The block diagram of rotational motioncompensation system for ship-borne phased array radar isshown in Figure 1 which consists of adaptive 120572-120573-120574 filterfor beam pointing error prediction ship coordinatesearthcoordinates conversion earth coordinatesship coordinatesconversion and adaptive 120572-120573-120574 filter for target trackingThe roll angle 120601
119877(119905) and the pitch angle 120579
119875(119905) of the ship
angularmotion aremeasured by the gyroscope (120579119860 120601119860) is the
antenna beam pointing angle relative to the ship body where120579119860is the angle off antenna boresight and 120601
119860is the azimuth
angle counterclockwise from the bowof the ship Assume thatthe beam steering angle at a certain time instant is (120579
0 1206010) the
antenna point angle offset caused by the ship motion (pitch
International Journal of Antennas and Propagation 3
Table 1 Sea state parameters
Roll Pitch119860119877(∘) 119879
119877(sec) 119860
119875(∘) 119879
119875(sec)
Sea state 2 1 25 02 12Sea state 3 1 20 02 06
roll) is (120579119890 120601119890) and then the current antenna pointing angle
is (120579119864 120601119864) with respect to the earth coordinates
120579119864= 1205790+ 120579119890 120601
119864= 1206010+ 120601119890 (1)
If the beampointing error is predicted as (Δ120579119864 Δ120601119864) then the
beam pointing angle is corrected as
120579119864119862
= 1205790+ 120579119890minus Δ120579119864 120601
119864119862= 1206010+ 120601119890minus Δ120601119864 (2)
The value of (120579119864119862 120601119864119862) is approximate to (120579
0 1206010)
21 Ship Rotational Motion Model Ship is affected by thewaves in the ocean resulting in six degrees of freedomof movement In this paper assuming zero yaw angle thesimplified roll and pitch model [1 13] is adopted to describethe ship rotational motion in the earth coordinates Shiprsquosrotational motion is modeled with sinusoidal signal The rollangle is
120601119877 (119905) = 119860
119877sin (120596
119877119905) + 119899
119877 (119905) (3)
The pitch angle is
120579119875 (119905) = 119860
119875sin (120596
119875119905) + 119899
119875 (119905) (4)
where 119860119877and 119860
119875are the amplitude of shiprsquos roll and pitch
angles 119899119877(119905) and 119899
119875(119905) are assumed to be zeromeanGaussian
noise 119879119877and 119879
119875are the roll and pitch periods and 120596
119877=
2120587119879119877and120596
119875= 2120587119879
119875are the roll and pitch angular frequen-
ciesThe random roll and pitch angle errors caused by other
unknown factors including the change of shiprsquos travelingdirection and weather in the actual ship navigation envi-ronment will be considered into the standard deviation ofthe noise terms The amplitude and period parameters of seastates 2 and 3 are listed in Table 1 [1]
22 Coordinates Transform Since the phased array antennais installed on the ship the beam steering control employsthe ship body coordinates But the target tracking of adaptive120572-120573-120574 filter employs the Earth coordinates The Euler coor-dinates transform formula is used to convert antenna beampointing angle relative to the earth (120579
119864 120601119864) where 120579
119864is the
angle from vertical and 120601119864is the angle clockwise from true
north into antenna beam pointing angle relative to the shipbody (120579
119860 120601119860) [13]
[
[
sin 120579119860cos120601119860
sin 120579119860sin120601119860
cos 120579119860
]
]
= 119879 (120601) 119879 (120579) 119879 (120595)[
[
sin 120579119864cos120601119864
minus sin 120579119864sin120601119864
cos 120579119864
]
]
(5)
where the coordinates transform matrices are defined as
119879 (120601) =[
[
1 0 0
0 cos120601 sin1206010 minus sin120601 cos120601
]
]
119879 (120579) =[
[
cos 120579 0 sin 1205790 1 0
minus sin 120579 0 cos 120579]
]
119879 (120595) =[
[
cos120595 minus sin120595 0
sin120595 cos120595 0
0 0 1
]
]
(6)
where the ship roll angle 120601 is positive with downward roll ofstarboard the ship pitch angle 120579 is positive with bow pitchupward and the ship yaw angle 120595 is positive clockwise fromthe north
23 Planar Array Antenna An 119872 times 119873 element planararray [6] is designed for the ship-borne phased array radarsystem which includes the beamforming (BF) mode anddirection of arrival (DOA) mode The planar array canproduce multibeams in the azimuth and elevation by usingthe beamformer network (BFN) which consists of a set ofpower dividers and phase shifters The planar array usingamplitude comparison method [14] generates the differencesignal patterns for theDOAestimationThedifference signalsobtained from two neighboring beams canmeasure the DOAof the target signal of the ship-borne phased array radarsystem The beam pattern is expressed as [14 15]
119860119865 (120579 120601) =
119873
sum
119899 = 1
1198681119899[
119872
sum
119898=1
1198681198981119890119895(119898minus 1)(119896119889
119909sin 120579 cos120601+120573
119909)]
times 119890119895(119899 minus 1)(119896119889
119910sin 120579 sin120601+120573
119910)= 119878119909119898119878119910119899
(7)
where
119878119909119898
=
119872
sum
119898=1
1198681198981119890119895(119898minus 1)(119896119889
119909sin 120579 cos120601+120573
119909)
119878119910119898
=
119873
sum
119899 = 1
1198681119899119890119895(119899 minus 1)(119896119889
119910sin 120579 sin120601+120573
119910)
120573119909= minus 119896119889
119909sin 120579119905cos120601119905
120573119910= minus 119896119889
119910sin 120579119905sin120601119905
(8)
where 1198681198981
= 1198681119899= Chebyshev weighting [13]119872 = the number
of array elements in the 119909-axis 119873 = the number of arrayelements in 119910-axis 120601
119905= the azimuth steering angle of the
main beam 120579119905= the elevation steering angle of the main
beam 119889119909= the interelement spacing in 119909-axis 119889
119910= the
interelement spacing in 119910-axis 119896 = 2120587120582 = the phasepropagation constant and 120582 is the wavelength of the carrier
The phase of the RF signal at each array element isadjusted to steer the beam to the coordinates (120579
119905 120601119905)
4 International Journal of Antennas and Propagation
Estimation Prediction
Delay
Pitch gain estimator
Roll gain estimator
Estimation Prediction
Delay
Roll pitchs anglemeasurement
1205721-1205731-1205741 filter(g1r)
1205721-1205731-1205741 filter(g1p)
120579mr(k)
120579mp(k)
r(k)
r(k)
ar(k)
p(k)
p(k)
ap(k)
r(k + 1)r(k + 1)ar(k + 1)
p(k + 1)
p(k + 1)
ap(k + 1)
120579er(k)
120596er(k)
aer(k)
120579ep(k)
120596ep(k)
aep(k)
Figure 2 Parallel adaptive 120572-120573-120574 filter for beam pointing error prediction
3 Adaptive 120572-120573-120574 Filter
As shown in Figure 1 the adaptive 120572-120573-120574 filters are applied tobeam pointing error prediction and target tracking respec-tively
31 Parallel Adaptive 120572-120573-120574 Filter for Beam Pointing ErrorPrediction Figure 2 shows the beam pointing error predic-tion system which consists of separate adaptive 120572
1-1205731-1205741
filters to predict the roll and pitch angles of ship motionand their outputs are fed into the beam steering controller(BSC)The BSC compensates the beam pointing errors of thephased array antenna onboard the ship to reduce the impactof the roll and pitch motion on the phased array radar Rolland pitch data are random but they are similar in natureand they can be characterized with different amplitudes andfrequencies [1]The 120572
1-1205731-1205741filter predicts the next positions
using current error also known as innovation [8] Thisinnovation process has two steps prediction and estimation
The pitch prediction equations are expressed as
120579119901 (119896 + 1) = 120579
119890119901 (119896) + 1198791
120596119890119901 (
119896) +
1198792
1
2
119886119890119901 (
119896)
119901 (119896 + 1) = 120596
119890119901 (119896) + 1198791
119886119890119901 (
119896)
119886119901 (119896 + 1) = 119886
119890119901 (119896)
(9)
The pitch estimation equations are expressed as
120579119890119901 (
119896) =120579119901 (119896) + 1205721119901
[120579119898119901 (
119896) minus120579119901 (119896)] + 1198991119901 (
119896)
120596119890119901 (
119896) = 119901 (119896) +
1205731119901
1198791
[120579119898119901 (
119896) minus120579119901 (119896)]
119886119890119901 (
119896) = 119886119901 (119896) +
1205741119901
21198792
1
[120579119898119901 (
119896) minus120579119901 (119896)]
(10)
where 120579119901(119896 + 1) 119908
119901(119896 + 1) and 119886
119901(119896 + 1) are pitch angu-
lar position pitch angular velocity and pitch angular accel-eration values predicted at (119896 + 1)th interval respectivelyThe sampling time 119879
1= 01 sec 120579
119890119901(119896) 119908
119890119901(119896) and 119886
119890119901(119896)
are pitch angular position pitch angular velocity and pitchangular acceleration estimated at 119896th interval 120572
1119901 1205731119901 and
1205741119901
are the smoothing parameters whose region of stabilityand parameter constraint have been studied in [8] Therelationship among 120572
1119901 1205731119901 and 120574
1119901parameters can be
related to pitch gain 1198921119901 where 0 lt 119892
1119901lt 1 The formula
is described as follows
1205721119901
= 1 minus 1198923
1119901
1205731119901
= 15 sdot (1 + 1198921119901) (1 minus 119892
1119901)
2
1205741119901
= (1 minus 1198921119901)
3
(11)
The roll prediction equations roll estimation equations androll smoothing parameters have similar forms as the pitchprediction equations pitch estimation equations and pitchsmoothing parameters respectively
32 Adaptive 120572-120573-120574 Filter for Target Tracking in Three-Dimensional Space The GA based 120572
2-1205732-1205742filter algorithm
for target tracking is used by ship-borne phased array radarto predict the next target positions using current error andfurther improves its tracking accuracy The tracking processof 1205722-1205732-1205742filter is shown in Figure 3 where the prediction
equations are
X (119896 + 1) = X119890 (119896) + 1198792
V119890 (119896) +
1198792
2
2
A119890 (119896)
V (119896 + 1) = V119890 (119896) + 1198792
A119890 (119896)
A (119896 + 1) = A119890 (119896)
(12)
International Journal of Antennas and Propagation 5
Tracking gain estimator
Estimation Prediction
Positionmeasurement
Delay
1205722-1205732-1205742 filter(g2)
Xm(k)
X(k)
V(k)
A(k)
Xe(k)
Ve(k)
Ae(k)
X(k + 1)
V(k + 1)
A(k + 1)
Figure 3 Adaptive 120572-120573-120574 filter for target tracking in three-dimensional space
The estimation equations are
X119890 (119896) =
X (119896) + 1205722[X119898 (
119896) minusX (119896)] + n
2 (119896)
V119890 (119896) = V (119896) +
1205732
1198792
[X119898 (
119896) minus X (119896)]
A119890 (119896) = A (119896) +
1205742
21198792
2
[X119898 (
119896) minus X (119896)]
(13)
where X(119896) = [119909(119896) 119910(119896) 119911(119896)]119879 V(119896) = [V
119909(119896) V119910(119896)
V119911(119896)]119879 and A(119896) = [119886
119909(119896) 119886119910(119896) 119886119911(119896)]119879 are the position
vector velocity vector and acceleration vector respectivelyat time instant 119896 n
2(119896) is the zero mean white Gaussian noise
and the sampling time 1198792= 1 sec The relationship among
1205722 1205732 and 120574
2parameters can be related to gain factor vector
g2= [1198922 1198922 1198922]119879 The formula is described as follows
1205722=[
[
1205722119909
0 0
0 1205722119910
0
0 0 1205722119911
]
]
=[
[
1 minus 1198923
20 0
0 1 minus 1198923
20
0 0 1 minus 1198923
2
]
]
1205732=[
[
1205732119909
0 0
0 1205732119910
0
0 0 1205732119911
]
]
=[
[
[
15 (1 + 1198922) (1 minus 119892
2)2
0 0
0 15 (1 + 1198922) (1 minus 119892
2)2
0
0 0 15 (1 + 1198922) (1 minus 119892
2)2
]
]
]
1205742=[
[
1205742119909
0 0
0 1205742119910
0
0 0 1205742119911
]
]
=[
[
[
(1 minus 1198922)3
0 0
0 (1 minus 1198922)3
0
0 0 (1 minus 1198922)3
]
]
]
(14)
33 GA Gain Estimator The flow chart of GA rollpitch gainestimators of adaptive 120572-120573-120574 filter is shown in Figure 4 TheGA rollpitch gain estimators are used to estimate the roll andpitch angles gain (119892
11199031198921119901) of adaptive 120572
1-1205731-1205741filters The
GA tracking gain estimator is used to estimate the optimalgain g
2of adaptive 120572
2-1205732-1205742filter The flow chart of GA
tracking gain estimator is similar to Figure 4TheGAmethoduses selection crossover andmutation [16] techniques to findthe solutions At first the initial population which containsrandomly generated chromosomes is made Chromosomesare chains of 0 and 1 bits whose length is defined as stringlength Chromosomes number is defined as the populationsizeThese chromosomes from themating pool are shuffled to
create more randomness and then applied to the problem tofind their outputs and fitness scores accordingly Fitness valuedefines how well the chromosome solves the problem Twochromosomes are selected from the mating pool dependingon selection method Depending on the crossover rate allbits between two randomly chosen points selected from twodifferent chromosomes are interchanged which is known astwo-point crossover Chromosomes may undergo mutationwhere random bits of chromosomes are flipped dependingon the mutation rate Termination criteria are defined by thenumber of generations
Selection is done usually by using two methods asfollows roulette wheel selection and tournament selection
6 International Journal of Antennas and Propagation
Get optimal filter gain
Evaluate
YesNo
End
Roulette wheelselection
Crossover
Mutation
Generate the next generation
Is termination criteria met
Generate the initial population randomly
between range of 0 and 1
gr1gp1
Input the rollpitch sea state data
Figure 4 Flow chart of GA rollpitch gain estimator
In roulette wheel selection the chromosomes selected frompopulation are put into a mating pool The chromosomeselection probability is proportion to the fitness value Intournament selection chromosomes have to go throughseveral tournaments The chromosome which has the fittestvalue is selected and put into the mating pool Here theroulette wheel selection method is chosen to determine thesolution
Two-point crossover calls for two random points selectedfrom two different parent chromosome strings all bitsbetween two points are swapped for two parent chromo-somes Crossover rate is commonly chosen in a range of060 to 080 [16] Mutation rate describes the probability thata bit in a chromosome will be flipped (0 flips to 1 and 1flips to 0) Mutation brings diversity to the population asmutation of a chromosome is a random process which willbring randomness to present chromosome group Generationdefines iteration the GA method runs for a defined numberof generations and the final best solution is considered as theresult
34 PSO Gain Estimator [9 12 17] PSO method was firstproposed in [18] PSO is based on the number of particlesflying around the solution space to find the best solutionthat minimizes the cost function Particles move around thesolution space as soon as a particle detects a better solutionthe information is passed to other particles and then particleschange their position with respect to their best positionobserved among all the particles Figure 5 describes the PSOmethod in detail The PSO changes the velocity and positionof the particle according to (15) and (17) respectively toachieve the fitness [16]
V119894119889(119899 + 1)
= 119870 lowast [V119894119889(119899)
+ 1198881sdot rand ( ) sdot (119901119894119889(119899) minus 119909119894119889(119899))
+ 1198882sdot Rand ( ) sdot (119901119892119889(119899) minus 119909119894119889(119899))]
(15)
119870 =
2
1003816100381610038161003816100381610038161003816
2 minus 120593 minus radic1205932minus 4120593
1003816100381610038161003816100381610038161003816
where 120593 = 1198881+ 1198882 120593 gt 4 (16)
119909119894119889(119899 + 1)
= 119909119894119889(119899)
+ V119894119889(119899)
(17)
International Journal of Antennas and Propagation 7
Initialize particles with random positions and random velocities
For each particle calculate the fitness values
Is fitness value ltbest fitness value
Particle exhausted
Set global best fitness value as best fitness value
Update particle positions and velocities using (18) and (20)
Maximum iterationreached
Give global best fitnessvalue and position
Set global best as local best fitness value
Yes
No
No
No
Yes
Yes
Figure 5 Flow chart of PSO method
where 119909119894119889(119899)
and V119894119889(119899)
are position and velocity of (119894119889)thparticle at (119899)th iteration respectively 119901
119894119889(119899)and 119901
119892119889(119899)are
the best position and global best position of (119894119889)th particle at(119899)th iteration respectively Rand( ) and rand( ) are randomnumber between 0 and 1 respectively Equation (15) updatesthe velocity of the particles and (17) updates the position ofparticles 119870 means inertia 119888
1and 1198882are correction factors
and swarm size means number of particles for an iterationas described in [16]
Particles in the PSO method sometimes go out of con-straints while changing their position Reinitialization of theparticlersquos position randomly to fall within the constraints isdone as soon as a particle moves out of constraints Thisensures full use of particles and also increases the chance ofdiscovering a better solution To reduce the chance of gettingtrapped in a local minimum the advantages of using PSOmethod are that it is derivative-free easy to implement andeasy to comprehend and the solution is independent of theinitial point [17] One of the drawbacks of the PSOmethod isthat it does not guarantee success that is the solution foundmay not be the best solution
35 GAPSO Based Adaptive 120572-120573-120574 Filter The flow chart ofthe adaptive 120572-120573-120574 filter algorithm is shown in Figure 6where 119873 is the number of sampling data to determine thefilter gain In this algorithm the parameters120572120573 and 120574 whichare related to the filter gain 119892 are optimized by minimizingthe RMSE using GA or PSO method The proposed adaptive120572-120573-120574 filter processes the current sampled data 119909
119894which
predicts the next sample data 119909119894+1
as iteration continuesThe adaptive GA or PSO process only occurs at certain timeintervalThe previous119873 data that is 119909
119901(119894minus119873) sdot sdot sdot 119909
119901(119894minus1) is
sent to the adaptiveGAor PSO algorithmwhere the optimumvalue of filter gain (119892) is determined for the 120572-120573-120574 filterThen 119909
119894+1is predicted and the process continues The value
of119873 determines running time and accuracy of the algorithmso 119873 must have small enough value to support real timeimplementation and large enough value to provide acceptableaccuracy
Since the roll and pitch angles of ship motion changeover time randomly in order to speed up the processingtime and to reach the objectives of minimum estimationerror the filter gain values are real time updated by using thepipelined architectureThe recorded roll and pitch signals are
8 International Journal of Antennas and Propagation
Collect previous N data using PSO or GA algorithm
Predicted value for120572-120573-120574 filterxi
xi+1
ximinus1 ximinusFind g for Nximinus1 ximinusN
that minimizes RMSE
Figure 6 Block diagram of adaptive 120572-120573-120574 filter
segmented into block of 1000 sampled data (100 sec) for lineartrajectory target and 420 sampled data (42 sec) for circulartrajectory target respectivelyThe gain value of 120572
2-1205732-1205742filter
in 101ndash200 sec is estimated by GA and PSO methods usingthe collected data during 1ndash100 sec the gain value of 120572
2-
1205732-1205742filter in 201ndash300 sec is estimated by the GA and PSO
methods using the collected data in 101ndash200 sec and so onSimilarly the measured flight target signals are segmentedinto block of 100 sampled data (100 sec) for linear trajectorytarget and 42 sampled data (42 sec) for circular trajectorytarget respectively The gain value of 120572
2-1205732-1205742filter in 101ndash
200 sec is estimated by the GA and PSO methods using thecollected data during 1ndash100 sec the gain value of 120572
2-1205732-1205742
filter in the interval of 201ndash300 sec is estimated by the GA andPSO methods using the collected data in 101ndash200 sec and soon
For the PSO process it was found that gain parameterconverged within 100 iterations so the number of iterationsfor an adaptive 120572-120573-120574 filter was kept at constant 100 iterationsper processing cycle swarm size = 30 inertia = 07298 andcorrection factors 119888
1= 21 119888
2= 2 For the GA process we
have considered population size = 8 string length = 12crossover rate = 08 and mutation rate = 005
If 120579119894119901 119894 = 1 2 119873 and
120579119894119901 119894 = 1 2 119873 are real
measurement pitchrsquos values and estimated pitchrsquos values ofadaptive 120572
1-1205731-1205741filter respectively then 120579
119894119903 119894 = 1 2 119873
and 120579119894119903 119894 = 1 2 119873 are realmeasurement rollrsquos values and
estimated rollrsquos values of adaptive 1205721-1205731-1205741filter respectively
where 119894 is the current iteration and 119873 is the total number ofsampling data The RMSEs of pitch and roll estimations aredefined as follows
RMSE119901= radic
1
119873
119873
sum
119894 = 1
(120579119894119901minus 120579119894119901)
2
(18)
RMSE119903= radic
1
119873
119873
sum
119894 = 1
(120579119894119903minus 120579119894119903)
2
(19)
where the main purpose of GA and PSO methods is to findthe optimum values of 119892
1119901and 119892
1119903which minimize the
RMSE119901and RMSE
119903 respectively
If X119894and X
119894 119894 = 1 2 119873 are real measurement target
position vector values and estimated target position vector
values of adaptive 1205722-1205732-1205742filter the RMSE of target tracking
is defined as
RMSE = radic1
119873
119873
sum
119894 = 1
((119909119894minus 119909119894)2+ (119910119894minus 119910119894)2+ (119911119894minus 119894)2) (20)
where 119909119894 119910119894 and 119911
119894are the components of target position
vector X119894in 119909 119910 and 119911 axes The main purpose of GA and
PSO algorithms is to find the optimum gain value of g2which
minimizes the RMSE
4 Experimental Results
Six scenarios are used to simulate the tracking performanceof ship-borne phased array radar using the proposed adaptive120572-120573-120574 filter for beam pointing error compensation and targettracking in three-dimensional space
Case 1 (roll and pitch estimations using GA based adaptive120572-120573-120574 filter and measured data) The roll and pitch datarecorded by a gyroscope of the ship were used in the exper-iments The total number of data is 1000 and the samplingtime is 01 sec The angle variation of roll and pitch signalsmeasured by ship gyroscope are shown in Figure 7 whichwillresult in the beam pointing errors of phased array antennaTheGA based adaptive 120572-120573-120574 filter is used to estimate the rolland pitch angles Figures 8(a) and 8(b) show that the conver-gent time of roll and pitch angles are 3 sec and 5 sec respec-tively under simulated sea states 2 and 3 and measured dataThe shiprsquos roll and pitch signals are simulated according to (3)and (4) with roll and pitch angle parameters of sea states 2and 3 listed in Table 1 and standard deviation of 001 Itconcludes that the proposed GA based adaptive 120572-120573-120574 filter issuitable to compensate the ship motion As shown in Table 1the period of sea state 2 is longer than sea state 3 Thereforethe convergent RMSE (about 019∘ for roll about 01∘ forpitch) of sea state 2 is less than sea state 3 (about 029∘ for rollabout 022∘ for pitch) because the roll and pitch signals withshorter period in sea state 3 have rapid oscillations which aremore difficult to be predicted precisely
Case 2 (tracking linear moving target in three-dimensionalspace using GA based adaptive 120572-120573-120574 filter (without beampointing error)) Assuming the ship is not affected by thesea waves (antenna beam pointing error is zero) the trackingperformance of GA based adaptive 120572-120573-120574 filter for a straightflight target trajectory is simulated The initial position ofradar is (0 0 0) The ship moves with a speed of 10msec in
International Journal of Antennas and Propagation 9
0 100 200 300 400 500 600 700 800 900 1000
0
02
04
06
08
Pitc
h an
gle (
deg)
Time (10minus1 s)
minus12
minus08
minus06
minus04
minus02
minus1
(a)
0
1
2
3
Roll
angl
e (de
g)
0 100 200 300 400 500 600 700 800 900 1000Time (10minus1 s)
minus1
minus2
minus3
minus4
(b)
Figure 7 Demonstrating angle variation of (a) pitch and (b) roll signals measured by gyroscope
0
005
01
015
02
025
03
035
04
RMSE
(deg
)
Sea 3Sea 2Measurement
0 100 200 300 400 500 600 700 800 900 1000Time (10minus1 s)
(a)
0
005
01
015
02
025
03
035
04RM
SE (d
eg)
Sea 3Sea 2Measurement
Time (10minus1 s)0 100 200 300 400 500 600 700 800 900 1000
(b)
Figure 8 RMSE of (a) roll and (b) pitch under simulated data of sea states 2 and 3 and measured data
the 119884 axial direction The linear path equation of the ship isgiven by
X119904(119909119905 119910119905 119911119905) = X0119904+ V119904119905 (21)
where 119905 = 1 2 3 1000 sec V119904is ship velocity vector
and X0119904is initial ship position vector in three-dimensional
spaceThe radar position is updated every secondThe initialposition of flying target is (minus74840 minus129620 9100)m whichis about 150 km from the radar The flight speed of target is300msec (119883 axial velocity of 150msec and 119884 axial velocityof 260msec) The target location is updated every secondThe linear target trajectory equation is
X119898(119909119905 119910119905 119911119905) = X0119898
+ V119898119905 (22)
where V119898
is the target velocity vector and X0119898
is theinitial target position vector in three-dimensional space Thetracking accuracy of GA based adaptive 120572-120573-120574 filter for astraight flight target is shown in Figure 9 where the trajectoryestimation error is calculated by
119864119905= radic(119909
119905minus 119909119905)2+ (119910119905minus 119910119905)2+ (119911119905minus 119905)2 (23)
where (119909119905 119910119905 119911119905) and (119909
119905 119910119905 119905) denote the real target location
and estimated target location respectively It shows that theGA based adaptive 120572-120573-120574 filter converges to less than about2m after 5 sec
10 International Journal of Antennas and Propagation
Table 2 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0574 06399 0652 06904 06537 0674 06286 06313 06044 053921198921199011
0629 06664 0622 07756 0664 07399 0663 06652 06816 064981198922
0285 05827 03282 03851 00054 03480 02916 03736 0547 04864
Table 3 Estimated gain parameters for PSO based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0557 05986 0357 05345 06142 05774 05544 0529 04592 056261198921199011
0612 05898 04129 05675 05286 03953 05565 06199 05214 062851198922
0326 007 01136 0146 0098 02365 021 02547 02474 01245
0 20 40 60 80 100 1200
20
40
60
80
100
120
140
160
180
Time (s)
Et
(m)
Average estimation error = 15176m
Figure 9 Estimation error for linear target trajectory without beampointing error
Case 3 (tracking circular moving target in three-dimensionalspace using GA based adaptive 120572-120573-120574 filter (without beampointing error)) Figures 10(a) and 10(b) show the simulationscenario of beam pointing error compensation for circularmaneuvering air target and linear moving ship The shiphas the same linear moving speed and path equation asCase 2The flying target is parallel to the119883119884 plane the initialposition of flying target is (8885 4590 9100)m the velocity is300msec and the angle (120579
0) between the initial position and
the119883-axis is 30∘The trajectory of the flight vehicle is a circlewhich has a radius of 10 km The center of circular trajectoryis 119884-axis The rotational cycle time of the flight target is 209seconds and the total observation time is 419 sec
The circular trajectory equation is
X119898(119909119905 119910119905 119911119905) = (119903 cos (120579
0minus 120596119905) 119903 sin (120579
0minus 120596119905) 9100) (24)
where 119905 = 1 2 3 418 419 sec 119903 is the circular flight radiusin meter 120579
0is the initial angle between the flight vehicle
and the 119883-axis 120596 is the angular speed of flight vehicle Theaccuracy of GA based adaptive 120572-120573-120574 filter for tracking acircular flight target is shown in Figure 11 where theGAbasedadaptive 120572-120573-120574 filter converges to less than 3m after 8 secTherefore using the GA based adaptive 120572-120573-120574 filter to trackthe circular trajectory target has the larger average estimationerror and longer convergent time than linear trajectory target
Case 4 (estimating roll and pitch angles and tracking lineartrajectory target using GAPSO based adaptive 120572-120573-120574 filterand measured data (with beam pointing error)) The mea-sured roll and pitch signals shown in Figure 7 are used toevaluate the performance of GAPSO based adaptive 120572-120573-120574filter for estimating the roll and pitch angles and tracking thelinear trajectory targetThe beamof 6times6 planar array antennais steered to track the linear trajectory target The samplingfrequency is set as 10Hz Therefore the shiprsquos roll and pitchsignals are sampled 10000 points within 1000 seconds Asthe scenario described in Case 2 the ship is along the 119884
axis of 119883119884 plane in the earth coordinates the ship speedis 10msec and the flying target speed is 300msec Theflying target is parallel to the 119883119884 plane 91 km above theground and the angle between the flight direction and the119884-axis is 120∘ The largest reconnaissance distance of ship-borne radar is 150 km When the air target is flying withinthe radar detection range the beampointing error estimationand linear trajectory target tracking of ship-borne phasedarray radar are simulated
Figure 12 is a simulation flow chart used to verifythe effect of beam pointing error compensation on thetracking accuracy of adaptive 120572-120573-120574 filter Tables 2 and 3demonstrate the optimal gain parameters of adaptive 120572-120573-120574 filter for roll 119892
1199031 pitch 119892
1199011 and target tracking 119892
2
obtained by the GA and PSO methods respectively thatminimizes the RMSE Figures 13ndash17 show the estimationerrors for five different gain values which are summarizedin Table 4 It demonstrates that the greater gain value(119892 = 01 075 and 09) will generate the greater error andneed the longer convergence time The estimation errors
International Journal of Antennas and Propagation 11
Table 4 Estimation errors for different gains
Calculated by GA Calculated by PSO Fixed gain1198921199031
05392sim06904 0357sim06142 01 075 091198921199011
0622sim07756 03953sim06285 01 075 091198922
0054sim05827 007sim03259 01 075 09Average 119864
119905(m) 14781 163224 169663 534804 3991188
Table 5 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0543 05653 05424 05455 05665 056 058 0532 05414 053851198921199011
0464 0484 04459 0422 04667 04654 04509 04459 04308 046511198922
0161 00929 01937 00865 00241 00609 01174 01897 01713 01727
Y
X
Z
(a)
Y
X10 km
2730∘
(b)
Figure 10 Simulation scenarios of Case 3 for (a) 3D and (b) top view diagrams
0 10 20 30 40 50 60 70 80 90 1000
50
100
150
200
250
Time (s)
Et
(m)
Average estimation error = 24905m
Figure 11 Estimation error for circular target trajectory without beam pointing error
12 International Journal of Antennas and Propagation
Generate the target detection data
Using GA or PSO
to track the target
Generate beam steering angle
Measure or simulate roll and pitch angles
Using GA or PSO
roll and pitch angles
Simulate beam pointing error compensation
Verify tracking accuracy ofGAPSO adaptive 120572-120573-120574 filter
120572-120573-120574 filter 120572-120573-120574 filter to estimate
Figure 12 Flow chart of tracking performance simulation forCase 4experiment
0 100 200 300 400 500 600 700 800 900 10000
50
100
150
200
250
300
Time (s)
Et
(m)
Average estimation error = 14781m
Figure 13 Estimation error of GA based adaptive 120572-120573-120574 filter
presented in Figures 16 and 17 are too large to be appli-cable when the gain values are equal to 075 and 09The GA and PSO have smaller errors compared with thefixed gain values and the GA obtains the best estimationaccuracy The estimation errors of GA and PSO meth-ods were found to be close From Figures 13 14 and 15we can observe that there is a peak value that occurredat the time duration about 510 to 515 seconds becausethe horizontal beam direction of ship-borne phased array
0
50
100
150
200
250
300
Et
(m)
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 163224 m
Figure 14 Estimation error of PSO based adaptive 120572-120573-120574 filter
0
20
40
60
80
100
120
140
160
180
200Et
(m)
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 169663 m
Figure 15 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
radar is steered according to the linear trajectory targetwhich is flying through sky just above the ship at thetime instant of 510 second As shown in Figure 18 thehorizontal beam steering angle of ship-borne phased arrayradar at about the time of 510 second is changing fromdescending into ascending When the GA based adaptive120572-120573-120574 filter tracks the linear trajectory target the esti-mation error values are less than 30 meters or less ifthe peak estimation errors during these 5 seconds areignored
Case 5 (estimating roll and pitch angles and tracking lineartrajectory target using GAPSO based adaptive 120572-120573-120574 filterand simulated data (with beam pointing error)) The shiprsquosroll and pitch signals are simulated according to (3) and (4)with parameters of sea state 2 listed in Table 1 and standard
International Journal of Antennas and Propagation 13
0 200 400 600 800 10000
500
1000
1500
Time (s)
Et
(m)
Average estimation error = 534804 m
Figure 16 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0 200 400 600 800 10000
1000
2000
3000
4000
5000
6000
7000
8000
9000
Time (s)
Et
(m)
Average estimation error = 3991188 m
Figure 17 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
deviation of 001 The simulated shiprsquos roll and pitch signalsare applied for five different methods to estimate the gain ofGAPSO based adaptive 120572-120573-120574 filter The total observationtime is 1000 sec and sampling time is 01 sec The simulationreplicates 10000 times The input samples are updated byone new sample for the next iteration The other simulationscenario and parameters are the same asCase 4 Tables 5 and 6demonstrate the optimal gain parameters of adaptive 120572-120573-120574filter for roll gain 119892
1199031 pitch gain 119892
1199011 and target tracking gain
vector g2obtained by the GA and PSOmethods respectively
that minimizes RMSE Figures 19 20 21 22 and 23 showthe estimation errors for five different gain values which aresummarized in Table 7 The estimation error of GA basedadaptive 120572-120573-120574 filter for tracking a linear trajectory targetis shown in Figure 19 where the convergence time of GA
0 100 200 300 400 500 600 700 800 900 1000Time (s)
0
05
1
15
2
25
Azi
mut
h ste
erin
g an
gle (
rad)
With beam pointing errorWithout beam pointing error
Figure 18 Azimuth beam steering angle of ship-borne phased arrayradar
0
10
20
30
40
50
60
70
80
90
100
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 160756 m
Et
(m)
Figure 19 Estimation error of GA based adaptive 120572-120573-120574 filter
based adaptive 120572-120573-120574 filter is about 3 sec and the averageestimation error is about 160756m It concludes that theexperimental results in Case 5 have the same trend as inCase 4 therefore the simulated data can be used to observethe effect of shipmotion on the tracking performance of ship-borne phased array radar under different sea state conditionsThe convergence time and average estimation error of GAbased adaptive 120572-120573-120574 filter are better than themethod used in[6] where the tracking error of AEKF converges to less thanabout 20m at 550 iterations (seciteration) and the estimationerror will remain within the range of about 20m when thebeam pointing error is compensated by FBFN controller
Case 6 (estimating roll and pitch angles and tracking circularmoving target using GA based adaptive 120572-120573-120574 filter and
14 International Journal of Antennas and Propagation
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 187684 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 20 Estimation error of PSO based adaptive 120572-120573-120574 filter
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 199594 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 21 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
measured data (with beam pointing error)) As the scenariodescribed in Case 3 the measured roll and pitch signalsshown in Figure 7 are used to evaluate the performance ofGA based adaptive 120572-120573-120574 filter for estimating the roll andpitch angles and tracking the circular trajectory target Theestimation errors of 120572-120573-120574 filter with fixed gains of 119892
1199031=
1198921199011
= 1198922= 01 and 119892
1199031= 1198921199011
= 1198922= 075 are shown in
Figures 24 and 25 respectivelyThe average estimation errorsfor fixed filter gains of 01 and 075 are about 41638m and572386m respectively The optimal roll gain 119892
1199031 pitch gain
1198921199011 and target tracking gain 119892
2of adaptive 120572-120573-120574 filter
estimated by the GA method during the observation timeof 419 sec are listed in Table 8 The estimation error of GAbased adaptive 120572-120573-120574 filter for tracking a circular flight targetis shown in Figure 26 where the convergence time of GAbased adaptive 120572-120573-120574 filter is about 5 sec and the average
0
500
1000
1500
0 200 400 600 800 1000Time (s)
Average estimation error = 1080515 m
Et
(m)
Figure 22 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
0 200 400 600 800 1000Time (s)
Average estimation error = 4279449 m
Et
(m)
Figure 23 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
estimation error is about 36667m It concludes that the GAbased adaptive 120572-120573-120574 filter can greatly improve the trackingaccuracy and convergence time of the conventional 120572-120573-120574filter for tracking the circular trajectory target
5 Conclusions
This paper proposed an intelligent beam pointing error com-pensationmechanism for ship-borne phased array radarTheGA based adaptive 120572-120573-120574 filter estimates the roll and pitchangles of the ship moving in the sea and thus compensatesfor the antenna beam pointing error in order to enhance thetracking accuracy of phased array radar system
International Journal of Antennas and Propagation 15
Table 6 Estimated gain parameters for PSO based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0366 03645 0532 04519 05151 05002 04797 05254 04146 053511198921199011
0366 03647 03634 03644 03643 0363 03644 03684 03643 036421198922
0125 01271 02286 02166 01169 01799 01735 01564 00767 01346
Table 7 Estimation errors for different gains
Calculated by GA Calculated by PSO Fixed gain1198921199031
0459sim0583 03645sim05352 01 075 091198921199011
0498sim0667 0363sim03684 01 075 091198922
00586sim01697 00367sim02291 01 075 09Average 119864
119905(m) 160756 187684 199594 1080515 4279449
Table 8 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim42 43sim84 85sim126 127sim168 169sim210 211sim252 253sim294 295sim336 337sim378 379sim4191198921199031
0619 04996 05316 0644 05052 05568 05316 04459 05092 059241198921199011
0582 06727 07201 04913 04168 07169 05983 02036 05858 06211198922
0214 02215 01612 02474 01822 0157 02073 01211 02278 01074
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 41638 m
Figure 24 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
Six experimental cases are conducted to verify the perfor-mance of ship-borne phased array radar using the proposedGA based adaptive 120572-120573-120574 filter In Case 1 the GA basedadaptive 120572-120573-120574 filter is used to estimate the roll and pitchangles with the simulated and measured roll and pitchsignals respectively The simulation results show that themeasurement data has the minimum estimation error theestimation error in the sea state 2 is the second and theestimation error in the sea state 3 is the worst Cases 2 and 3show that when the GA based adaptive 120572-120573-120574 filter is appliedto estimate the beam pointing error and to track the targetin a ship-borne phased array radar the circular trajectory
0 50 100 150 200 250 300 350 400 450Time (s)
0
500
1000
1500
Average estimation error = 572386 m
Et
(m)
Figure 25 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
target will be tracked with the larger average estimation errorand longer convergence time than linear trajectory targetTheexperimental results ofCases 4 and 5demonstrate thatwhen alinear trajectory target is tracked by ship-borne phased arrayradar the average estimation error and convergence time ofship-borne phased array radar using the GA based adaptive120572-120573-120574 filter are better than PSO based adaptive 120572-120573-120574 filterand fixed filter gain 120572-120573-120574 filter The convergence time andtracking accuracy of ship-borne phased array radar usingthe proposed GA based adaptive 120572-120573-120574 filter are superiorto the AEKF significantly In conclusions the proposed GAbased adaptive 120572-120573-120574 filter is a real time applicable algorithm
16 International Journal of Antennas and Propagation
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 36667 m
Figure 26 Estimation error of GA based adaptive 120572-120573-120574 filter
whose accuracy is good with relatively less complexity Inaddition the roll and pitch signalsmeasured in real operationenvironments can be replaced with the simulated roll andpitch signals to test all the relevant properties of practical shipmotion compensation and air target tracking for ship-bornephased array radar
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
This work is supported in part by research grants fromNational Science Council China (NSC 102-2218-E-155-001)
References
[1] Z Fang and Q Rundong Ship-borne Phased Array RadarMotion Compensation System Engineering and ElectronicTechnologies 1998
[2] L J Love J F Jansen and F G Pin ldquoCompensation of waveinduced motion and force phenomenon for ship based highperformance robotic and human amplifying systemsrdquo OakRidge National Laboratory ORNLTM-2003233 2003
[3] R J Hill Motion Compensation for Ship Borne Radars andLidars US Department of Commerce National Oceanic andAtmospheric Administration Office of Oceanic and Atmo-spheric Research Earth System Research Laboratory PhysicalSciences Division 2005
[4] T I Fossen and T Perez ldquoKalman filtering for positioning andheading control of ships and offshore rigsrdquo IEEEControl SystemsMagazine vol 29 no 6 pp 32ndash46 2009
[5] S Kuchler C Pregizer J K Eberharter K Schneider and OSawodny ldquoReal-time estimation of a shiprsquos attituderdquo in Proceed-ings of theAmericanControl Conference (ACC rsquo11) pp 2411ndash2416San Francisco Calif USA July 2011
[6] J Mar K C Tsai Y T Wang and M B Basnet ldquoIntelligentmotion compensation for improving the tracking performanceof shipborne phased array radarrdquo International Journal ofAntennas and Propagation vol 2013 Article ID 384756 14pages 2013
[7] T Dirk and S Tarunraj ldquoOptimal design of 120572-120573-(120574) filtersrdquo inAmerican Control Conference vol 6 pp 4348ndash4352 ChicagoIll USA June 2000
[8] D Tenne and T Singh ldquoCharacterizing performance of 120572-120573-120574filtersrdquo IEEE Transactions on Aerospace and Electronic Systemsvol 38 no 3 pp 1072ndash1087 2002
[9] T E Lee J P Su and K W Yu ldquoParameter optimization fora third-order sampled-data trackerrdquo in Proceedings of the 2ndInternational Conference on Innovative Computing Informationand Control p 336 Kumamoto Japan September 2007
[10] Y H Lho and J H Painter ldquoA fuzzy tuned adaptive Kalmanfilterrdquo in Industrial FuzzyControl and Intelligent System pp 144ndash148 December 1993
[11] N Ramakoti A Vinay and R K Jatoth ldquoParticle swarm opti-mization aided Kalman filter for object trackingrdquo in Proceedingsof the International Conference on Advances in ComputingControl and Telecommunication Technologies (ACT rsquo09) pp 531ndash533 Kerela India December 2009
[12] T-E Lee J-P Su andK-WYu ldquoAparticle swarmoptimization-based state estimation scheme formoving objectsrdquoTransactionsof the Institute of Measurement and Control vol 34 no 2-3 pp236ndash254 2012
[13] D C Law S A McLaughlin M J Post et al ldquoAn electronicallystabilized phased array system for shipborne atmospheric windprofilingrdquo Journal of Atmospheric and Oceanic Technology vol19 no 6 pp 924ndash933 2002
[14] C A Balanis AntennaTheory Analysis and Design JohnWileyamp Sons New York NY USA 3rd edition 2005
[15] J Mar and Y R Lin ldquoImplementation of SDR digital beam-former for microsatellite SARrdquo IEEE Geoscience and RemoteSensing Letters vol 6 no 1 pp 92ndash96 2009
[16] R C Eberhart and Y Shi Computational Intelligence Conceptsto Implementations ElsevierMorgan Kaufmann 2007
[17] K Y Lee and J-B Park ldquoApplication of particle swarm optimi-zation to economic dispatch problem advantages anddisadvan-tagesrdquo in Proceedings of the IEEE PES Power Systems Conferenceand Exposition (PSCE rsquo06) pp 188ndash192 Atlanta Ga USANovember 2006
[18] R C Eberhart and J Kennedy ldquoNew optimizer using particleswarm theoryrdquo in Proceedings of the 6th International Sympo-sium onMicroMachine and Human Science pp 39ndash43 NagoyaJapan October 1995
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
International Journal of Antennas and Propagation 3
Table 1 Sea state parameters
Roll Pitch119860119877(∘) 119879
119877(sec) 119860
119875(∘) 119879
119875(sec)
Sea state 2 1 25 02 12Sea state 3 1 20 02 06
roll) is (120579119890 120601119890) and then the current antenna pointing angle
is (120579119864 120601119864) with respect to the earth coordinates
120579119864= 1205790+ 120579119890 120601
119864= 1206010+ 120601119890 (1)
If the beampointing error is predicted as (Δ120579119864 Δ120601119864) then the
beam pointing angle is corrected as
120579119864119862
= 1205790+ 120579119890minus Δ120579119864 120601
119864119862= 1206010+ 120601119890minus Δ120601119864 (2)
The value of (120579119864119862 120601119864119862) is approximate to (120579
0 1206010)
21 Ship Rotational Motion Model Ship is affected by thewaves in the ocean resulting in six degrees of freedomof movement In this paper assuming zero yaw angle thesimplified roll and pitch model [1 13] is adopted to describethe ship rotational motion in the earth coordinates Shiprsquosrotational motion is modeled with sinusoidal signal The rollangle is
120601119877 (119905) = 119860
119877sin (120596
119877119905) + 119899
119877 (119905) (3)
The pitch angle is
120579119875 (119905) = 119860
119875sin (120596
119875119905) + 119899
119875 (119905) (4)
where 119860119877and 119860
119875are the amplitude of shiprsquos roll and pitch
angles 119899119877(119905) and 119899
119875(119905) are assumed to be zeromeanGaussian
noise 119879119877and 119879
119875are the roll and pitch periods and 120596
119877=
2120587119879119877and120596
119875= 2120587119879
119875are the roll and pitch angular frequen-
ciesThe random roll and pitch angle errors caused by other
unknown factors including the change of shiprsquos travelingdirection and weather in the actual ship navigation envi-ronment will be considered into the standard deviation ofthe noise terms The amplitude and period parameters of seastates 2 and 3 are listed in Table 1 [1]
22 Coordinates Transform Since the phased array antennais installed on the ship the beam steering control employsthe ship body coordinates But the target tracking of adaptive120572-120573-120574 filter employs the Earth coordinates The Euler coor-dinates transform formula is used to convert antenna beampointing angle relative to the earth (120579
119864 120601119864) where 120579
119864is the
angle from vertical and 120601119864is the angle clockwise from true
north into antenna beam pointing angle relative to the shipbody (120579
119860 120601119860) [13]
[
[
sin 120579119860cos120601119860
sin 120579119860sin120601119860
cos 120579119860
]
]
= 119879 (120601) 119879 (120579) 119879 (120595)[
[
sin 120579119864cos120601119864
minus sin 120579119864sin120601119864
cos 120579119864
]
]
(5)
where the coordinates transform matrices are defined as
119879 (120601) =[
[
1 0 0
0 cos120601 sin1206010 minus sin120601 cos120601
]
]
119879 (120579) =[
[
cos 120579 0 sin 1205790 1 0
minus sin 120579 0 cos 120579]
]
119879 (120595) =[
[
cos120595 minus sin120595 0
sin120595 cos120595 0
0 0 1
]
]
(6)
where the ship roll angle 120601 is positive with downward roll ofstarboard the ship pitch angle 120579 is positive with bow pitchupward and the ship yaw angle 120595 is positive clockwise fromthe north
23 Planar Array Antenna An 119872 times 119873 element planararray [6] is designed for the ship-borne phased array radarsystem which includes the beamforming (BF) mode anddirection of arrival (DOA) mode The planar array canproduce multibeams in the azimuth and elevation by usingthe beamformer network (BFN) which consists of a set ofpower dividers and phase shifters The planar array usingamplitude comparison method [14] generates the differencesignal patterns for theDOAestimationThedifference signalsobtained from two neighboring beams canmeasure the DOAof the target signal of the ship-borne phased array radarsystem The beam pattern is expressed as [14 15]
119860119865 (120579 120601) =
119873
sum
119899 = 1
1198681119899[
119872
sum
119898=1
1198681198981119890119895(119898minus 1)(119896119889
119909sin 120579 cos120601+120573
119909)]
times 119890119895(119899 minus 1)(119896119889
119910sin 120579 sin120601+120573
119910)= 119878119909119898119878119910119899
(7)
where
119878119909119898
=
119872
sum
119898=1
1198681198981119890119895(119898minus 1)(119896119889
119909sin 120579 cos120601+120573
119909)
119878119910119898
=
119873
sum
119899 = 1
1198681119899119890119895(119899 minus 1)(119896119889
119910sin 120579 sin120601+120573
119910)
120573119909= minus 119896119889
119909sin 120579119905cos120601119905
120573119910= minus 119896119889
119910sin 120579119905sin120601119905
(8)
where 1198681198981
= 1198681119899= Chebyshev weighting [13]119872 = the number
of array elements in the 119909-axis 119873 = the number of arrayelements in 119910-axis 120601
119905= the azimuth steering angle of the
main beam 120579119905= the elevation steering angle of the main
beam 119889119909= the interelement spacing in 119909-axis 119889
119910= the
interelement spacing in 119910-axis 119896 = 2120587120582 = the phasepropagation constant and 120582 is the wavelength of the carrier
The phase of the RF signal at each array element isadjusted to steer the beam to the coordinates (120579
119905 120601119905)
4 International Journal of Antennas and Propagation
Estimation Prediction
Delay
Pitch gain estimator
Roll gain estimator
Estimation Prediction
Delay
Roll pitchs anglemeasurement
1205721-1205731-1205741 filter(g1r)
1205721-1205731-1205741 filter(g1p)
120579mr(k)
120579mp(k)
r(k)
r(k)
ar(k)
p(k)
p(k)
ap(k)
r(k + 1)r(k + 1)ar(k + 1)
p(k + 1)
p(k + 1)
ap(k + 1)
120579er(k)
120596er(k)
aer(k)
120579ep(k)
120596ep(k)
aep(k)
Figure 2 Parallel adaptive 120572-120573-120574 filter for beam pointing error prediction
3 Adaptive 120572-120573-120574 Filter
As shown in Figure 1 the adaptive 120572-120573-120574 filters are applied tobeam pointing error prediction and target tracking respec-tively
31 Parallel Adaptive 120572-120573-120574 Filter for Beam Pointing ErrorPrediction Figure 2 shows the beam pointing error predic-tion system which consists of separate adaptive 120572
1-1205731-1205741
filters to predict the roll and pitch angles of ship motionand their outputs are fed into the beam steering controller(BSC)The BSC compensates the beam pointing errors of thephased array antenna onboard the ship to reduce the impactof the roll and pitch motion on the phased array radar Rolland pitch data are random but they are similar in natureand they can be characterized with different amplitudes andfrequencies [1]The 120572
1-1205731-1205741filter predicts the next positions
using current error also known as innovation [8] Thisinnovation process has two steps prediction and estimation
The pitch prediction equations are expressed as
120579119901 (119896 + 1) = 120579
119890119901 (119896) + 1198791
120596119890119901 (
119896) +
1198792
1
2
119886119890119901 (
119896)
119901 (119896 + 1) = 120596
119890119901 (119896) + 1198791
119886119890119901 (
119896)
119886119901 (119896 + 1) = 119886
119890119901 (119896)
(9)
The pitch estimation equations are expressed as
120579119890119901 (
119896) =120579119901 (119896) + 1205721119901
[120579119898119901 (
119896) minus120579119901 (119896)] + 1198991119901 (
119896)
120596119890119901 (
119896) = 119901 (119896) +
1205731119901
1198791
[120579119898119901 (
119896) minus120579119901 (119896)]
119886119890119901 (
119896) = 119886119901 (119896) +
1205741119901
21198792
1
[120579119898119901 (
119896) minus120579119901 (119896)]
(10)
where 120579119901(119896 + 1) 119908
119901(119896 + 1) and 119886
119901(119896 + 1) are pitch angu-
lar position pitch angular velocity and pitch angular accel-eration values predicted at (119896 + 1)th interval respectivelyThe sampling time 119879
1= 01 sec 120579
119890119901(119896) 119908
119890119901(119896) and 119886
119890119901(119896)
are pitch angular position pitch angular velocity and pitchangular acceleration estimated at 119896th interval 120572
1119901 1205731119901 and
1205741119901
are the smoothing parameters whose region of stabilityand parameter constraint have been studied in [8] Therelationship among 120572
1119901 1205731119901 and 120574
1119901parameters can be
related to pitch gain 1198921119901 where 0 lt 119892
1119901lt 1 The formula
is described as follows
1205721119901
= 1 minus 1198923
1119901
1205731119901
= 15 sdot (1 + 1198921119901) (1 minus 119892
1119901)
2
1205741119901
= (1 minus 1198921119901)
3
(11)
The roll prediction equations roll estimation equations androll smoothing parameters have similar forms as the pitchprediction equations pitch estimation equations and pitchsmoothing parameters respectively
32 Adaptive 120572-120573-120574 Filter for Target Tracking in Three-Dimensional Space The GA based 120572
2-1205732-1205742filter algorithm
for target tracking is used by ship-borne phased array radarto predict the next target positions using current error andfurther improves its tracking accuracy The tracking processof 1205722-1205732-1205742filter is shown in Figure 3 where the prediction
equations are
X (119896 + 1) = X119890 (119896) + 1198792
V119890 (119896) +
1198792
2
2
A119890 (119896)
V (119896 + 1) = V119890 (119896) + 1198792
A119890 (119896)
A (119896 + 1) = A119890 (119896)
(12)
International Journal of Antennas and Propagation 5
Tracking gain estimator
Estimation Prediction
Positionmeasurement
Delay
1205722-1205732-1205742 filter(g2)
Xm(k)
X(k)
V(k)
A(k)
Xe(k)
Ve(k)
Ae(k)
X(k + 1)
V(k + 1)
A(k + 1)
Figure 3 Adaptive 120572-120573-120574 filter for target tracking in three-dimensional space
The estimation equations are
X119890 (119896) =
X (119896) + 1205722[X119898 (
119896) minusX (119896)] + n
2 (119896)
V119890 (119896) = V (119896) +
1205732
1198792
[X119898 (
119896) minus X (119896)]
A119890 (119896) = A (119896) +
1205742
21198792
2
[X119898 (
119896) minus X (119896)]
(13)
where X(119896) = [119909(119896) 119910(119896) 119911(119896)]119879 V(119896) = [V
119909(119896) V119910(119896)
V119911(119896)]119879 and A(119896) = [119886
119909(119896) 119886119910(119896) 119886119911(119896)]119879 are the position
vector velocity vector and acceleration vector respectivelyat time instant 119896 n
2(119896) is the zero mean white Gaussian noise
and the sampling time 1198792= 1 sec The relationship among
1205722 1205732 and 120574
2parameters can be related to gain factor vector
g2= [1198922 1198922 1198922]119879 The formula is described as follows
1205722=[
[
1205722119909
0 0
0 1205722119910
0
0 0 1205722119911
]
]
=[
[
1 minus 1198923
20 0
0 1 minus 1198923
20
0 0 1 minus 1198923
2
]
]
1205732=[
[
1205732119909
0 0
0 1205732119910
0
0 0 1205732119911
]
]
=[
[
[
15 (1 + 1198922) (1 minus 119892
2)2
0 0
0 15 (1 + 1198922) (1 minus 119892
2)2
0
0 0 15 (1 + 1198922) (1 minus 119892
2)2
]
]
]
1205742=[
[
1205742119909
0 0
0 1205742119910
0
0 0 1205742119911
]
]
=[
[
[
(1 minus 1198922)3
0 0
0 (1 minus 1198922)3
0
0 0 (1 minus 1198922)3
]
]
]
(14)
33 GA Gain Estimator The flow chart of GA rollpitch gainestimators of adaptive 120572-120573-120574 filter is shown in Figure 4 TheGA rollpitch gain estimators are used to estimate the roll andpitch angles gain (119892
11199031198921119901) of adaptive 120572
1-1205731-1205741filters The
GA tracking gain estimator is used to estimate the optimalgain g
2of adaptive 120572
2-1205732-1205742filter The flow chart of GA
tracking gain estimator is similar to Figure 4TheGAmethoduses selection crossover andmutation [16] techniques to findthe solutions At first the initial population which containsrandomly generated chromosomes is made Chromosomesare chains of 0 and 1 bits whose length is defined as stringlength Chromosomes number is defined as the populationsizeThese chromosomes from themating pool are shuffled to
create more randomness and then applied to the problem tofind their outputs and fitness scores accordingly Fitness valuedefines how well the chromosome solves the problem Twochromosomes are selected from the mating pool dependingon selection method Depending on the crossover rate allbits between two randomly chosen points selected from twodifferent chromosomes are interchanged which is known astwo-point crossover Chromosomes may undergo mutationwhere random bits of chromosomes are flipped dependingon the mutation rate Termination criteria are defined by thenumber of generations
Selection is done usually by using two methods asfollows roulette wheel selection and tournament selection
6 International Journal of Antennas and Propagation
Get optimal filter gain
Evaluate
YesNo
End
Roulette wheelselection
Crossover
Mutation
Generate the next generation
Is termination criteria met
Generate the initial population randomly
between range of 0 and 1
gr1gp1
Input the rollpitch sea state data
Figure 4 Flow chart of GA rollpitch gain estimator
In roulette wheel selection the chromosomes selected frompopulation are put into a mating pool The chromosomeselection probability is proportion to the fitness value Intournament selection chromosomes have to go throughseveral tournaments The chromosome which has the fittestvalue is selected and put into the mating pool Here theroulette wheel selection method is chosen to determine thesolution
Two-point crossover calls for two random points selectedfrom two different parent chromosome strings all bitsbetween two points are swapped for two parent chromo-somes Crossover rate is commonly chosen in a range of060 to 080 [16] Mutation rate describes the probability thata bit in a chromosome will be flipped (0 flips to 1 and 1flips to 0) Mutation brings diversity to the population asmutation of a chromosome is a random process which willbring randomness to present chromosome group Generationdefines iteration the GA method runs for a defined numberof generations and the final best solution is considered as theresult
34 PSO Gain Estimator [9 12 17] PSO method was firstproposed in [18] PSO is based on the number of particlesflying around the solution space to find the best solutionthat minimizes the cost function Particles move around thesolution space as soon as a particle detects a better solutionthe information is passed to other particles and then particleschange their position with respect to their best positionobserved among all the particles Figure 5 describes the PSOmethod in detail The PSO changes the velocity and positionof the particle according to (15) and (17) respectively toachieve the fitness [16]
V119894119889(119899 + 1)
= 119870 lowast [V119894119889(119899)
+ 1198881sdot rand ( ) sdot (119901119894119889(119899) minus 119909119894119889(119899))
+ 1198882sdot Rand ( ) sdot (119901119892119889(119899) minus 119909119894119889(119899))]
(15)
119870 =
2
1003816100381610038161003816100381610038161003816
2 minus 120593 minus radic1205932minus 4120593
1003816100381610038161003816100381610038161003816
where 120593 = 1198881+ 1198882 120593 gt 4 (16)
119909119894119889(119899 + 1)
= 119909119894119889(119899)
+ V119894119889(119899)
(17)
International Journal of Antennas and Propagation 7
Initialize particles with random positions and random velocities
For each particle calculate the fitness values
Is fitness value ltbest fitness value
Particle exhausted
Set global best fitness value as best fitness value
Update particle positions and velocities using (18) and (20)
Maximum iterationreached
Give global best fitnessvalue and position
Set global best as local best fitness value
Yes
No
No
No
Yes
Yes
Figure 5 Flow chart of PSO method
where 119909119894119889(119899)
and V119894119889(119899)
are position and velocity of (119894119889)thparticle at (119899)th iteration respectively 119901
119894119889(119899)and 119901
119892119889(119899)are
the best position and global best position of (119894119889)th particle at(119899)th iteration respectively Rand( ) and rand( ) are randomnumber between 0 and 1 respectively Equation (15) updatesthe velocity of the particles and (17) updates the position ofparticles 119870 means inertia 119888
1and 1198882are correction factors
and swarm size means number of particles for an iterationas described in [16]
Particles in the PSO method sometimes go out of con-straints while changing their position Reinitialization of theparticlersquos position randomly to fall within the constraints isdone as soon as a particle moves out of constraints Thisensures full use of particles and also increases the chance ofdiscovering a better solution To reduce the chance of gettingtrapped in a local minimum the advantages of using PSOmethod are that it is derivative-free easy to implement andeasy to comprehend and the solution is independent of theinitial point [17] One of the drawbacks of the PSOmethod isthat it does not guarantee success that is the solution foundmay not be the best solution
35 GAPSO Based Adaptive 120572-120573-120574 Filter The flow chart ofthe adaptive 120572-120573-120574 filter algorithm is shown in Figure 6where 119873 is the number of sampling data to determine thefilter gain In this algorithm the parameters120572120573 and 120574 whichare related to the filter gain 119892 are optimized by minimizingthe RMSE using GA or PSO method The proposed adaptive120572-120573-120574 filter processes the current sampled data 119909
119894which
predicts the next sample data 119909119894+1
as iteration continuesThe adaptive GA or PSO process only occurs at certain timeintervalThe previous119873 data that is 119909
119901(119894minus119873) sdot sdot sdot 119909
119901(119894minus1) is
sent to the adaptiveGAor PSO algorithmwhere the optimumvalue of filter gain (119892) is determined for the 120572-120573-120574 filterThen 119909
119894+1is predicted and the process continues The value
of119873 determines running time and accuracy of the algorithmso 119873 must have small enough value to support real timeimplementation and large enough value to provide acceptableaccuracy
Since the roll and pitch angles of ship motion changeover time randomly in order to speed up the processingtime and to reach the objectives of minimum estimationerror the filter gain values are real time updated by using thepipelined architectureThe recorded roll and pitch signals are
8 International Journal of Antennas and Propagation
Collect previous N data using PSO or GA algorithm
Predicted value for120572-120573-120574 filterxi
xi+1
ximinus1 ximinusFind g for Nximinus1 ximinusN
that minimizes RMSE
Figure 6 Block diagram of adaptive 120572-120573-120574 filter
segmented into block of 1000 sampled data (100 sec) for lineartrajectory target and 420 sampled data (42 sec) for circulartrajectory target respectivelyThe gain value of 120572
2-1205732-1205742filter
in 101ndash200 sec is estimated by GA and PSO methods usingthe collected data during 1ndash100 sec the gain value of 120572
2-
1205732-1205742filter in 201ndash300 sec is estimated by the GA and PSO
methods using the collected data in 101ndash200 sec and so onSimilarly the measured flight target signals are segmentedinto block of 100 sampled data (100 sec) for linear trajectorytarget and 42 sampled data (42 sec) for circular trajectorytarget respectively The gain value of 120572
2-1205732-1205742filter in 101ndash
200 sec is estimated by the GA and PSO methods using thecollected data during 1ndash100 sec the gain value of 120572
2-1205732-1205742
filter in the interval of 201ndash300 sec is estimated by the GA andPSO methods using the collected data in 101ndash200 sec and soon
For the PSO process it was found that gain parameterconverged within 100 iterations so the number of iterationsfor an adaptive 120572-120573-120574 filter was kept at constant 100 iterationsper processing cycle swarm size = 30 inertia = 07298 andcorrection factors 119888
1= 21 119888
2= 2 For the GA process we
have considered population size = 8 string length = 12crossover rate = 08 and mutation rate = 005
If 120579119894119901 119894 = 1 2 119873 and
120579119894119901 119894 = 1 2 119873 are real
measurement pitchrsquos values and estimated pitchrsquos values ofadaptive 120572
1-1205731-1205741filter respectively then 120579
119894119903 119894 = 1 2 119873
and 120579119894119903 119894 = 1 2 119873 are realmeasurement rollrsquos values and
estimated rollrsquos values of adaptive 1205721-1205731-1205741filter respectively
where 119894 is the current iteration and 119873 is the total number ofsampling data The RMSEs of pitch and roll estimations aredefined as follows
RMSE119901= radic
1
119873
119873
sum
119894 = 1
(120579119894119901minus 120579119894119901)
2
(18)
RMSE119903= radic
1
119873
119873
sum
119894 = 1
(120579119894119903minus 120579119894119903)
2
(19)
where the main purpose of GA and PSO methods is to findthe optimum values of 119892
1119901and 119892
1119903which minimize the
RMSE119901and RMSE
119903 respectively
If X119894and X
119894 119894 = 1 2 119873 are real measurement target
position vector values and estimated target position vector
values of adaptive 1205722-1205732-1205742filter the RMSE of target tracking
is defined as
RMSE = radic1
119873
119873
sum
119894 = 1
((119909119894minus 119909119894)2+ (119910119894minus 119910119894)2+ (119911119894minus 119894)2) (20)
where 119909119894 119910119894 and 119911
119894are the components of target position
vector X119894in 119909 119910 and 119911 axes The main purpose of GA and
PSO algorithms is to find the optimum gain value of g2which
minimizes the RMSE
4 Experimental Results
Six scenarios are used to simulate the tracking performanceof ship-borne phased array radar using the proposed adaptive120572-120573-120574 filter for beam pointing error compensation and targettracking in three-dimensional space
Case 1 (roll and pitch estimations using GA based adaptive120572-120573-120574 filter and measured data) The roll and pitch datarecorded by a gyroscope of the ship were used in the exper-iments The total number of data is 1000 and the samplingtime is 01 sec The angle variation of roll and pitch signalsmeasured by ship gyroscope are shown in Figure 7 whichwillresult in the beam pointing errors of phased array antennaTheGA based adaptive 120572-120573-120574 filter is used to estimate the rolland pitch angles Figures 8(a) and 8(b) show that the conver-gent time of roll and pitch angles are 3 sec and 5 sec respec-tively under simulated sea states 2 and 3 and measured dataThe shiprsquos roll and pitch signals are simulated according to (3)and (4) with roll and pitch angle parameters of sea states 2and 3 listed in Table 1 and standard deviation of 001 Itconcludes that the proposed GA based adaptive 120572-120573-120574 filter issuitable to compensate the ship motion As shown in Table 1the period of sea state 2 is longer than sea state 3 Thereforethe convergent RMSE (about 019∘ for roll about 01∘ forpitch) of sea state 2 is less than sea state 3 (about 029∘ for rollabout 022∘ for pitch) because the roll and pitch signals withshorter period in sea state 3 have rapid oscillations which aremore difficult to be predicted precisely
Case 2 (tracking linear moving target in three-dimensionalspace using GA based adaptive 120572-120573-120574 filter (without beampointing error)) Assuming the ship is not affected by thesea waves (antenna beam pointing error is zero) the trackingperformance of GA based adaptive 120572-120573-120574 filter for a straightflight target trajectory is simulated The initial position ofradar is (0 0 0) The ship moves with a speed of 10msec in
International Journal of Antennas and Propagation 9
0 100 200 300 400 500 600 700 800 900 1000
0
02
04
06
08
Pitc
h an
gle (
deg)
Time (10minus1 s)
minus12
minus08
minus06
minus04
minus02
minus1
(a)
0
1
2
3
Roll
angl
e (de
g)
0 100 200 300 400 500 600 700 800 900 1000Time (10minus1 s)
minus1
minus2
minus3
minus4
(b)
Figure 7 Demonstrating angle variation of (a) pitch and (b) roll signals measured by gyroscope
0
005
01
015
02
025
03
035
04
RMSE
(deg
)
Sea 3Sea 2Measurement
0 100 200 300 400 500 600 700 800 900 1000Time (10minus1 s)
(a)
0
005
01
015
02
025
03
035
04RM
SE (d
eg)
Sea 3Sea 2Measurement
Time (10minus1 s)0 100 200 300 400 500 600 700 800 900 1000
(b)
Figure 8 RMSE of (a) roll and (b) pitch under simulated data of sea states 2 and 3 and measured data
the 119884 axial direction The linear path equation of the ship isgiven by
X119904(119909119905 119910119905 119911119905) = X0119904+ V119904119905 (21)
where 119905 = 1 2 3 1000 sec V119904is ship velocity vector
and X0119904is initial ship position vector in three-dimensional
spaceThe radar position is updated every secondThe initialposition of flying target is (minus74840 minus129620 9100)m whichis about 150 km from the radar The flight speed of target is300msec (119883 axial velocity of 150msec and 119884 axial velocityof 260msec) The target location is updated every secondThe linear target trajectory equation is
X119898(119909119905 119910119905 119911119905) = X0119898
+ V119898119905 (22)
where V119898
is the target velocity vector and X0119898
is theinitial target position vector in three-dimensional space Thetracking accuracy of GA based adaptive 120572-120573-120574 filter for astraight flight target is shown in Figure 9 where the trajectoryestimation error is calculated by
119864119905= radic(119909
119905minus 119909119905)2+ (119910119905minus 119910119905)2+ (119911119905minus 119905)2 (23)
where (119909119905 119910119905 119911119905) and (119909
119905 119910119905 119905) denote the real target location
and estimated target location respectively It shows that theGA based adaptive 120572-120573-120574 filter converges to less than about2m after 5 sec
10 International Journal of Antennas and Propagation
Table 2 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0574 06399 0652 06904 06537 0674 06286 06313 06044 053921198921199011
0629 06664 0622 07756 0664 07399 0663 06652 06816 064981198922
0285 05827 03282 03851 00054 03480 02916 03736 0547 04864
Table 3 Estimated gain parameters for PSO based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0557 05986 0357 05345 06142 05774 05544 0529 04592 056261198921199011
0612 05898 04129 05675 05286 03953 05565 06199 05214 062851198922
0326 007 01136 0146 0098 02365 021 02547 02474 01245
0 20 40 60 80 100 1200
20
40
60
80
100
120
140
160
180
Time (s)
Et
(m)
Average estimation error = 15176m
Figure 9 Estimation error for linear target trajectory without beampointing error
Case 3 (tracking circular moving target in three-dimensionalspace using GA based adaptive 120572-120573-120574 filter (without beampointing error)) Figures 10(a) and 10(b) show the simulationscenario of beam pointing error compensation for circularmaneuvering air target and linear moving ship The shiphas the same linear moving speed and path equation asCase 2The flying target is parallel to the119883119884 plane the initialposition of flying target is (8885 4590 9100)m the velocity is300msec and the angle (120579
0) between the initial position and
the119883-axis is 30∘The trajectory of the flight vehicle is a circlewhich has a radius of 10 km The center of circular trajectoryis 119884-axis The rotational cycle time of the flight target is 209seconds and the total observation time is 419 sec
The circular trajectory equation is
X119898(119909119905 119910119905 119911119905) = (119903 cos (120579
0minus 120596119905) 119903 sin (120579
0minus 120596119905) 9100) (24)
where 119905 = 1 2 3 418 419 sec 119903 is the circular flight radiusin meter 120579
0is the initial angle between the flight vehicle
and the 119883-axis 120596 is the angular speed of flight vehicle Theaccuracy of GA based adaptive 120572-120573-120574 filter for tracking acircular flight target is shown in Figure 11 where theGAbasedadaptive 120572-120573-120574 filter converges to less than 3m after 8 secTherefore using the GA based adaptive 120572-120573-120574 filter to trackthe circular trajectory target has the larger average estimationerror and longer convergent time than linear trajectory target
Case 4 (estimating roll and pitch angles and tracking lineartrajectory target using GAPSO based adaptive 120572-120573-120574 filterand measured data (with beam pointing error)) The mea-sured roll and pitch signals shown in Figure 7 are used toevaluate the performance of GAPSO based adaptive 120572-120573-120574filter for estimating the roll and pitch angles and tracking thelinear trajectory targetThe beamof 6times6 planar array antennais steered to track the linear trajectory target The samplingfrequency is set as 10Hz Therefore the shiprsquos roll and pitchsignals are sampled 10000 points within 1000 seconds Asthe scenario described in Case 2 the ship is along the 119884
axis of 119883119884 plane in the earth coordinates the ship speedis 10msec and the flying target speed is 300msec Theflying target is parallel to the 119883119884 plane 91 km above theground and the angle between the flight direction and the119884-axis is 120∘ The largest reconnaissance distance of ship-borne radar is 150 km When the air target is flying withinthe radar detection range the beampointing error estimationand linear trajectory target tracking of ship-borne phasedarray radar are simulated
Figure 12 is a simulation flow chart used to verifythe effect of beam pointing error compensation on thetracking accuracy of adaptive 120572-120573-120574 filter Tables 2 and 3demonstrate the optimal gain parameters of adaptive 120572-120573-120574 filter for roll 119892
1199031 pitch 119892
1199011 and target tracking 119892
2
obtained by the GA and PSO methods respectively thatminimizes the RMSE Figures 13ndash17 show the estimationerrors for five different gain values which are summarizedin Table 4 It demonstrates that the greater gain value(119892 = 01 075 and 09) will generate the greater error andneed the longer convergence time The estimation errors
International Journal of Antennas and Propagation 11
Table 4 Estimation errors for different gains
Calculated by GA Calculated by PSO Fixed gain1198921199031
05392sim06904 0357sim06142 01 075 091198921199011
0622sim07756 03953sim06285 01 075 091198922
0054sim05827 007sim03259 01 075 09Average 119864
119905(m) 14781 163224 169663 534804 3991188
Table 5 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0543 05653 05424 05455 05665 056 058 0532 05414 053851198921199011
0464 0484 04459 0422 04667 04654 04509 04459 04308 046511198922
0161 00929 01937 00865 00241 00609 01174 01897 01713 01727
Y
X
Z
(a)
Y
X10 km
2730∘
(b)
Figure 10 Simulation scenarios of Case 3 for (a) 3D and (b) top view diagrams
0 10 20 30 40 50 60 70 80 90 1000
50
100
150
200
250
Time (s)
Et
(m)
Average estimation error = 24905m
Figure 11 Estimation error for circular target trajectory without beam pointing error
12 International Journal of Antennas and Propagation
Generate the target detection data
Using GA or PSO
to track the target
Generate beam steering angle
Measure or simulate roll and pitch angles
Using GA or PSO
roll and pitch angles
Simulate beam pointing error compensation
Verify tracking accuracy ofGAPSO adaptive 120572-120573-120574 filter
120572-120573-120574 filter 120572-120573-120574 filter to estimate
Figure 12 Flow chart of tracking performance simulation forCase 4experiment
0 100 200 300 400 500 600 700 800 900 10000
50
100
150
200
250
300
Time (s)
Et
(m)
Average estimation error = 14781m
Figure 13 Estimation error of GA based adaptive 120572-120573-120574 filter
presented in Figures 16 and 17 are too large to be appli-cable when the gain values are equal to 075 and 09The GA and PSO have smaller errors compared with thefixed gain values and the GA obtains the best estimationaccuracy The estimation errors of GA and PSO meth-ods were found to be close From Figures 13 14 and 15we can observe that there is a peak value that occurredat the time duration about 510 to 515 seconds becausethe horizontal beam direction of ship-borne phased array
0
50
100
150
200
250
300
Et
(m)
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 163224 m
Figure 14 Estimation error of PSO based adaptive 120572-120573-120574 filter
0
20
40
60
80
100
120
140
160
180
200Et
(m)
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 169663 m
Figure 15 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
radar is steered according to the linear trajectory targetwhich is flying through sky just above the ship at thetime instant of 510 second As shown in Figure 18 thehorizontal beam steering angle of ship-borne phased arrayradar at about the time of 510 second is changing fromdescending into ascending When the GA based adaptive120572-120573-120574 filter tracks the linear trajectory target the esti-mation error values are less than 30 meters or less ifthe peak estimation errors during these 5 seconds areignored
Case 5 (estimating roll and pitch angles and tracking lineartrajectory target using GAPSO based adaptive 120572-120573-120574 filterand simulated data (with beam pointing error)) The shiprsquosroll and pitch signals are simulated according to (3) and (4)with parameters of sea state 2 listed in Table 1 and standard
International Journal of Antennas and Propagation 13
0 200 400 600 800 10000
500
1000
1500
Time (s)
Et
(m)
Average estimation error = 534804 m
Figure 16 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0 200 400 600 800 10000
1000
2000
3000
4000
5000
6000
7000
8000
9000
Time (s)
Et
(m)
Average estimation error = 3991188 m
Figure 17 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
deviation of 001 The simulated shiprsquos roll and pitch signalsare applied for five different methods to estimate the gain ofGAPSO based adaptive 120572-120573-120574 filter The total observationtime is 1000 sec and sampling time is 01 sec The simulationreplicates 10000 times The input samples are updated byone new sample for the next iteration The other simulationscenario and parameters are the same asCase 4 Tables 5 and 6demonstrate the optimal gain parameters of adaptive 120572-120573-120574filter for roll gain 119892
1199031 pitch gain 119892
1199011 and target tracking gain
vector g2obtained by the GA and PSOmethods respectively
that minimizes RMSE Figures 19 20 21 22 and 23 showthe estimation errors for five different gain values which aresummarized in Table 7 The estimation error of GA basedadaptive 120572-120573-120574 filter for tracking a linear trajectory targetis shown in Figure 19 where the convergence time of GA
0 100 200 300 400 500 600 700 800 900 1000Time (s)
0
05
1
15
2
25
Azi
mut
h ste
erin
g an
gle (
rad)
With beam pointing errorWithout beam pointing error
Figure 18 Azimuth beam steering angle of ship-borne phased arrayradar
0
10
20
30
40
50
60
70
80
90
100
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 160756 m
Et
(m)
Figure 19 Estimation error of GA based adaptive 120572-120573-120574 filter
based adaptive 120572-120573-120574 filter is about 3 sec and the averageestimation error is about 160756m It concludes that theexperimental results in Case 5 have the same trend as inCase 4 therefore the simulated data can be used to observethe effect of shipmotion on the tracking performance of ship-borne phased array radar under different sea state conditionsThe convergence time and average estimation error of GAbased adaptive 120572-120573-120574 filter are better than themethod used in[6] where the tracking error of AEKF converges to less thanabout 20m at 550 iterations (seciteration) and the estimationerror will remain within the range of about 20m when thebeam pointing error is compensated by FBFN controller
Case 6 (estimating roll and pitch angles and tracking circularmoving target using GA based adaptive 120572-120573-120574 filter and
14 International Journal of Antennas and Propagation
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 187684 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 20 Estimation error of PSO based adaptive 120572-120573-120574 filter
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 199594 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 21 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
measured data (with beam pointing error)) As the scenariodescribed in Case 3 the measured roll and pitch signalsshown in Figure 7 are used to evaluate the performance ofGA based adaptive 120572-120573-120574 filter for estimating the roll andpitch angles and tracking the circular trajectory target Theestimation errors of 120572-120573-120574 filter with fixed gains of 119892
1199031=
1198921199011
= 1198922= 01 and 119892
1199031= 1198921199011
= 1198922= 075 are shown in
Figures 24 and 25 respectivelyThe average estimation errorsfor fixed filter gains of 01 and 075 are about 41638m and572386m respectively The optimal roll gain 119892
1199031 pitch gain
1198921199011 and target tracking gain 119892
2of adaptive 120572-120573-120574 filter
estimated by the GA method during the observation timeof 419 sec are listed in Table 8 The estimation error of GAbased adaptive 120572-120573-120574 filter for tracking a circular flight targetis shown in Figure 26 where the convergence time of GAbased adaptive 120572-120573-120574 filter is about 5 sec and the average
0
500
1000
1500
0 200 400 600 800 1000Time (s)
Average estimation error = 1080515 m
Et
(m)
Figure 22 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
0 200 400 600 800 1000Time (s)
Average estimation error = 4279449 m
Et
(m)
Figure 23 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
estimation error is about 36667m It concludes that the GAbased adaptive 120572-120573-120574 filter can greatly improve the trackingaccuracy and convergence time of the conventional 120572-120573-120574filter for tracking the circular trajectory target
5 Conclusions
This paper proposed an intelligent beam pointing error com-pensationmechanism for ship-borne phased array radarTheGA based adaptive 120572-120573-120574 filter estimates the roll and pitchangles of the ship moving in the sea and thus compensatesfor the antenna beam pointing error in order to enhance thetracking accuracy of phased array radar system
International Journal of Antennas and Propagation 15
Table 6 Estimated gain parameters for PSO based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0366 03645 0532 04519 05151 05002 04797 05254 04146 053511198921199011
0366 03647 03634 03644 03643 0363 03644 03684 03643 036421198922
0125 01271 02286 02166 01169 01799 01735 01564 00767 01346
Table 7 Estimation errors for different gains
Calculated by GA Calculated by PSO Fixed gain1198921199031
0459sim0583 03645sim05352 01 075 091198921199011
0498sim0667 0363sim03684 01 075 091198922
00586sim01697 00367sim02291 01 075 09Average 119864
119905(m) 160756 187684 199594 1080515 4279449
Table 8 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim42 43sim84 85sim126 127sim168 169sim210 211sim252 253sim294 295sim336 337sim378 379sim4191198921199031
0619 04996 05316 0644 05052 05568 05316 04459 05092 059241198921199011
0582 06727 07201 04913 04168 07169 05983 02036 05858 06211198922
0214 02215 01612 02474 01822 0157 02073 01211 02278 01074
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 41638 m
Figure 24 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
Six experimental cases are conducted to verify the perfor-mance of ship-borne phased array radar using the proposedGA based adaptive 120572-120573-120574 filter In Case 1 the GA basedadaptive 120572-120573-120574 filter is used to estimate the roll and pitchangles with the simulated and measured roll and pitchsignals respectively The simulation results show that themeasurement data has the minimum estimation error theestimation error in the sea state 2 is the second and theestimation error in the sea state 3 is the worst Cases 2 and 3show that when the GA based adaptive 120572-120573-120574 filter is appliedto estimate the beam pointing error and to track the targetin a ship-borne phased array radar the circular trajectory
0 50 100 150 200 250 300 350 400 450Time (s)
0
500
1000
1500
Average estimation error = 572386 m
Et
(m)
Figure 25 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
target will be tracked with the larger average estimation errorand longer convergence time than linear trajectory targetTheexperimental results ofCases 4 and 5demonstrate thatwhen alinear trajectory target is tracked by ship-borne phased arrayradar the average estimation error and convergence time ofship-borne phased array radar using the GA based adaptive120572-120573-120574 filter are better than PSO based adaptive 120572-120573-120574 filterand fixed filter gain 120572-120573-120574 filter The convergence time andtracking accuracy of ship-borne phased array radar usingthe proposed GA based adaptive 120572-120573-120574 filter are superiorto the AEKF significantly In conclusions the proposed GAbased adaptive 120572-120573-120574 filter is a real time applicable algorithm
16 International Journal of Antennas and Propagation
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 36667 m
Figure 26 Estimation error of GA based adaptive 120572-120573-120574 filter
whose accuracy is good with relatively less complexity Inaddition the roll and pitch signalsmeasured in real operationenvironments can be replaced with the simulated roll andpitch signals to test all the relevant properties of practical shipmotion compensation and air target tracking for ship-bornephased array radar
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
This work is supported in part by research grants fromNational Science Council China (NSC 102-2218-E-155-001)
References
[1] Z Fang and Q Rundong Ship-borne Phased Array RadarMotion Compensation System Engineering and ElectronicTechnologies 1998
[2] L J Love J F Jansen and F G Pin ldquoCompensation of waveinduced motion and force phenomenon for ship based highperformance robotic and human amplifying systemsrdquo OakRidge National Laboratory ORNLTM-2003233 2003
[3] R J Hill Motion Compensation for Ship Borne Radars andLidars US Department of Commerce National Oceanic andAtmospheric Administration Office of Oceanic and Atmo-spheric Research Earth System Research Laboratory PhysicalSciences Division 2005
[4] T I Fossen and T Perez ldquoKalman filtering for positioning andheading control of ships and offshore rigsrdquo IEEEControl SystemsMagazine vol 29 no 6 pp 32ndash46 2009
[5] S Kuchler C Pregizer J K Eberharter K Schneider and OSawodny ldquoReal-time estimation of a shiprsquos attituderdquo in Proceed-ings of theAmericanControl Conference (ACC rsquo11) pp 2411ndash2416San Francisco Calif USA July 2011
[6] J Mar K C Tsai Y T Wang and M B Basnet ldquoIntelligentmotion compensation for improving the tracking performanceof shipborne phased array radarrdquo International Journal ofAntennas and Propagation vol 2013 Article ID 384756 14pages 2013
[7] T Dirk and S Tarunraj ldquoOptimal design of 120572-120573-(120574) filtersrdquo inAmerican Control Conference vol 6 pp 4348ndash4352 ChicagoIll USA June 2000
[8] D Tenne and T Singh ldquoCharacterizing performance of 120572-120573-120574filtersrdquo IEEE Transactions on Aerospace and Electronic Systemsvol 38 no 3 pp 1072ndash1087 2002
[9] T E Lee J P Su and K W Yu ldquoParameter optimization fora third-order sampled-data trackerrdquo in Proceedings of the 2ndInternational Conference on Innovative Computing Informationand Control p 336 Kumamoto Japan September 2007
[10] Y H Lho and J H Painter ldquoA fuzzy tuned adaptive Kalmanfilterrdquo in Industrial FuzzyControl and Intelligent System pp 144ndash148 December 1993
[11] N Ramakoti A Vinay and R K Jatoth ldquoParticle swarm opti-mization aided Kalman filter for object trackingrdquo in Proceedingsof the International Conference on Advances in ComputingControl and Telecommunication Technologies (ACT rsquo09) pp 531ndash533 Kerela India December 2009
[12] T-E Lee J-P Su andK-WYu ldquoAparticle swarmoptimization-based state estimation scheme formoving objectsrdquoTransactionsof the Institute of Measurement and Control vol 34 no 2-3 pp236ndash254 2012
[13] D C Law S A McLaughlin M J Post et al ldquoAn electronicallystabilized phased array system for shipborne atmospheric windprofilingrdquo Journal of Atmospheric and Oceanic Technology vol19 no 6 pp 924ndash933 2002
[14] C A Balanis AntennaTheory Analysis and Design JohnWileyamp Sons New York NY USA 3rd edition 2005
[15] J Mar and Y R Lin ldquoImplementation of SDR digital beam-former for microsatellite SARrdquo IEEE Geoscience and RemoteSensing Letters vol 6 no 1 pp 92ndash96 2009
[16] R C Eberhart and Y Shi Computational Intelligence Conceptsto Implementations ElsevierMorgan Kaufmann 2007
[17] K Y Lee and J-B Park ldquoApplication of particle swarm optimi-zation to economic dispatch problem advantages anddisadvan-tagesrdquo in Proceedings of the IEEE PES Power Systems Conferenceand Exposition (PSCE rsquo06) pp 188ndash192 Atlanta Ga USANovember 2006
[18] R C Eberhart and J Kennedy ldquoNew optimizer using particleswarm theoryrdquo in Proceedings of the 6th International Sympo-sium onMicroMachine and Human Science pp 39ndash43 NagoyaJapan October 1995
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
4 International Journal of Antennas and Propagation
Estimation Prediction
Delay
Pitch gain estimator
Roll gain estimator
Estimation Prediction
Delay
Roll pitchs anglemeasurement
1205721-1205731-1205741 filter(g1r)
1205721-1205731-1205741 filter(g1p)
120579mr(k)
120579mp(k)
r(k)
r(k)
ar(k)
p(k)
p(k)
ap(k)
r(k + 1)r(k + 1)ar(k + 1)
p(k + 1)
p(k + 1)
ap(k + 1)
120579er(k)
120596er(k)
aer(k)
120579ep(k)
120596ep(k)
aep(k)
Figure 2 Parallel adaptive 120572-120573-120574 filter for beam pointing error prediction
3 Adaptive 120572-120573-120574 Filter
As shown in Figure 1 the adaptive 120572-120573-120574 filters are applied tobeam pointing error prediction and target tracking respec-tively
31 Parallel Adaptive 120572-120573-120574 Filter for Beam Pointing ErrorPrediction Figure 2 shows the beam pointing error predic-tion system which consists of separate adaptive 120572
1-1205731-1205741
filters to predict the roll and pitch angles of ship motionand their outputs are fed into the beam steering controller(BSC)The BSC compensates the beam pointing errors of thephased array antenna onboard the ship to reduce the impactof the roll and pitch motion on the phased array radar Rolland pitch data are random but they are similar in natureand they can be characterized with different amplitudes andfrequencies [1]The 120572
1-1205731-1205741filter predicts the next positions
using current error also known as innovation [8] Thisinnovation process has two steps prediction and estimation
The pitch prediction equations are expressed as
120579119901 (119896 + 1) = 120579
119890119901 (119896) + 1198791
120596119890119901 (
119896) +
1198792
1
2
119886119890119901 (
119896)
119901 (119896 + 1) = 120596
119890119901 (119896) + 1198791
119886119890119901 (
119896)
119886119901 (119896 + 1) = 119886
119890119901 (119896)
(9)
The pitch estimation equations are expressed as
120579119890119901 (
119896) =120579119901 (119896) + 1205721119901
[120579119898119901 (
119896) minus120579119901 (119896)] + 1198991119901 (
119896)
120596119890119901 (
119896) = 119901 (119896) +
1205731119901
1198791
[120579119898119901 (
119896) minus120579119901 (119896)]
119886119890119901 (
119896) = 119886119901 (119896) +
1205741119901
21198792
1
[120579119898119901 (
119896) minus120579119901 (119896)]
(10)
where 120579119901(119896 + 1) 119908
119901(119896 + 1) and 119886
119901(119896 + 1) are pitch angu-
lar position pitch angular velocity and pitch angular accel-eration values predicted at (119896 + 1)th interval respectivelyThe sampling time 119879
1= 01 sec 120579
119890119901(119896) 119908
119890119901(119896) and 119886
119890119901(119896)
are pitch angular position pitch angular velocity and pitchangular acceleration estimated at 119896th interval 120572
1119901 1205731119901 and
1205741119901
are the smoothing parameters whose region of stabilityand parameter constraint have been studied in [8] Therelationship among 120572
1119901 1205731119901 and 120574
1119901parameters can be
related to pitch gain 1198921119901 where 0 lt 119892
1119901lt 1 The formula
is described as follows
1205721119901
= 1 minus 1198923
1119901
1205731119901
= 15 sdot (1 + 1198921119901) (1 minus 119892
1119901)
2
1205741119901
= (1 minus 1198921119901)
3
(11)
The roll prediction equations roll estimation equations androll smoothing parameters have similar forms as the pitchprediction equations pitch estimation equations and pitchsmoothing parameters respectively
32 Adaptive 120572-120573-120574 Filter for Target Tracking in Three-Dimensional Space The GA based 120572
2-1205732-1205742filter algorithm
for target tracking is used by ship-borne phased array radarto predict the next target positions using current error andfurther improves its tracking accuracy The tracking processof 1205722-1205732-1205742filter is shown in Figure 3 where the prediction
equations are
X (119896 + 1) = X119890 (119896) + 1198792
V119890 (119896) +
1198792
2
2
A119890 (119896)
V (119896 + 1) = V119890 (119896) + 1198792
A119890 (119896)
A (119896 + 1) = A119890 (119896)
(12)
International Journal of Antennas and Propagation 5
Tracking gain estimator
Estimation Prediction
Positionmeasurement
Delay
1205722-1205732-1205742 filter(g2)
Xm(k)
X(k)
V(k)
A(k)
Xe(k)
Ve(k)
Ae(k)
X(k + 1)
V(k + 1)
A(k + 1)
Figure 3 Adaptive 120572-120573-120574 filter for target tracking in three-dimensional space
The estimation equations are
X119890 (119896) =
X (119896) + 1205722[X119898 (
119896) minusX (119896)] + n
2 (119896)
V119890 (119896) = V (119896) +
1205732
1198792
[X119898 (
119896) minus X (119896)]
A119890 (119896) = A (119896) +
1205742
21198792
2
[X119898 (
119896) minus X (119896)]
(13)
where X(119896) = [119909(119896) 119910(119896) 119911(119896)]119879 V(119896) = [V
119909(119896) V119910(119896)
V119911(119896)]119879 and A(119896) = [119886
119909(119896) 119886119910(119896) 119886119911(119896)]119879 are the position
vector velocity vector and acceleration vector respectivelyat time instant 119896 n
2(119896) is the zero mean white Gaussian noise
and the sampling time 1198792= 1 sec The relationship among
1205722 1205732 and 120574
2parameters can be related to gain factor vector
g2= [1198922 1198922 1198922]119879 The formula is described as follows
1205722=[
[
1205722119909
0 0
0 1205722119910
0
0 0 1205722119911
]
]
=[
[
1 minus 1198923
20 0
0 1 minus 1198923
20
0 0 1 minus 1198923
2
]
]
1205732=[
[
1205732119909
0 0
0 1205732119910
0
0 0 1205732119911
]
]
=[
[
[
15 (1 + 1198922) (1 minus 119892
2)2
0 0
0 15 (1 + 1198922) (1 minus 119892
2)2
0
0 0 15 (1 + 1198922) (1 minus 119892
2)2
]
]
]
1205742=[
[
1205742119909
0 0
0 1205742119910
0
0 0 1205742119911
]
]
=[
[
[
(1 minus 1198922)3
0 0
0 (1 minus 1198922)3
0
0 0 (1 minus 1198922)3
]
]
]
(14)
33 GA Gain Estimator The flow chart of GA rollpitch gainestimators of adaptive 120572-120573-120574 filter is shown in Figure 4 TheGA rollpitch gain estimators are used to estimate the roll andpitch angles gain (119892
11199031198921119901) of adaptive 120572
1-1205731-1205741filters The
GA tracking gain estimator is used to estimate the optimalgain g
2of adaptive 120572
2-1205732-1205742filter The flow chart of GA
tracking gain estimator is similar to Figure 4TheGAmethoduses selection crossover andmutation [16] techniques to findthe solutions At first the initial population which containsrandomly generated chromosomes is made Chromosomesare chains of 0 and 1 bits whose length is defined as stringlength Chromosomes number is defined as the populationsizeThese chromosomes from themating pool are shuffled to
create more randomness and then applied to the problem tofind their outputs and fitness scores accordingly Fitness valuedefines how well the chromosome solves the problem Twochromosomes are selected from the mating pool dependingon selection method Depending on the crossover rate allbits between two randomly chosen points selected from twodifferent chromosomes are interchanged which is known astwo-point crossover Chromosomes may undergo mutationwhere random bits of chromosomes are flipped dependingon the mutation rate Termination criteria are defined by thenumber of generations
Selection is done usually by using two methods asfollows roulette wheel selection and tournament selection
6 International Journal of Antennas and Propagation
Get optimal filter gain
Evaluate
YesNo
End
Roulette wheelselection
Crossover
Mutation
Generate the next generation
Is termination criteria met
Generate the initial population randomly
between range of 0 and 1
gr1gp1
Input the rollpitch sea state data
Figure 4 Flow chart of GA rollpitch gain estimator
In roulette wheel selection the chromosomes selected frompopulation are put into a mating pool The chromosomeselection probability is proportion to the fitness value Intournament selection chromosomes have to go throughseveral tournaments The chromosome which has the fittestvalue is selected and put into the mating pool Here theroulette wheel selection method is chosen to determine thesolution
Two-point crossover calls for two random points selectedfrom two different parent chromosome strings all bitsbetween two points are swapped for two parent chromo-somes Crossover rate is commonly chosen in a range of060 to 080 [16] Mutation rate describes the probability thata bit in a chromosome will be flipped (0 flips to 1 and 1flips to 0) Mutation brings diversity to the population asmutation of a chromosome is a random process which willbring randomness to present chromosome group Generationdefines iteration the GA method runs for a defined numberof generations and the final best solution is considered as theresult
34 PSO Gain Estimator [9 12 17] PSO method was firstproposed in [18] PSO is based on the number of particlesflying around the solution space to find the best solutionthat minimizes the cost function Particles move around thesolution space as soon as a particle detects a better solutionthe information is passed to other particles and then particleschange their position with respect to their best positionobserved among all the particles Figure 5 describes the PSOmethod in detail The PSO changes the velocity and positionof the particle according to (15) and (17) respectively toachieve the fitness [16]
V119894119889(119899 + 1)
= 119870 lowast [V119894119889(119899)
+ 1198881sdot rand ( ) sdot (119901119894119889(119899) minus 119909119894119889(119899))
+ 1198882sdot Rand ( ) sdot (119901119892119889(119899) minus 119909119894119889(119899))]
(15)
119870 =
2
1003816100381610038161003816100381610038161003816
2 minus 120593 minus radic1205932minus 4120593
1003816100381610038161003816100381610038161003816
where 120593 = 1198881+ 1198882 120593 gt 4 (16)
119909119894119889(119899 + 1)
= 119909119894119889(119899)
+ V119894119889(119899)
(17)
International Journal of Antennas and Propagation 7
Initialize particles with random positions and random velocities
For each particle calculate the fitness values
Is fitness value ltbest fitness value
Particle exhausted
Set global best fitness value as best fitness value
Update particle positions and velocities using (18) and (20)
Maximum iterationreached
Give global best fitnessvalue and position
Set global best as local best fitness value
Yes
No
No
No
Yes
Yes
Figure 5 Flow chart of PSO method
where 119909119894119889(119899)
and V119894119889(119899)
are position and velocity of (119894119889)thparticle at (119899)th iteration respectively 119901
119894119889(119899)and 119901
119892119889(119899)are
the best position and global best position of (119894119889)th particle at(119899)th iteration respectively Rand( ) and rand( ) are randomnumber between 0 and 1 respectively Equation (15) updatesthe velocity of the particles and (17) updates the position ofparticles 119870 means inertia 119888
1and 1198882are correction factors
and swarm size means number of particles for an iterationas described in [16]
Particles in the PSO method sometimes go out of con-straints while changing their position Reinitialization of theparticlersquos position randomly to fall within the constraints isdone as soon as a particle moves out of constraints Thisensures full use of particles and also increases the chance ofdiscovering a better solution To reduce the chance of gettingtrapped in a local minimum the advantages of using PSOmethod are that it is derivative-free easy to implement andeasy to comprehend and the solution is independent of theinitial point [17] One of the drawbacks of the PSOmethod isthat it does not guarantee success that is the solution foundmay not be the best solution
35 GAPSO Based Adaptive 120572-120573-120574 Filter The flow chart ofthe adaptive 120572-120573-120574 filter algorithm is shown in Figure 6where 119873 is the number of sampling data to determine thefilter gain In this algorithm the parameters120572120573 and 120574 whichare related to the filter gain 119892 are optimized by minimizingthe RMSE using GA or PSO method The proposed adaptive120572-120573-120574 filter processes the current sampled data 119909
119894which
predicts the next sample data 119909119894+1
as iteration continuesThe adaptive GA or PSO process only occurs at certain timeintervalThe previous119873 data that is 119909
119901(119894minus119873) sdot sdot sdot 119909
119901(119894minus1) is
sent to the adaptiveGAor PSO algorithmwhere the optimumvalue of filter gain (119892) is determined for the 120572-120573-120574 filterThen 119909
119894+1is predicted and the process continues The value
of119873 determines running time and accuracy of the algorithmso 119873 must have small enough value to support real timeimplementation and large enough value to provide acceptableaccuracy
Since the roll and pitch angles of ship motion changeover time randomly in order to speed up the processingtime and to reach the objectives of minimum estimationerror the filter gain values are real time updated by using thepipelined architectureThe recorded roll and pitch signals are
8 International Journal of Antennas and Propagation
Collect previous N data using PSO or GA algorithm
Predicted value for120572-120573-120574 filterxi
xi+1
ximinus1 ximinusFind g for Nximinus1 ximinusN
that minimizes RMSE
Figure 6 Block diagram of adaptive 120572-120573-120574 filter
segmented into block of 1000 sampled data (100 sec) for lineartrajectory target and 420 sampled data (42 sec) for circulartrajectory target respectivelyThe gain value of 120572
2-1205732-1205742filter
in 101ndash200 sec is estimated by GA and PSO methods usingthe collected data during 1ndash100 sec the gain value of 120572
2-
1205732-1205742filter in 201ndash300 sec is estimated by the GA and PSO
methods using the collected data in 101ndash200 sec and so onSimilarly the measured flight target signals are segmentedinto block of 100 sampled data (100 sec) for linear trajectorytarget and 42 sampled data (42 sec) for circular trajectorytarget respectively The gain value of 120572
2-1205732-1205742filter in 101ndash
200 sec is estimated by the GA and PSO methods using thecollected data during 1ndash100 sec the gain value of 120572
2-1205732-1205742
filter in the interval of 201ndash300 sec is estimated by the GA andPSO methods using the collected data in 101ndash200 sec and soon
For the PSO process it was found that gain parameterconverged within 100 iterations so the number of iterationsfor an adaptive 120572-120573-120574 filter was kept at constant 100 iterationsper processing cycle swarm size = 30 inertia = 07298 andcorrection factors 119888
1= 21 119888
2= 2 For the GA process we
have considered population size = 8 string length = 12crossover rate = 08 and mutation rate = 005
If 120579119894119901 119894 = 1 2 119873 and
120579119894119901 119894 = 1 2 119873 are real
measurement pitchrsquos values and estimated pitchrsquos values ofadaptive 120572
1-1205731-1205741filter respectively then 120579
119894119903 119894 = 1 2 119873
and 120579119894119903 119894 = 1 2 119873 are realmeasurement rollrsquos values and
estimated rollrsquos values of adaptive 1205721-1205731-1205741filter respectively
where 119894 is the current iteration and 119873 is the total number ofsampling data The RMSEs of pitch and roll estimations aredefined as follows
RMSE119901= radic
1
119873
119873
sum
119894 = 1
(120579119894119901minus 120579119894119901)
2
(18)
RMSE119903= radic
1
119873
119873
sum
119894 = 1
(120579119894119903minus 120579119894119903)
2
(19)
where the main purpose of GA and PSO methods is to findthe optimum values of 119892
1119901and 119892
1119903which minimize the
RMSE119901and RMSE
119903 respectively
If X119894and X
119894 119894 = 1 2 119873 are real measurement target
position vector values and estimated target position vector
values of adaptive 1205722-1205732-1205742filter the RMSE of target tracking
is defined as
RMSE = radic1
119873
119873
sum
119894 = 1
((119909119894minus 119909119894)2+ (119910119894minus 119910119894)2+ (119911119894minus 119894)2) (20)
where 119909119894 119910119894 and 119911
119894are the components of target position
vector X119894in 119909 119910 and 119911 axes The main purpose of GA and
PSO algorithms is to find the optimum gain value of g2which
minimizes the RMSE
4 Experimental Results
Six scenarios are used to simulate the tracking performanceof ship-borne phased array radar using the proposed adaptive120572-120573-120574 filter for beam pointing error compensation and targettracking in three-dimensional space
Case 1 (roll and pitch estimations using GA based adaptive120572-120573-120574 filter and measured data) The roll and pitch datarecorded by a gyroscope of the ship were used in the exper-iments The total number of data is 1000 and the samplingtime is 01 sec The angle variation of roll and pitch signalsmeasured by ship gyroscope are shown in Figure 7 whichwillresult in the beam pointing errors of phased array antennaTheGA based adaptive 120572-120573-120574 filter is used to estimate the rolland pitch angles Figures 8(a) and 8(b) show that the conver-gent time of roll and pitch angles are 3 sec and 5 sec respec-tively under simulated sea states 2 and 3 and measured dataThe shiprsquos roll and pitch signals are simulated according to (3)and (4) with roll and pitch angle parameters of sea states 2and 3 listed in Table 1 and standard deviation of 001 Itconcludes that the proposed GA based adaptive 120572-120573-120574 filter issuitable to compensate the ship motion As shown in Table 1the period of sea state 2 is longer than sea state 3 Thereforethe convergent RMSE (about 019∘ for roll about 01∘ forpitch) of sea state 2 is less than sea state 3 (about 029∘ for rollabout 022∘ for pitch) because the roll and pitch signals withshorter period in sea state 3 have rapid oscillations which aremore difficult to be predicted precisely
Case 2 (tracking linear moving target in three-dimensionalspace using GA based adaptive 120572-120573-120574 filter (without beampointing error)) Assuming the ship is not affected by thesea waves (antenna beam pointing error is zero) the trackingperformance of GA based adaptive 120572-120573-120574 filter for a straightflight target trajectory is simulated The initial position ofradar is (0 0 0) The ship moves with a speed of 10msec in
International Journal of Antennas and Propagation 9
0 100 200 300 400 500 600 700 800 900 1000
0
02
04
06
08
Pitc
h an
gle (
deg)
Time (10minus1 s)
minus12
minus08
minus06
minus04
minus02
minus1
(a)
0
1
2
3
Roll
angl
e (de
g)
0 100 200 300 400 500 600 700 800 900 1000Time (10minus1 s)
minus1
minus2
minus3
minus4
(b)
Figure 7 Demonstrating angle variation of (a) pitch and (b) roll signals measured by gyroscope
0
005
01
015
02
025
03
035
04
RMSE
(deg
)
Sea 3Sea 2Measurement
0 100 200 300 400 500 600 700 800 900 1000Time (10minus1 s)
(a)
0
005
01
015
02
025
03
035
04RM
SE (d
eg)
Sea 3Sea 2Measurement
Time (10minus1 s)0 100 200 300 400 500 600 700 800 900 1000
(b)
Figure 8 RMSE of (a) roll and (b) pitch under simulated data of sea states 2 and 3 and measured data
the 119884 axial direction The linear path equation of the ship isgiven by
X119904(119909119905 119910119905 119911119905) = X0119904+ V119904119905 (21)
where 119905 = 1 2 3 1000 sec V119904is ship velocity vector
and X0119904is initial ship position vector in three-dimensional
spaceThe radar position is updated every secondThe initialposition of flying target is (minus74840 minus129620 9100)m whichis about 150 km from the radar The flight speed of target is300msec (119883 axial velocity of 150msec and 119884 axial velocityof 260msec) The target location is updated every secondThe linear target trajectory equation is
X119898(119909119905 119910119905 119911119905) = X0119898
+ V119898119905 (22)
where V119898
is the target velocity vector and X0119898
is theinitial target position vector in three-dimensional space Thetracking accuracy of GA based adaptive 120572-120573-120574 filter for astraight flight target is shown in Figure 9 where the trajectoryestimation error is calculated by
119864119905= radic(119909
119905minus 119909119905)2+ (119910119905minus 119910119905)2+ (119911119905minus 119905)2 (23)
where (119909119905 119910119905 119911119905) and (119909
119905 119910119905 119905) denote the real target location
and estimated target location respectively It shows that theGA based adaptive 120572-120573-120574 filter converges to less than about2m after 5 sec
10 International Journal of Antennas and Propagation
Table 2 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0574 06399 0652 06904 06537 0674 06286 06313 06044 053921198921199011
0629 06664 0622 07756 0664 07399 0663 06652 06816 064981198922
0285 05827 03282 03851 00054 03480 02916 03736 0547 04864
Table 3 Estimated gain parameters for PSO based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0557 05986 0357 05345 06142 05774 05544 0529 04592 056261198921199011
0612 05898 04129 05675 05286 03953 05565 06199 05214 062851198922
0326 007 01136 0146 0098 02365 021 02547 02474 01245
0 20 40 60 80 100 1200
20
40
60
80
100
120
140
160
180
Time (s)
Et
(m)
Average estimation error = 15176m
Figure 9 Estimation error for linear target trajectory without beampointing error
Case 3 (tracking circular moving target in three-dimensionalspace using GA based adaptive 120572-120573-120574 filter (without beampointing error)) Figures 10(a) and 10(b) show the simulationscenario of beam pointing error compensation for circularmaneuvering air target and linear moving ship The shiphas the same linear moving speed and path equation asCase 2The flying target is parallel to the119883119884 plane the initialposition of flying target is (8885 4590 9100)m the velocity is300msec and the angle (120579
0) between the initial position and
the119883-axis is 30∘The trajectory of the flight vehicle is a circlewhich has a radius of 10 km The center of circular trajectoryis 119884-axis The rotational cycle time of the flight target is 209seconds and the total observation time is 419 sec
The circular trajectory equation is
X119898(119909119905 119910119905 119911119905) = (119903 cos (120579
0minus 120596119905) 119903 sin (120579
0minus 120596119905) 9100) (24)
where 119905 = 1 2 3 418 419 sec 119903 is the circular flight radiusin meter 120579
0is the initial angle between the flight vehicle
and the 119883-axis 120596 is the angular speed of flight vehicle Theaccuracy of GA based adaptive 120572-120573-120574 filter for tracking acircular flight target is shown in Figure 11 where theGAbasedadaptive 120572-120573-120574 filter converges to less than 3m after 8 secTherefore using the GA based adaptive 120572-120573-120574 filter to trackthe circular trajectory target has the larger average estimationerror and longer convergent time than linear trajectory target
Case 4 (estimating roll and pitch angles and tracking lineartrajectory target using GAPSO based adaptive 120572-120573-120574 filterand measured data (with beam pointing error)) The mea-sured roll and pitch signals shown in Figure 7 are used toevaluate the performance of GAPSO based adaptive 120572-120573-120574filter for estimating the roll and pitch angles and tracking thelinear trajectory targetThe beamof 6times6 planar array antennais steered to track the linear trajectory target The samplingfrequency is set as 10Hz Therefore the shiprsquos roll and pitchsignals are sampled 10000 points within 1000 seconds Asthe scenario described in Case 2 the ship is along the 119884
axis of 119883119884 plane in the earth coordinates the ship speedis 10msec and the flying target speed is 300msec Theflying target is parallel to the 119883119884 plane 91 km above theground and the angle between the flight direction and the119884-axis is 120∘ The largest reconnaissance distance of ship-borne radar is 150 km When the air target is flying withinthe radar detection range the beampointing error estimationand linear trajectory target tracking of ship-borne phasedarray radar are simulated
Figure 12 is a simulation flow chart used to verifythe effect of beam pointing error compensation on thetracking accuracy of adaptive 120572-120573-120574 filter Tables 2 and 3demonstrate the optimal gain parameters of adaptive 120572-120573-120574 filter for roll 119892
1199031 pitch 119892
1199011 and target tracking 119892
2
obtained by the GA and PSO methods respectively thatminimizes the RMSE Figures 13ndash17 show the estimationerrors for five different gain values which are summarizedin Table 4 It demonstrates that the greater gain value(119892 = 01 075 and 09) will generate the greater error andneed the longer convergence time The estimation errors
International Journal of Antennas and Propagation 11
Table 4 Estimation errors for different gains
Calculated by GA Calculated by PSO Fixed gain1198921199031
05392sim06904 0357sim06142 01 075 091198921199011
0622sim07756 03953sim06285 01 075 091198922
0054sim05827 007sim03259 01 075 09Average 119864
119905(m) 14781 163224 169663 534804 3991188
Table 5 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0543 05653 05424 05455 05665 056 058 0532 05414 053851198921199011
0464 0484 04459 0422 04667 04654 04509 04459 04308 046511198922
0161 00929 01937 00865 00241 00609 01174 01897 01713 01727
Y
X
Z
(a)
Y
X10 km
2730∘
(b)
Figure 10 Simulation scenarios of Case 3 for (a) 3D and (b) top view diagrams
0 10 20 30 40 50 60 70 80 90 1000
50
100
150
200
250
Time (s)
Et
(m)
Average estimation error = 24905m
Figure 11 Estimation error for circular target trajectory without beam pointing error
12 International Journal of Antennas and Propagation
Generate the target detection data
Using GA or PSO
to track the target
Generate beam steering angle
Measure or simulate roll and pitch angles
Using GA or PSO
roll and pitch angles
Simulate beam pointing error compensation
Verify tracking accuracy ofGAPSO adaptive 120572-120573-120574 filter
120572-120573-120574 filter 120572-120573-120574 filter to estimate
Figure 12 Flow chart of tracking performance simulation forCase 4experiment
0 100 200 300 400 500 600 700 800 900 10000
50
100
150
200
250
300
Time (s)
Et
(m)
Average estimation error = 14781m
Figure 13 Estimation error of GA based adaptive 120572-120573-120574 filter
presented in Figures 16 and 17 are too large to be appli-cable when the gain values are equal to 075 and 09The GA and PSO have smaller errors compared with thefixed gain values and the GA obtains the best estimationaccuracy The estimation errors of GA and PSO meth-ods were found to be close From Figures 13 14 and 15we can observe that there is a peak value that occurredat the time duration about 510 to 515 seconds becausethe horizontal beam direction of ship-borne phased array
0
50
100
150
200
250
300
Et
(m)
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 163224 m
Figure 14 Estimation error of PSO based adaptive 120572-120573-120574 filter
0
20
40
60
80
100
120
140
160
180
200Et
(m)
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 169663 m
Figure 15 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
radar is steered according to the linear trajectory targetwhich is flying through sky just above the ship at thetime instant of 510 second As shown in Figure 18 thehorizontal beam steering angle of ship-borne phased arrayradar at about the time of 510 second is changing fromdescending into ascending When the GA based adaptive120572-120573-120574 filter tracks the linear trajectory target the esti-mation error values are less than 30 meters or less ifthe peak estimation errors during these 5 seconds areignored
Case 5 (estimating roll and pitch angles and tracking lineartrajectory target using GAPSO based adaptive 120572-120573-120574 filterand simulated data (with beam pointing error)) The shiprsquosroll and pitch signals are simulated according to (3) and (4)with parameters of sea state 2 listed in Table 1 and standard
International Journal of Antennas and Propagation 13
0 200 400 600 800 10000
500
1000
1500
Time (s)
Et
(m)
Average estimation error = 534804 m
Figure 16 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0 200 400 600 800 10000
1000
2000
3000
4000
5000
6000
7000
8000
9000
Time (s)
Et
(m)
Average estimation error = 3991188 m
Figure 17 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
deviation of 001 The simulated shiprsquos roll and pitch signalsare applied for five different methods to estimate the gain ofGAPSO based adaptive 120572-120573-120574 filter The total observationtime is 1000 sec and sampling time is 01 sec The simulationreplicates 10000 times The input samples are updated byone new sample for the next iteration The other simulationscenario and parameters are the same asCase 4 Tables 5 and 6demonstrate the optimal gain parameters of adaptive 120572-120573-120574filter for roll gain 119892
1199031 pitch gain 119892
1199011 and target tracking gain
vector g2obtained by the GA and PSOmethods respectively
that minimizes RMSE Figures 19 20 21 22 and 23 showthe estimation errors for five different gain values which aresummarized in Table 7 The estimation error of GA basedadaptive 120572-120573-120574 filter for tracking a linear trajectory targetis shown in Figure 19 where the convergence time of GA
0 100 200 300 400 500 600 700 800 900 1000Time (s)
0
05
1
15
2
25
Azi
mut
h ste
erin
g an
gle (
rad)
With beam pointing errorWithout beam pointing error
Figure 18 Azimuth beam steering angle of ship-borne phased arrayradar
0
10
20
30
40
50
60
70
80
90
100
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 160756 m
Et
(m)
Figure 19 Estimation error of GA based adaptive 120572-120573-120574 filter
based adaptive 120572-120573-120574 filter is about 3 sec and the averageestimation error is about 160756m It concludes that theexperimental results in Case 5 have the same trend as inCase 4 therefore the simulated data can be used to observethe effect of shipmotion on the tracking performance of ship-borne phased array radar under different sea state conditionsThe convergence time and average estimation error of GAbased adaptive 120572-120573-120574 filter are better than themethod used in[6] where the tracking error of AEKF converges to less thanabout 20m at 550 iterations (seciteration) and the estimationerror will remain within the range of about 20m when thebeam pointing error is compensated by FBFN controller
Case 6 (estimating roll and pitch angles and tracking circularmoving target using GA based adaptive 120572-120573-120574 filter and
14 International Journal of Antennas and Propagation
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 187684 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 20 Estimation error of PSO based adaptive 120572-120573-120574 filter
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 199594 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 21 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
measured data (with beam pointing error)) As the scenariodescribed in Case 3 the measured roll and pitch signalsshown in Figure 7 are used to evaluate the performance ofGA based adaptive 120572-120573-120574 filter for estimating the roll andpitch angles and tracking the circular trajectory target Theestimation errors of 120572-120573-120574 filter with fixed gains of 119892
1199031=
1198921199011
= 1198922= 01 and 119892
1199031= 1198921199011
= 1198922= 075 are shown in
Figures 24 and 25 respectivelyThe average estimation errorsfor fixed filter gains of 01 and 075 are about 41638m and572386m respectively The optimal roll gain 119892
1199031 pitch gain
1198921199011 and target tracking gain 119892
2of adaptive 120572-120573-120574 filter
estimated by the GA method during the observation timeof 419 sec are listed in Table 8 The estimation error of GAbased adaptive 120572-120573-120574 filter for tracking a circular flight targetis shown in Figure 26 where the convergence time of GAbased adaptive 120572-120573-120574 filter is about 5 sec and the average
0
500
1000
1500
0 200 400 600 800 1000Time (s)
Average estimation error = 1080515 m
Et
(m)
Figure 22 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
0 200 400 600 800 1000Time (s)
Average estimation error = 4279449 m
Et
(m)
Figure 23 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
estimation error is about 36667m It concludes that the GAbased adaptive 120572-120573-120574 filter can greatly improve the trackingaccuracy and convergence time of the conventional 120572-120573-120574filter for tracking the circular trajectory target
5 Conclusions
This paper proposed an intelligent beam pointing error com-pensationmechanism for ship-borne phased array radarTheGA based adaptive 120572-120573-120574 filter estimates the roll and pitchangles of the ship moving in the sea and thus compensatesfor the antenna beam pointing error in order to enhance thetracking accuracy of phased array radar system
International Journal of Antennas and Propagation 15
Table 6 Estimated gain parameters for PSO based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0366 03645 0532 04519 05151 05002 04797 05254 04146 053511198921199011
0366 03647 03634 03644 03643 0363 03644 03684 03643 036421198922
0125 01271 02286 02166 01169 01799 01735 01564 00767 01346
Table 7 Estimation errors for different gains
Calculated by GA Calculated by PSO Fixed gain1198921199031
0459sim0583 03645sim05352 01 075 091198921199011
0498sim0667 0363sim03684 01 075 091198922
00586sim01697 00367sim02291 01 075 09Average 119864
119905(m) 160756 187684 199594 1080515 4279449
Table 8 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim42 43sim84 85sim126 127sim168 169sim210 211sim252 253sim294 295sim336 337sim378 379sim4191198921199031
0619 04996 05316 0644 05052 05568 05316 04459 05092 059241198921199011
0582 06727 07201 04913 04168 07169 05983 02036 05858 06211198922
0214 02215 01612 02474 01822 0157 02073 01211 02278 01074
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 41638 m
Figure 24 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
Six experimental cases are conducted to verify the perfor-mance of ship-borne phased array radar using the proposedGA based adaptive 120572-120573-120574 filter In Case 1 the GA basedadaptive 120572-120573-120574 filter is used to estimate the roll and pitchangles with the simulated and measured roll and pitchsignals respectively The simulation results show that themeasurement data has the minimum estimation error theestimation error in the sea state 2 is the second and theestimation error in the sea state 3 is the worst Cases 2 and 3show that when the GA based adaptive 120572-120573-120574 filter is appliedto estimate the beam pointing error and to track the targetin a ship-borne phased array radar the circular trajectory
0 50 100 150 200 250 300 350 400 450Time (s)
0
500
1000
1500
Average estimation error = 572386 m
Et
(m)
Figure 25 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
target will be tracked with the larger average estimation errorand longer convergence time than linear trajectory targetTheexperimental results ofCases 4 and 5demonstrate thatwhen alinear trajectory target is tracked by ship-borne phased arrayradar the average estimation error and convergence time ofship-borne phased array radar using the GA based adaptive120572-120573-120574 filter are better than PSO based adaptive 120572-120573-120574 filterand fixed filter gain 120572-120573-120574 filter The convergence time andtracking accuracy of ship-borne phased array radar usingthe proposed GA based adaptive 120572-120573-120574 filter are superiorto the AEKF significantly In conclusions the proposed GAbased adaptive 120572-120573-120574 filter is a real time applicable algorithm
16 International Journal of Antennas and Propagation
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 36667 m
Figure 26 Estimation error of GA based adaptive 120572-120573-120574 filter
whose accuracy is good with relatively less complexity Inaddition the roll and pitch signalsmeasured in real operationenvironments can be replaced with the simulated roll andpitch signals to test all the relevant properties of practical shipmotion compensation and air target tracking for ship-bornephased array radar
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
This work is supported in part by research grants fromNational Science Council China (NSC 102-2218-E-155-001)
References
[1] Z Fang and Q Rundong Ship-borne Phased Array RadarMotion Compensation System Engineering and ElectronicTechnologies 1998
[2] L J Love J F Jansen and F G Pin ldquoCompensation of waveinduced motion and force phenomenon for ship based highperformance robotic and human amplifying systemsrdquo OakRidge National Laboratory ORNLTM-2003233 2003
[3] R J Hill Motion Compensation for Ship Borne Radars andLidars US Department of Commerce National Oceanic andAtmospheric Administration Office of Oceanic and Atmo-spheric Research Earth System Research Laboratory PhysicalSciences Division 2005
[4] T I Fossen and T Perez ldquoKalman filtering for positioning andheading control of ships and offshore rigsrdquo IEEEControl SystemsMagazine vol 29 no 6 pp 32ndash46 2009
[5] S Kuchler C Pregizer J K Eberharter K Schneider and OSawodny ldquoReal-time estimation of a shiprsquos attituderdquo in Proceed-ings of theAmericanControl Conference (ACC rsquo11) pp 2411ndash2416San Francisco Calif USA July 2011
[6] J Mar K C Tsai Y T Wang and M B Basnet ldquoIntelligentmotion compensation for improving the tracking performanceof shipborne phased array radarrdquo International Journal ofAntennas and Propagation vol 2013 Article ID 384756 14pages 2013
[7] T Dirk and S Tarunraj ldquoOptimal design of 120572-120573-(120574) filtersrdquo inAmerican Control Conference vol 6 pp 4348ndash4352 ChicagoIll USA June 2000
[8] D Tenne and T Singh ldquoCharacterizing performance of 120572-120573-120574filtersrdquo IEEE Transactions on Aerospace and Electronic Systemsvol 38 no 3 pp 1072ndash1087 2002
[9] T E Lee J P Su and K W Yu ldquoParameter optimization fora third-order sampled-data trackerrdquo in Proceedings of the 2ndInternational Conference on Innovative Computing Informationand Control p 336 Kumamoto Japan September 2007
[10] Y H Lho and J H Painter ldquoA fuzzy tuned adaptive Kalmanfilterrdquo in Industrial FuzzyControl and Intelligent System pp 144ndash148 December 1993
[11] N Ramakoti A Vinay and R K Jatoth ldquoParticle swarm opti-mization aided Kalman filter for object trackingrdquo in Proceedingsof the International Conference on Advances in ComputingControl and Telecommunication Technologies (ACT rsquo09) pp 531ndash533 Kerela India December 2009
[12] T-E Lee J-P Su andK-WYu ldquoAparticle swarmoptimization-based state estimation scheme formoving objectsrdquoTransactionsof the Institute of Measurement and Control vol 34 no 2-3 pp236ndash254 2012
[13] D C Law S A McLaughlin M J Post et al ldquoAn electronicallystabilized phased array system for shipborne atmospheric windprofilingrdquo Journal of Atmospheric and Oceanic Technology vol19 no 6 pp 924ndash933 2002
[14] C A Balanis AntennaTheory Analysis and Design JohnWileyamp Sons New York NY USA 3rd edition 2005
[15] J Mar and Y R Lin ldquoImplementation of SDR digital beam-former for microsatellite SARrdquo IEEE Geoscience and RemoteSensing Letters vol 6 no 1 pp 92ndash96 2009
[16] R C Eberhart and Y Shi Computational Intelligence Conceptsto Implementations ElsevierMorgan Kaufmann 2007
[17] K Y Lee and J-B Park ldquoApplication of particle swarm optimi-zation to economic dispatch problem advantages anddisadvan-tagesrdquo in Proceedings of the IEEE PES Power Systems Conferenceand Exposition (PSCE rsquo06) pp 188ndash192 Atlanta Ga USANovember 2006
[18] R C Eberhart and J Kennedy ldquoNew optimizer using particleswarm theoryrdquo in Proceedings of the 6th International Sympo-sium onMicroMachine and Human Science pp 39ndash43 NagoyaJapan October 1995
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
International Journal of Antennas and Propagation 5
Tracking gain estimator
Estimation Prediction
Positionmeasurement
Delay
1205722-1205732-1205742 filter(g2)
Xm(k)
X(k)
V(k)
A(k)
Xe(k)
Ve(k)
Ae(k)
X(k + 1)
V(k + 1)
A(k + 1)
Figure 3 Adaptive 120572-120573-120574 filter for target tracking in three-dimensional space
The estimation equations are
X119890 (119896) =
X (119896) + 1205722[X119898 (
119896) minusX (119896)] + n
2 (119896)
V119890 (119896) = V (119896) +
1205732
1198792
[X119898 (
119896) minus X (119896)]
A119890 (119896) = A (119896) +
1205742
21198792
2
[X119898 (
119896) minus X (119896)]
(13)
where X(119896) = [119909(119896) 119910(119896) 119911(119896)]119879 V(119896) = [V
119909(119896) V119910(119896)
V119911(119896)]119879 and A(119896) = [119886
119909(119896) 119886119910(119896) 119886119911(119896)]119879 are the position
vector velocity vector and acceleration vector respectivelyat time instant 119896 n
2(119896) is the zero mean white Gaussian noise
and the sampling time 1198792= 1 sec The relationship among
1205722 1205732 and 120574
2parameters can be related to gain factor vector
g2= [1198922 1198922 1198922]119879 The formula is described as follows
1205722=[
[
1205722119909
0 0
0 1205722119910
0
0 0 1205722119911
]
]
=[
[
1 minus 1198923
20 0
0 1 minus 1198923
20
0 0 1 minus 1198923
2
]
]
1205732=[
[
1205732119909
0 0
0 1205732119910
0
0 0 1205732119911
]
]
=[
[
[
15 (1 + 1198922) (1 minus 119892
2)2
0 0
0 15 (1 + 1198922) (1 minus 119892
2)2
0
0 0 15 (1 + 1198922) (1 minus 119892
2)2
]
]
]
1205742=[
[
1205742119909
0 0
0 1205742119910
0
0 0 1205742119911
]
]
=[
[
[
(1 minus 1198922)3
0 0
0 (1 minus 1198922)3
0
0 0 (1 minus 1198922)3
]
]
]
(14)
33 GA Gain Estimator The flow chart of GA rollpitch gainestimators of adaptive 120572-120573-120574 filter is shown in Figure 4 TheGA rollpitch gain estimators are used to estimate the roll andpitch angles gain (119892
11199031198921119901) of adaptive 120572
1-1205731-1205741filters The
GA tracking gain estimator is used to estimate the optimalgain g
2of adaptive 120572
2-1205732-1205742filter The flow chart of GA
tracking gain estimator is similar to Figure 4TheGAmethoduses selection crossover andmutation [16] techniques to findthe solutions At first the initial population which containsrandomly generated chromosomes is made Chromosomesare chains of 0 and 1 bits whose length is defined as stringlength Chromosomes number is defined as the populationsizeThese chromosomes from themating pool are shuffled to
create more randomness and then applied to the problem tofind their outputs and fitness scores accordingly Fitness valuedefines how well the chromosome solves the problem Twochromosomes are selected from the mating pool dependingon selection method Depending on the crossover rate allbits between two randomly chosen points selected from twodifferent chromosomes are interchanged which is known astwo-point crossover Chromosomes may undergo mutationwhere random bits of chromosomes are flipped dependingon the mutation rate Termination criteria are defined by thenumber of generations
Selection is done usually by using two methods asfollows roulette wheel selection and tournament selection
6 International Journal of Antennas and Propagation
Get optimal filter gain
Evaluate
YesNo
End
Roulette wheelselection
Crossover
Mutation
Generate the next generation
Is termination criteria met
Generate the initial population randomly
between range of 0 and 1
gr1gp1
Input the rollpitch sea state data
Figure 4 Flow chart of GA rollpitch gain estimator
In roulette wheel selection the chromosomes selected frompopulation are put into a mating pool The chromosomeselection probability is proportion to the fitness value Intournament selection chromosomes have to go throughseveral tournaments The chromosome which has the fittestvalue is selected and put into the mating pool Here theroulette wheel selection method is chosen to determine thesolution
Two-point crossover calls for two random points selectedfrom two different parent chromosome strings all bitsbetween two points are swapped for two parent chromo-somes Crossover rate is commonly chosen in a range of060 to 080 [16] Mutation rate describes the probability thata bit in a chromosome will be flipped (0 flips to 1 and 1flips to 0) Mutation brings diversity to the population asmutation of a chromosome is a random process which willbring randomness to present chromosome group Generationdefines iteration the GA method runs for a defined numberof generations and the final best solution is considered as theresult
34 PSO Gain Estimator [9 12 17] PSO method was firstproposed in [18] PSO is based on the number of particlesflying around the solution space to find the best solutionthat minimizes the cost function Particles move around thesolution space as soon as a particle detects a better solutionthe information is passed to other particles and then particleschange their position with respect to their best positionobserved among all the particles Figure 5 describes the PSOmethod in detail The PSO changes the velocity and positionof the particle according to (15) and (17) respectively toachieve the fitness [16]
V119894119889(119899 + 1)
= 119870 lowast [V119894119889(119899)
+ 1198881sdot rand ( ) sdot (119901119894119889(119899) minus 119909119894119889(119899))
+ 1198882sdot Rand ( ) sdot (119901119892119889(119899) minus 119909119894119889(119899))]
(15)
119870 =
2
1003816100381610038161003816100381610038161003816
2 minus 120593 minus radic1205932minus 4120593
1003816100381610038161003816100381610038161003816
where 120593 = 1198881+ 1198882 120593 gt 4 (16)
119909119894119889(119899 + 1)
= 119909119894119889(119899)
+ V119894119889(119899)
(17)
International Journal of Antennas and Propagation 7
Initialize particles with random positions and random velocities
For each particle calculate the fitness values
Is fitness value ltbest fitness value
Particle exhausted
Set global best fitness value as best fitness value
Update particle positions and velocities using (18) and (20)
Maximum iterationreached
Give global best fitnessvalue and position
Set global best as local best fitness value
Yes
No
No
No
Yes
Yes
Figure 5 Flow chart of PSO method
where 119909119894119889(119899)
and V119894119889(119899)
are position and velocity of (119894119889)thparticle at (119899)th iteration respectively 119901
119894119889(119899)and 119901
119892119889(119899)are
the best position and global best position of (119894119889)th particle at(119899)th iteration respectively Rand( ) and rand( ) are randomnumber between 0 and 1 respectively Equation (15) updatesthe velocity of the particles and (17) updates the position ofparticles 119870 means inertia 119888
1and 1198882are correction factors
and swarm size means number of particles for an iterationas described in [16]
Particles in the PSO method sometimes go out of con-straints while changing their position Reinitialization of theparticlersquos position randomly to fall within the constraints isdone as soon as a particle moves out of constraints Thisensures full use of particles and also increases the chance ofdiscovering a better solution To reduce the chance of gettingtrapped in a local minimum the advantages of using PSOmethod are that it is derivative-free easy to implement andeasy to comprehend and the solution is independent of theinitial point [17] One of the drawbacks of the PSOmethod isthat it does not guarantee success that is the solution foundmay not be the best solution
35 GAPSO Based Adaptive 120572-120573-120574 Filter The flow chart ofthe adaptive 120572-120573-120574 filter algorithm is shown in Figure 6where 119873 is the number of sampling data to determine thefilter gain In this algorithm the parameters120572120573 and 120574 whichare related to the filter gain 119892 are optimized by minimizingthe RMSE using GA or PSO method The proposed adaptive120572-120573-120574 filter processes the current sampled data 119909
119894which
predicts the next sample data 119909119894+1
as iteration continuesThe adaptive GA or PSO process only occurs at certain timeintervalThe previous119873 data that is 119909
119901(119894minus119873) sdot sdot sdot 119909
119901(119894minus1) is
sent to the adaptiveGAor PSO algorithmwhere the optimumvalue of filter gain (119892) is determined for the 120572-120573-120574 filterThen 119909
119894+1is predicted and the process continues The value
of119873 determines running time and accuracy of the algorithmso 119873 must have small enough value to support real timeimplementation and large enough value to provide acceptableaccuracy
Since the roll and pitch angles of ship motion changeover time randomly in order to speed up the processingtime and to reach the objectives of minimum estimationerror the filter gain values are real time updated by using thepipelined architectureThe recorded roll and pitch signals are
8 International Journal of Antennas and Propagation
Collect previous N data using PSO or GA algorithm
Predicted value for120572-120573-120574 filterxi
xi+1
ximinus1 ximinusFind g for Nximinus1 ximinusN
that minimizes RMSE
Figure 6 Block diagram of adaptive 120572-120573-120574 filter
segmented into block of 1000 sampled data (100 sec) for lineartrajectory target and 420 sampled data (42 sec) for circulartrajectory target respectivelyThe gain value of 120572
2-1205732-1205742filter
in 101ndash200 sec is estimated by GA and PSO methods usingthe collected data during 1ndash100 sec the gain value of 120572
2-
1205732-1205742filter in 201ndash300 sec is estimated by the GA and PSO
methods using the collected data in 101ndash200 sec and so onSimilarly the measured flight target signals are segmentedinto block of 100 sampled data (100 sec) for linear trajectorytarget and 42 sampled data (42 sec) for circular trajectorytarget respectively The gain value of 120572
2-1205732-1205742filter in 101ndash
200 sec is estimated by the GA and PSO methods using thecollected data during 1ndash100 sec the gain value of 120572
2-1205732-1205742
filter in the interval of 201ndash300 sec is estimated by the GA andPSO methods using the collected data in 101ndash200 sec and soon
For the PSO process it was found that gain parameterconverged within 100 iterations so the number of iterationsfor an adaptive 120572-120573-120574 filter was kept at constant 100 iterationsper processing cycle swarm size = 30 inertia = 07298 andcorrection factors 119888
1= 21 119888
2= 2 For the GA process we
have considered population size = 8 string length = 12crossover rate = 08 and mutation rate = 005
If 120579119894119901 119894 = 1 2 119873 and
120579119894119901 119894 = 1 2 119873 are real
measurement pitchrsquos values and estimated pitchrsquos values ofadaptive 120572
1-1205731-1205741filter respectively then 120579
119894119903 119894 = 1 2 119873
and 120579119894119903 119894 = 1 2 119873 are realmeasurement rollrsquos values and
estimated rollrsquos values of adaptive 1205721-1205731-1205741filter respectively
where 119894 is the current iteration and 119873 is the total number ofsampling data The RMSEs of pitch and roll estimations aredefined as follows
RMSE119901= radic
1
119873
119873
sum
119894 = 1
(120579119894119901minus 120579119894119901)
2
(18)
RMSE119903= radic
1
119873
119873
sum
119894 = 1
(120579119894119903minus 120579119894119903)
2
(19)
where the main purpose of GA and PSO methods is to findthe optimum values of 119892
1119901and 119892
1119903which minimize the
RMSE119901and RMSE
119903 respectively
If X119894and X
119894 119894 = 1 2 119873 are real measurement target
position vector values and estimated target position vector
values of adaptive 1205722-1205732-1205742filter the RMSE of target tracking
is defined as
RMSE = radic1
119873
119873
sum
119894 = 1
((119909119894minus 119909119894)2+ (119910119894minus 119910119894)2+ (119911119894minus 119894)2) (20)
where 119909119894 119910119894 and 119911
119894are the components of target position
vector X119894in 119909 119910 and 119911 axes The main purpose of GA and
PSO algorithms is to find the optimum gain value of g2which
minimizes the RMSE
4 Experimental Results
Six scenarios are used to simulate the tracking performanceof ship-borne phased array radar using the proposed adaptive120572-120573-120574 filter for beam pointing error compensation and targettracking in three-dimensional space
Case 1 (roll and pitch estimations using GA based adaptive120572-120573-120574 filter and measured data) The roll and pitch datarecorded by a gyroscope of the ship were used in the exper-iments The total number of data is 1000 and the samplingtime is 01 sec The angle variation of roll and pitch signalsmeasured by ship gyroscope are shown in Figure 7 whichwillresult in the beam pointing errors of phased array antennaTheGA based adaptive 120572-120573-120574 filter is used to estimate the rolland pitch angles Figures 8(a) and 8(b) show that the conver-gent time of roll and pitch angles are 3 sec and 5 sec respec-tively under simulated sea states 2 and 3 and measured dataThe shiprsquos roll and pitch signals are simulated according to (3)and (4) with roll and pitch angle parameters of sea states 2and 3 listed in Table 1 and standard deviation of 001 Itconcludes that the proposed GA based adaptive 120572-120573-120574 filter issuitable to compensate the ship motion As shown in Table 1the period of sea state 2 is longer than sea state 3 Thereforethe convergent RMSE (about 019∘ for roll about 01∘ forpitch) of sea state 2 is less than sea state 3 (about 029∘ for rollabout 022∘ for pitch) because the roll and pitch signals withshorter period in sea state 3 have rapid oscillations which aremore difficult to be predicted precisely
Case 2 (tracking linear moving target in three-dimensionalspace using GA based adaptive 120572-120573-120574 filter (without beampointing error)) Assuming the ship is not affected by thesea waves (antenna beam pointing error is zero) the trackingperformance of GA based adaptive 120572-120573-120574 filter for a straightflight target trajectory is simulated The initial position ofradar is (0 0 0) The ship moves with a speed of 10msec in
International Journal of Antennas and Propagation 9
0 100 200 300 400 500 600 700 800 900 1000
0
02
04
06
08
Pitc
h an
gle (
deg)
Time (10minus1 s)
minus12
minus08
minus06
minus04
minus02
minus1
(a)
0
1
2
3
Roll
angl
e (de
g)
0 100 200 300 400 500 600 700 800 900 1000Time (10minus1 s)
minus1
minus2
minus3
minus4
(b)
Figure 7 Demonstrating angle variation of (a) pitch and (b) roll signals measured by gyroscope
0
005
01
015
02
025
03
035
04
RMSE
(deg
)
Sea 3Sea 2Measurement
0 100 200 300 400 500 600 700 800 900 1000Time (10minus1 s)
(a)
0
005
01
015
02
025
03
035
04RM
SE (d
eg)
Sea 3Sea 2Measurement
Time (10minus1 s)0 100 200 300 400 500 600 700 800 900 1000
(b)
Figure 8 RMSE of (a) roll and (b) pitch under simulated data of sea states 2 and 3 and measured data
the 119884 axial direction The linear path equation of the ship isgiven by
X119904(119909119905 119910119905 119911119905) = X0119904+ V119904119905 (21)
where 119905 = 1 2 3 1000 sec V119904is ship velocity vector
and X0119904is initial ship position vector in three-dimensional
spaceThe radar position is updated every secondThe initialposition of flying target is (minus74840 minus129620 9100)m whichis about 150 km from the radar The flight speed of target is300msec (119883 axial velocity of 150msec and 119884 axial velocityof 260msec) The target location is updated every secondThe linear target trajectory equation is
X119898(119909119905 119910119905 119911119905) = X0119898
+ V119898119905 (22)
where V119898
is the target velocity vector and X0119898
is theinitial target position vector in three-dimensional space Thetracking accuracy of GA based adaptive 120572-120573-120574 filter for astraight flight target is shown in Figure 9 where the trajectoryestimation error is calculated by
119864119905= radic(119909
119905minus 119909119905)2+ (119910119905minus 119910119905)2+ (119911119905minus 119905)2 (23)
where (119909119905 119910119905 119911119905) and (119909
119905 119910119905 119905) denote the real target location
and estimated target location respectively It shows that theGA based adaptive 120572-120573-120574 filter converges to less than about2m after 5 sec
10 International Journal of Antennas and Propagation
Table 2 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0574 06399 0652 06904 06537 0674 06286 06313 06044 053921198921199011
0629 06664 0622 07756 0664 07399 0663 06652 06816 064981198922
0285 05827 03282 03851 00054 03480 02916 03736 0547 04864
Table 3 Estimated gain parameters for PSO based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0557 05986 0357 05345 06142 05774 05544 0529 04592 056261198921199011
0612 05898 04129 05675 05286 03953 05565 06199 05214 062851198922
0326 007 01136 0146 0098 02365 021 02547 02474 01245
0 20 40 60 80 100 1200
20
40
60
80
100
120
140
160
180
Time (s)
Et
(m)
Average estimation error = 15176m
Figure 9 Estimation error for linear target trajectory without beampointing error
Case 3 (tracking circular moving target in three-dimensionalspace using GA based adaptive 120572-120573-120574 filter (without beampointing error)) Figures 10(a) and 10(b) show the simulationscenario of beam pointing error compensation for circularmaneuvering air target and linear moving ship The shiphas the same linear moving speed and path equation asCase 2The flying target is parallel to the119883119884 plane the initialposition of flying target is (8885 4590 9100)m the velocity is300msec and the angle (120579
0) between the initial position and
the119883-axis is 30∘The trajectory of the flight vehicle is a circlewhich has a radius of 10 km The center of circular trajectoryis 119884-axis The rotational cycle time of the flight target is 209seconds and the total observation time is 419 sec
The circular trajectory equation is
X119898(119909119905 119910119905 119911119905) = (119903 cos (120579
0minus 120596119905) 119903 sin (120579
0minus 120596119905) 9100) (24)
where 119905 = 1 2 3 418 419 sec 119903 is the circular flight radiusin meter 120579
0is the initial angle between the flight vehicle
and the 119883-axis 120596 is the angular speed of flight vehicle Theaccuracy of GA based adaptive 120572-120573-120574 filter for tracking acircular flight target is shown in Figure 11 where theGAbasedadaptive 120572-120573-120574 filter converges to less than 3m after 8 secTherefore using the GA based adaptive 120572-120573-120574 filter to trackthe circular trajectory target has the larger average estimationerror and longer convergent time than linear trajectory target
Case 4 (estimating roll and pitch angles and tracking lineartrajectory target using GAPSO based adaptive 120572-120573-120574 filterand measured data (with beam pointing error)) The mea-sured roll and pitch signals shown in Figure 7 are used toevaluate the performance of GAPSO based adaptive 120572-120573-120574filter for estimating the roll and pitch angles and tracking thelinear trajectory targetThe beamof 6times6 planar array antennais steered to track the linear trajectory target The samplingfrequency is set as 10Hz Therefore the shiprsquos roll and pitchsignals are sampled 10000 points within 1000 seconds Asthe scenario described in Case 2 the ship is along the 119884
axis of 119883119884 plane in the earth coordinates the ship speedis 10msec and the flying target speed is 300msec Theflying target is parallel to the 119883119884 plane 91 km above theground and the angle between the flight direction and the119884-axis is 120∘ The largest reconnaissance distance of ship-borne radar is 150 km When the air target is flying withinthe radar detection range the beampointing error estimationand linear trajectory target tracking of ship-borne phasedarray radar are simulated
Figure 12 is a simulation flow chart used to verifythe effect of beam pointing error compensation on thetracking accuracy of adaptive 120572-120573-120574 filter Tables 2 and 3demonstrate the optimal gain parameters of adaptive 120572-120573-120574 filter for roll 119892
1199031 pitch 119892
1199011 and target tracking 119892
2
obtained by the GA and PSO methods respectively thatminimizes the RMSE Figures 13ndash17 show the estimationerrors for five different gain values which are summarizedin Table 4 It demonstrates that the greater gain value(119892 = 01 075 and 09) will generate the greater error andneed the longer convergence time The estimation errors
International Journal of Antennas and Propagation 11
Table 4 Estimation errors for different gains
Calculated by GA Calculated by PSO Fixed gain1198921199031
05392sim06904 0357sim06142 01 075 091198921199011
0622sim07756 03953sim06285 01 075 091198922
0054sim05827 007sim03259 01 075 09Average 119864
119905(m) 14781 163224 169663 534804 3991188
Table 5 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0543 05653 05424 05455 05665 056 058 0532 05414 053851198921199011
0464 0484 04459 0422 04667 04654 04509 04459 04308 046511198922
0161 00929 01937 00865 00241 00609 01174 01897 01713 01727
Y
X
Z
(a)
Y
X10 km
2730∘
(b)
Figure 10 Simulation scenarios of Case 3 for (a) 3D and (b) top view diagrams
0 10 20 30 40 50 60 70 80 90 1000
50
100
150
200
250
Time (s)
Et
(m)
Average estimation error = 24905m
Figure 11 Estimation error for circular target trajectory without beam pointing error
12 International Journal of Antennas and Propagation
Generate the target detection data
Using GA or PSO
to track the target
Generate beam steering angle
Measure or simulate roll and pitch angles
Using GA or PSO
roll and pitch angles
Simulate beam pointing error compensation
Verify tracking accuracy ofGAPSO adaptive 120572-120573-120574 filter
120572-120573-120574 filter 120572-120573-120574 filter to estimate
Figure 12 Flow chart of tracking performance simulation forCase 4experiment
0 100 200 300 400 500 600 700 800 900 10000
50
100
150
200
250
300
Time (s)
Et
(m)
Average estimation error = 14781m
Figure 13 Estimation error of GA based adaptive 120572-120573-120574 filter
presented in Figures 16 and 17 are too large to be appli-cable when the gain values are equal to 075 and 09The GA and PSO have smaller errors compared with thefixed gain values and the GA obtains the best estimationaccuracy The estimation errors of GA and PSO meth-ods were found to be close From Figures 13 14 and 15we can observe that there is a peak value that occurredat the time duration about 510 to 515 seconds becausethe horizontal beam direction of ship-borne phased array
0
50
100
150
200
250
300
Et
(m)
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 163224 m
Figure 14 Estimation error of PSO based adaptive 120572-120573-120574 filter
0
20
40
60
80
100
120
140
160
180
200Et
(m)
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 169663 m
Figure 15 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
radar is steered according to the linear trajectory targetwhich is flying through sky just above the ship at thetime instant of 510 second As shown in Figure 18 thehorizontal beam steering angle of ship-borne phased arrayradar at about the time of 510 second is changing fromdescending into ascending When the GA based adaptive120572-120573-120574 filter tracks the linear trajectory target the esti-mation error values are less than 30 meters or less ifthe peak estimation errors during these 5 seconds areignored
Case 5 (estimating roll and pitch angles and tracking lineartrajectory target using GAPSO based adaptive 120572-120573-120574 filterand simulated data (with beam pointing error)) The shiprsquosroll and pitch signals are simulated according to (3) and (4)with parameters of sea state 2 listed in Table 1 and standard
International Journal of Antennas and Propagation 13
0 200 400 600 800 10000
500
1000
1500
Time (s)
Et
(m)
Average estimation error = 534804 m
Figure 16 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0 200 400 600 800 10000
1000
2000
3000
4000
5000
6000
7000
8000
9000
Time (s)
Et
(m)
Average estimation error = 3991188 m
Figure 17 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
deviation of 001 The simulated shiprsquos roll and pitch signalsare applied for five different methods to estimate the gain ofGAPSO based adaptive 120572-120573-120574 filter The total observationtime is 1000 sec and sampling time is 01 sec The simulationreplicates 10000 times The input samples are updated byone new sample for the next iteration The other simulationscenario and parameters are the same asCase 4 Tables 5 and 6demonstrate the optimal gain parameters of adaptive 120572-120573-120574filter for roll gain 119892
1199031 pitch gain 119892
1199011 and target tracking gain
vector g2obtained by the GA and PSOmethods respectively
that minimizes RMSE Figures 19 20 21 22 and 23 showthe estimation errors for five different gain values which aresummarized in Table 7 The estimation error of GA basedadaptive 120572-120573-120574 filter for tracking a linear trajectory targetis shown in Figure 19 where the convergence time of GA
0 100 200 300 400 500 600 700 800 900 1000Time (s)
0
05
1
15
2
25
Azi
mut
h ste
erin
g an
gle (
rad)
With beam pointing errorWithout beam pointing error
Figure 18 Azimuth beam steering angle of ship-borne phased arrayradar
0
10
20
30
40
50
60
70
80
90
100
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 160756 m
Et
(m)
Figure 19 Estimation error of GA based adaptive 120572-120573-120574 filter
based adaptive 120572-120573-120574 filter is about 3 sec and the averageestimation error is about 160756m It concludes that theexperimental results in Case 5 have the same trend as inCase 4 therefore the simulated data can be used to observethe effect of shipmotion on the tracking performance of ship-borne phased array radar under different sea state conditionsThe convergence time and average estimation error of GAbased adaptive 120572-120573-120574 filter are better than themethod used in[6] where the tracking error of AEKF converges to less thanabout 20m at 550 iterations (seciteration) and the estimationerror will remain within the range of about 20m when thebeam pointing error is compensated by FBFN controller
Case 6 (estimating roll and pitch angles and tracking circularmoving target using GA based adaptive 120572-120573-120574 filter and
14 International Journal of Antennas and Propagation
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 187684 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 20 Estimation error of PSO based adaptive 120572-120573-120574 filter
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 199594 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 21 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
measured data (with beam pointing error)) As the scenariodescribed in Case 3 the measured roll and pitch signalsshown in Figure 7 are used to evaluate the performance ofGA based adaptive 120572-120573-120574 filter for estimating the roll andpitch angles and tracking the circular trajectory target Theestimation errors of 120572-120573-120574 filter with fixed gains of 119892
1199031=
1198921199011
= 1198922= 01 and 119892
1199031= 1198921199011
= 1198922= 075 are shown in
Figures 24 and 25 respectivelyThe average estimation errorsfor fixed filter gains of 01 and 075 are about 41638m and572386m respectively The optimal roll gain 119892
1199031 pitch gain
1198921199011 and target tracking gain 119892
2of adaptive 120572-120573-120574 filter
estimated by the GA method during the observation timeof 419 sec are listed in Table 8 The estimation error of GAbased adaptive 120572-120573-120574 filter for tracking a circular flight targetis shown in Figure 26 where the convergence time of GAbased adaptive 120572-120573-120574 filter is about 5 sec and the average
0
500
1000
1500
0 200 400 600 800 1000Time (s)
Average estimation error = 1080515 m
Et
(m)
Figure 22 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
0 200 400 600 800 1000Time (s)
Average estimation error = 4279449 m
Et
(m)
Figure 23 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
estimation error is about 36667m It concludes that the GAbased adaptive 120572-120573-120574 filter can greatly improve the trackingaccuracy and convergence time of the conventional 120572-120573-120574filter for tracking the circular trajectory target
5 Conclusions
This paper proposed an intelligent beam pointing error com-pensationmechanism for ship-borne phased array radarTheGA based adaptive 120572-120573-120574 filter estimates the roll and pitchangles of the ship moving in the sea and thus compensatesfor the antenna beam pointing error in order to enhance thetracking accuracy of phased array radar system
International Journal of Antennas and Propagation 15
Table 6 Estimated gain parameters for PSO based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0366 03645 0532 04519 05151 05002 04797 05254 04146 053511198921199011
0366 03647 03634 03644 03643 0363 03644 03684 03643 036421198922
0125 01271 02286 02166 01169 01799 01735 01564 00767 01346
Table 7 Estimation errors for different gains
Calculated by GA Calculated by PSO Fixed gain1198921199031
0459sim0583 03645sim05352 01 075 091198921199011
0498sim0667 0363sim03684 01 075 091198922
00586sim01697 00367sim02291 01 075 09Average 119864
119905(m) 160756 187684 199594 1080515 4279449
Table 8 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim42 43sim84 85sim126 127sim168 169sim210 211sim252 253sim294 295sim336 337sim378 379sim4191198921199031
0619 04996 05316 0644 05052 05568 05316 04459 05092 059241198921199011
0582 06727 07201 04913 04168 07169 05983 02036 05858 06211198922
0214 02215 01612 02474 01822 0157 02073 01211 02278 01074
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 41638 m
Figure 24 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
Six experimental cases are conducted to verify the perfor-mance of ship-borne phased array radar using the proposedGA based adaptive 120572-120573-120574 filter In Case 1 the GA basedadaptive 120572-120573-120574 filter is used to estimate the roll and pitchangles with the simulated and measured roll and pitchsignals respectively The simulation results show that themeasurement data has the minimum estimation error theestimation error in the sea state 2 is the second and theestimation error in the sea state 3 is the worst Cases 2 and 3show that when the GA based adaptive 120572-120573-120574 filter is appliedto estimate the beam pointing error and to track the targetin a ship-borne phased array radar the circular trajectory
0 50 100 150 200 250 300 350 400 450Time (s)
0
500
1000
1500
Average estimation error = 572386 m
Et
(m)
Figure 25 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
target will be tracked with the larger average estimation errorand longer convergence time than linear trajectory targetTheexperimental results ofCases 4 and 5demonstrate thatwhen alinear trajectory target is tracked by ship-borne phased arrayradar the average estimation error and convergence time ofship-borne phased array radar using the GA based adaptive120572-120573-120574 filter are better than PSO based adaptive 120572-120573-120574 filterand fixed filter gain 120572-120573-120574 filter The convergence time andtracking accuracy of ship-borne phased array radar usingthe proposed GA based adaptive 120572-120573-120574 filter are superiorto the AEKF significantly In conclusions the proposed GAbased adaptive 120572-120573-120574 filter is a real time applicable algorithm
16 International Journal of Antennas and Propagation
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 36667 m
Figure 26 Estimation error of GA based adaptive 120572-120573-120574 filter
whose accuracy is good with relatively less complexity Inaddition the roll and pitch signalsmeasured in real operationenvironments can be replaced with the simulated roll andpitch signals to test all the relevant properties of practical shipmotion compensation and air target tracking for ship-bornephased array radar
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
This work is supported in part by research grants fromNational Science Council China (NSC 102-2218-E-155-001)
References
[1] Z Fang and Q Rundong Ship-borne Phased Array RadarMotion Compensation System Engineering and ElectronicTechnologies 1998
[2] L J Love J F Jansen and F G Pin ldquoCompensation of waveinduced motion and force phenomenon for ship based highperformance robotic and human amplifying systemsrdquo OakRidge National Laboratory ORNLTM-2003233 2003
[3] R J Hill Motion Compensation for Ship Borne Radars andLidars US Department of Commerce National Oceanic andAtmospheric Administration Office of Oceanic and Atmo-spheric Research Earth System Research Laboratory PhysicalSciences Division 2005
[4] T I Fossen and T Perez ldquoKalman filtering for positioning andheading control of ships and offshore rigsrdquo IEEEControl SystemsMagazine vol 29 no 6 pp 32ndash46 2009
[5] S Kuchler C Pregizer J K Eberharter K Schneider and OSawodny ldquoReal-time estimation of a shiprsquos attituderdquo in Proceed-ings of theAmericanControl Conference (ACC rsquo11) pp 2411ndash2416San Francisco Calif USA July 2011
[6] J Mar K C Tsai Y T Wang and M B Basnet ldquoIntelligentmotion compensation for improving the tracking performanceof shipborne phased array radarrdquo International Journal ofAntennas and Propagation vol 2013 Article ID 384756 14pages 2013
[7] T Dirk and S Tarunraj ldquoOptimal design of 120572-120573-(120574) filtersrdquo inAmerican Control Conference vol 6 pp 4348ndash4352 ChicagoIll USA June 2000
[8] D Tenne and T Singh ldquoCharacterizing performance of 120572-120573-120574filtersrdquo IEEE Transactions on Aerospace and Electronic Systemsvol 38 no 3 pp 1072ndash1087 2002
[9] T E Lee J P Su and K W Yu ldquoParameter optimization fora third-order sampled-data trackerrdquo in Proceedings of the 2ndInternational Conference on Innovative Computing Informationand Control p 336 Kumamoto Japan September 2007
[10] Y H Lho and J H Painter ldquoA fuzzy tuned adaptive Kalmanfilterrdquo in Industrial FuzzyControl and Intelligent System pp 144ndash148 December 1993
[11] N Ramakoti A Vinay and R K Jatoth ldquoParticle swarm opti-mization aided Kalman filter for object trackingrdquo in Proceedingsof the International Conference on Advances in ComputingControl and Telecommunication Technologies (ACT rsquo09) pp 531ndash533 Kerela India December 2009
[12] T-E Lee J-P Su andK-WYu ldquoAparticle swarmoptimization-based state estimation scheme formoving objectsrdquoTransactionsof the Institute of Measurement and Control vol 34 no 2-3 pp236ndash254 2012
[13] D C Law S A McLaughlin M J Post et al ldquoAn electronicallystabilized phased array system for shipborne atmospheric windprofilingrdquo Journal of Atmospheric and Oceanic Technology vol19 no 6 pp 924ndash933 2002
[14] C A Balanis AntennaTheory Analysis and Design JohnWileyamp Sons New York NY USA 3rd edition 2005
[15] J Mar and Y R Lin ldquoImplementation of SDR digital beam-former for microsatellite SARrdquo IEEE Geoscience and RemoteSensing Letters vol 6 no 1 pp 92ndash96 2009
[16] R C Eberhart and Y Shi Computational Intelligence Conceptsto Implementations ElsevierMorgan Kaufmann 2007
[17] K Y Lee and J-B Park ldquoApplication of particle swarm optimi-zation to economic dispatch problem advantages anddisadvan-tagesrdquo in Proceedings of the IEEE PES Power Systems Conferenceand Exposition (PSCE rsquo06) pp 188ndash192 Atlanta Ga USANovember 2006
[18] R C Eberhart and J Kennedy ldquoNew optimizer using particleswarm theoryrdquo in Proceedings of the 6th International Sympo-sium onMicroMachine and Human Science pp 39ndash43 NagoyaJapan October 1995
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
6 International Journal of Antennas and Propagation
Get optimal filter gain
Evaluate
YesNo
End
Roulette wheelselection
Crossover
Mutation
Generate the next generation
Is termination criteria met
Generate the initial population randomly
between range of 0 and 1
gr1gp1
Input the rollpitch sea state data
Figure 4 Flow chart of GA rollpitch gain estimator
In roulette wheel selection the chromosomes selected frompopulation are put into a mating pool The chromosomeselection probability is proportion to the fitness value Intournament selection chromosomes have to go throughseveral tournaments The chromosome which has the fittestvalue is selected and put into the mating pool Here theroulette wheel selection method is chosen to determine thesolution
Two-point crossover calls for two random points selectedfrom two different parent chromosome strings all bitsbetween two points are swapped for two parent chromo-somes Crossover rate is commonly chosen in a range of060 to 080 [16] Mutation rate describes the probability thata bit in a chromosome will be flipped (0 flips to 1 and 1flips to 0) Mutation brings diversity to the population asmutation of a chromosome is a random process which willbring randomness to present chromosome group Generationdefines iteration the GA method runs for a defined numberof generations and the final best solution is considered as theresult
34 PSO Gain Estimator [9 12 17] PSO method was firstproposed in [18] PSO is based on the number of particlesflying around the solution space to find the best solutionthat minimizes the cost function Particles move around thesolution space as soon as a particle detects a better solutionthe information is passed to other particles and then particleschange their position with respect to their best positionobserved among all the particles Figure 5 describes the PSOmethod in detail The PSO changes the velocity and positionof the particle according to (15) and (17) respectively toachieve the fitness [16]
V119894119889(119899 + 1)
= 119870 lowast [V119894119889(119899)
+ 1198881sdot rand ( ) sdot (119901119894119889(119899) minus 119909119894119889(119899))
+ 1198882sdot Rand ( ) sdot (119901119892119889(119899) minus 119909119894119889(119899))]
(15)
119870 =
2
1003816100381610038161003816100381610038161003816
2 minus 120593 minus radic1205932minus 4120593
1003816100381610038161003816100381610038161003816
where 120593 = 1198881+ 1198882 120593 gt 4 (16)
119909119894119889(119899 + 1)
= 119909119894119889(119899)
+ V119894119889(119899)
(17)
International Journal of Antennas and Propagation 7
Initialize particles with random positions and random velocities
For each particle calculate the fitness values
Is fitness value ltbest fitness value
Particle exhausted
Set global best fitness value as best fitness value
Update particle positions and velocities using (18) and (20)
Maximum iterationreached
Give global best fitnessvalue and position
Set global best as local best fitness value
Yes
No
No
No
Yes
Yes
Figure 5 Flow chart of PSO method
where 119909119894119889(119899)
and V119894119889(119899)
are position and velocity of (119894119889)thparticle at (119899)th iteration respectively 119901
119894119889(119899)and 119901
119892119889(119899)are
the best position and global best position of (119894119889)th particle at(119899)th iteration respectively Rand( ) and rand( ) are randomnumber between 0 and 1 respectively Equation (15) updatesthe velocity of the particles and (17) updates the position ofparticles 119870 means inertia 119888
1and 1198882are correction factors
and swarm size means number of particles for an iterationas described in [16]
Particles in the PSO method sometimes go out of con-straints while changing their position Reinitialization of theparticlersquos position randomly to fall within the constraints isdone as soon as a particle moves out of constraints Thisensures full use of particles and also increases the chance ofdiscovering a better solution To reduce the chance of gettingtrapped in a local minimum the advantages of using PSOmethod are that it is derivative-free easy to implement andeasy to comprehend and the solution is independent of theinitial point [17] One of the drawbacks of the PSOmethod isthat it does not guarantee success that is the solution foundmay not be the best solution
35 GAPSO Based Adaptive 120572-120573-120574 Filter The flow chart ofthe adaptive 120572-120573-120574 filter algorithm is shown in Figure 6where 119873 is the number of sampling data to determine thefilter gain In this algorithm the parameters120572120573 and 120574 whichare related to the filter gain 119892 are optimized by minimizingthe RMSE using GA or PSO method The proposed adaptive120572-120573-120574 filter processes the current sampled data 119909
119894which
predicts the next sample data 119909119894+1
as iteration continuesThe adaptive GA or PSO process only occurs at certain timeintervalThe previous119873 data that is 119909
119901(119894minus119873) sdot sdot sdot 119909
119901(119894minus1) is
sent to the adaptiveGAor PSO algorithmwhere the optimumvalue of filter gain (119892) is determined for the 120572-120573-120574 filterThen 119909
119894+1is predicted and the process continues The value
of119873 determines running time and accuracy of the algorithmso 119873 must have small enough value to support real timeimplementation and large enough value to provide acceptableaccuracy
Since the roll and pitch angles of ship motion changeover time randomly in order to speed up the processingtime and to reach the objectives of minimum estimationerror the filter gain values are real time updated by using thepipelined architectureThe recorded roll and pitch signals are
8 International Journal of Antennas and Propagation
Collect previous N data using PSO or GA algorithm
Predicted value for120572-120573-120574 filterxi
xi+1
ximinus1 ximinusFind g for Nximinus1 ximinusN
that minimizes RMSE
Figure 6 Block diagram of adaptive 120572-120573-120574 filter
segmented into block of 1000 sampled data (100 sec) for lineartrajectory target and 420 sampled data (42 sec) for circulartrajectory target respectivelyThe gain value of 120572
2-1205732-1205742filter
in 101ndash200 sec is estimated by GA and PSO methods usingthe collected data during 1ndash100 sec the gain value of 120572
2-
1205732-1205742filter in 201ndash300 sec is estimated by the GA and PSO
methods using the collected data in 101ndash200 sec and so onSimilarly the measured flight target signals are segmentedinto block of 100 sampled data (100 sec) for linear trajectorytarget and 42 sampled data (42 sec) for circular trajectorytarget respectively The gain value of 120572
2-1205732-1205742filter in 101ndash
200 sec is estimated by the GA and PSO methods using thecollected data during 1ndash100 sec the gain value of 120572
2-1205732-1205742
filter in the interval of 201ndash300 sec is estimated by the GA andPSO methods using the collected data in 101ndash200 sec and soon
For the PSO process it was found that gain parameterconverged within 100 iterations so the number of iterationsfor an adaptive 120572-120573-120574 filter was kept at constant 100 iterationsper processing cycle swarm size = 30 inertia = 07298 andcorrection factors 119888
1= 21 119888
2= 2 For the GA process we
have considered population size = 8 string length = 12crossover rate = 08 and mutation rate = 005
If 120579119894119901 119894 = 1 2 119873 and
120579119894119901 119894 = 1 2 119873 are real
measurement pitchrsquos values and estimated pitchrsquos values ofadaptive 120572
1-1205731-1205741filter respectively then 120579
119894119903 119894 = 1 2 119873
and 120579119894119903 119894 = 1 2 119873 are realmeasurement rollrsquos values and
estimated rollrsquos values of adaptive 1205721-1205731-1205741filter respectively
where 119894 is the current iteration and 119873 is the total number ofsampling data The RMSEs of pitch and roll estimations aredefined as follows
RMSE119901= radic
1
119873
119873
sum
119894 = 1
(120579119894119901minus 120579119894119901)
2
(18)
RMSE119903= radic
1
119873
119873
sum
119894 = 1
(120579119894119903minus 120579119894119903)
2
(19)
where the main purpose of GA and PSO methods is to findthe optimum values of 119892
1119901and 119892
1119903which minimize the
RMSE119901and RMSE
119903 respectively
If X119894and X
119894 119894 = 1 2 119873 are real measurement target
position vector values and estimated target position vector
values of adaptive 1205722-1205732-1205742filter the RMSE of target tracking
is defined as
RMSE = radic1
119873
119873
sum
119894 = 1
((119909119894minus 119909119894)2+ (119910119894minus 119910119894)2+ (119911119894minus 119894)2) (20)
where 119909119894 119910119894 and 119911
119894are the components of target position
vector X119894in 119909 119910 and 119911 axes The main purpose of GA and
PSO algorithms is to find the optimum gain value of g2which
minimizes the RMSE
4 Experimental Results
Six scenarios are used to simulate the tracking performanceof ship-borne phased array radar using the proposed adaptive120572-120573-120574 filter for beam pointing error compensation and targettracking in three-dimensional space
Case 1 (roll and pitch estimations using GA based adaptive120572-120573-120574 filter and measured data) The roll and pitch datarecorded by a gyroscope of the ship were used in the exper-iments The total number of data is 1000 and the samplingtime is 01 sec The angle variation of roll and pitch signalsmeasured by ship gyroscope are shown in Figure 7 whichwillresult in the beam pointing errors of phased array antennaTheGA based adaptive 120572-120573-120574 filter is used to estimate the rolland pitch angles Figures 8(a) and 8(b) show that the conver-gent time of roll and pitch angles are 3 sec and 5 sec respec-tively under simulated sea states 2 and 3 and measured dataThe shiprsquos roll and pitch signals are simulated according to (3)and (4) with roll and pitch angle parameters of sea states 2and 3 listed in Table 1 and standard deviation of 001 Itconcludes that the proposed GA based adaptive 120572-120573-120574 filter issuitable to compensate the ship motion As shown in Table 1the period of sea state 2 is longer than sea state 3 Thereforethe convergent RMSE (about 019∘ for roll about 01∘ forpitch) of sea state 2 is less than sea state 3 (about 029∘ for rollabout 022∘ for pitch) because the roll and pitch signals withshorter period in sea state 3 have rapid oscillations which aremore difficult to be predicted precisely
Case 2 (tracking linear moving target in three-dimensionalspace using GA based adaptive 120572-120573-120574 filter (without beampointing error)) Assuming the ship is not affected by thesea waves (antenna beam pointing error is zero) the trackingperformance of GA based adaptive 120572-120573-120574 filter for a straightflight target trajectory is simulated The initial position ofradar is (0 0 0) The ship moves with a speed of 10msec in
International Journal of Antennas and Propagation 9
0 100 200 300 400 500 600 700 800 900 1000
0
02
04
06
08
Pitc
h an
gle (
deg)
Time (10minus1 s)
minus12
minus08
minus06
minus04
minus02
minus1
(a)
0
1
2
3
Roll
angl
e (de
g)
0 100 200 300 400 500 600 700 800 900 1000Time (10minus1 s)
minus1
minus2
minus3
minus4
(b)
Figure 7 Demonstrating angle variation of (a) pitch and (b) roll signals measured by gyroscope
0
005
01
015
02
025
03
035
04
RMSE
(deg
)
Sea 3Sea 2Measurement
0 100 200 300 400 500 600 700 800 900 1000Time (10minus1 s)
(a)
0
005
01
015
02
025
03
035
04RM
SE (d
eg)
Sea 3Sea 2Measurement
Time (10minus1 s)0 100 200 300 400 500 600 700 800 900 1000
(b)
Figure 8 RMSE of (a) roll and (b) pitch under simulated data of sea states 2 and 3 and measured data
the 119884 axial direction The linear path equation of the ship isgiven by
X119904(119909119905 119910119905 119911119905) = X0119904+ V119904119905 (21)
where 119905 = 1 2 3 1000 sec V119904is ship velocity vector
and X0119904is initial ship position vector in three-dimensional
spaceThe radar position is updated every secondThe initialposition of flying target is (minus74840 minus129620 9100)m whichis about 150 km from the radar The flight speed of target is300msec (119883 axial velocity of 150msec and 119884 axial velocityof 260msec) The target location is updated every secondThe linear target trajectory equation is
X119898(119909119905 119910119905 119911119905) = X0119898
+ V119898119905 (22)
where V119898
is the target velocity vector and X0119898
is theinitial target position vector in three-dimensional space Thetracking accuracy of GA based adaptive 120572-120573-120574 filter for astraight flight target is shown in Figure 9 where the trajectoryestimation error is calculated by
119864119905= radic(119909
119905minus 119909119905)2+ (119910119905minus 119910119905)2+ (119911119905minus 119905)2 (23)
where (119909119905 119910119905 119911119905) and (119909
119905 119910119905 119905) denote the real target location
and estimated target location respectively It shows that theGA based adaptive 120572-120573-120574 filter converges to less than about2m after 5 sec
10 International Journal of Antennas and Propagation
Table 2 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0574 06399 0652 06904 06537 0674 06286 06313 06044 053921198921199011
0629 06664 0622 07756 0664 07399 0663 06652 06816 064981198922
0285 05827 03282 03851 00054 03480 02916 03736 0547 04864
Table 3 Estimated gain parameters for PSO based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0557 05986 0357 05345 06142 05774 05544 0529 04592 056261198921199011
0612 05898 04129 05675 05286 03953 05565 06199 05214 062851198922
0326 007 01136 0146 0098 02365 021 02547 02474 01245
0 20 40 60 80 100 1200
20
40
60
80
100
120
140
160
180
Time (s)
Et
(m)
Average estimation error = 15176m
Figure 9 Estimation error for linear target trajectory without beampointing error
Case 3 (tracking circular moving target in three-dimensionalspace using GA based adaptive 120572-120573-120574 filter (without beampointing error)) Figures 10(a) and 10(b) show the simulationscenario of beam pointing error compensation for circularmaneuvering air target and linear moving ship The shiphas the same linear moving speed and path equation asCase 2The flying target is parallel to the119883119884 plane the initialposition of flying target is (8885 4590 9100)m the velocity is300msec and the angle (120579
0) between the initial position and
the119883-axis is 30∘The trajectory of the flight vehicle is a circlewhich has a radius of 10 km The center of circular trajectoryis 119884-axis The rotational cycle time of the flight target is 209seconds and the total observation time is 419 sec
The circular trajectory equation is
X119898(119909119905 119910119905 119911119905) = (119903 cos (120579
0minus 120596119905) 119903 sin (120579
0minus 120596119905) 9100) (24)
where 119905 = 1 2 3 418 419 sec 119903 is the circular flight radiusin meter 120579
0is the initial angle between the flight vehicle
and the 119883-axis 120596 is the angular speed of flight vehicle Theaccuracy of GA based adaptive 120572-120573-120574 filter for tracking acircular flight target is shown in Figure 11 where theGAbasedadaptive 120572-120573-120574 filter converges to less than 3m after 8 secTherefore using the GA based adaptive 120572-120573-120574 filter to trackthe circular trajectory target has the larger average estimationerror and longer convergent time than linear trajectory target
Case 4 (estimating roll and pitch angles and tracking lineartrajectory target using GAPSO based adaptive 120572-120573-120574 filterand measured data (with beam pointing error)) The mea-sured roll and pitch signals shown in Figure 7 are used toevaluate the performance of GAPSO based adaptive 120572-120573-120574filter for estimating the roll and pitch angles and tracking thelinear trajectory targetThe beamof 6times6 planar array antennais steered to track the linear trajectory target The samplingfrequency is set as 10Hz Therefore the shiprsquos roll and pitchsignals are sampled 10000 points within 1000 seconds Asthe scenario described in Case 2 the ship is along the 119884
axis of 119883119884 plane in the earth coordinates the ship speedis 10msec and the flying target speed is 300msec Theflying target is parallel to the 119883119884 plane 91 km above theground and the angle between the flight direction and the119884-axis is 120∘ The largest reconnaissance distance of ship-borne radar is 150 km When the air target is flying withinthe radar detection range the beampointing error estimationand linear trajectory target tracking of ship-borne phasedarray radar are simulated
Figure 12 is a simulation flow chart used to verifythe effect of beam pointing error compensation on thetracking accuracy of adaptive 120572-120573-120574 filter Tables 2 and 3demonstrate the optimal gain parameters of adaptive 120572-120573-120574 filter for roll 119892
1199031 pitch 119892
1199011 and target tracking 119892
2
obtained by the GA and PSO methods respectively thatminimizes the RMSE Figures 13ndash17 show the estimationerrors for five different gain values which are summarizedin Table 4 It demonstrates that the greater gain value(119892 = 01 075 and 09) will generate the greater error andneed the longer convergence time The estimation errors
International Journal of Antennas and Propagation 11
Table 4 Estimation errors for different gains
Calculated by GA Calculated by PSO Fixed gain1198921199031
05392sim06904 0357sim06142 01 075 091198921199011
0622sim07756 03953sim06285 01 075 091198922
0054sim05827 007sim03259 01 075 09Average 119864
119905(m) 14781 163224 169663 534804 3991188
Table 5 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0543 05653 05424 05455 05665 056 058 0532 05414 053851198921199011
0464 0484 04459 0422 04667 04654 04509 04459 04308 046511198922
0161 00929 01937 00865 00241 00609 01174 01897 01713 01727
Y
X
Z
(a)
Y
X10 km
2730∘
(b)
Figure 10 Simulation scenarios of Case 3 for (a) 3D and (b) top view diagrams
0 10 20 30 40 50 60 70 80 90 1000
50
100
150
200
250
Time (s)
Et
(m)
Average estimation error = 24905m
Figure 11 Estimation error for circular target trajectory without beam pointing error
12 International Journal of Antennas and Propagation
Generate the target detection data
Using GA or PSO
to track the target
Generate beam steering angle
Measure or simulate roll and pitch angles
Using GA or PSO
roll and pitch angles
Simulate beam pointing error compensation
Verify tracking accuracy ofGAPSO adaptive 120572-120573-120574 filter
120572-120573-120574 filter 120572-120573-120574 filter to estimate
Figure 12 Flow chart of tracking performance simulation forCase 4experiment
0 100 200 300 400 500 600 700 800 900 10000
50
100
150
200
250
300
Time (s)
Et
(m)
Average estimation error = 14781m
Figure 13 Estimation error of GA based adaptive 120572-120573-120574 filter
presented in Figures 16 and 17 are too large to be appli-cable when the gain values are equal to 075 and 09The GA and PSO have smaller errors compared with thefixed gain values and the GA obtains the best estimationaccuracy The estimation errors of GA and PSO meth-ods were found to be close From Figures 13 14 and 15we can observe that there is a peak value that occurredat the time duration about 510 to 515 seconds becausethe horizontal beam direction of ship-borne phased array
0
50
100
150
200
250
300
Et
(m)
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 163224 m
Figure 14 Estimation error of PSO based adaptive 120572-120573-120574 filter
0
20
40
60
80
100
120
140
160
180
200Et
(m)
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 169663 m
Figure 15 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
radar is steered according to the linear trajectory targetwhich is flying through sky just above the ship at thetime instant of 510 second As shown in Figure 18 thehorizontal beam steering angle of ship-borne phased arrayradar at about the time of 510 second is changing fromdescending into ascending When the GA based adaptive120572-120573-120574 filter tracks the linear trajectory target the esti-mation error values are less than 30 meters or less ifthe peak estimation errors during these 5 seconds areignored
Case 5 (estimating roll and pitch angles and tracking lineartrajectory target using GAPSO based adaptive 120572-120573-120574 filterand simulated data (with beam pointing error)) The shiprsquosroll and pitch signals are simulated according to (3) and (4)with parameters of sea state 2 listed in Table 1 and standard
International Journal of Antennas and Propagation 13
0 200 400 600 800 10000
500
1000
1500
Time (s)
Et
(m)
Average estimation error = 534804 m
Figure 16 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0 200 400 600 800 10000
1000
2000
3000
4000
5000
6000
7000
8000
9000
Time (s)
Et
(m)
Average estimation error = 3991188 m
Figure 17 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
deviation of 001 The simulated shiprsquos roll and pitch signalsare applied for five different methods to estimate the gain ofGAPSO based adaptive 120572-120573-120574 filter The total observationtime is 1000 sec and sampling time is 01 sec The simulationreplicates 10000 times The input samples are updated byone new sample for the next iteration The other simulationscenario and parameters are the same asCase 4 Tables 5 and 6demonstrate the optimal gain parameters of adaptive 120572-120573-120574filter for roll gain 119892
1199031 pitch gain 119892
1199011 and target tracking gain
vector g2obtained by the GA and PSOmethods respectively
that minimizes RMSE Figures 19 20 21 22 and 23 showthe estimation errors for five different gain values which aresummarized in Table 7 The estimation error of GA basedadaptive 120572-120573-120574 filter for tracking a linear trajectory targetis shown in Figure 19 where the convergence time of GA
0 100 200 300 400 500 600 700 800 900 1000Time (s)
0
05
1
15
2
25
Azi
mut
h ste
erin
g an
gle (
rad)
With beam pointing errorWithout beam pointing error
Figure 18 Azimuth beam steering angle of ship-borne phased arrayradar
0
10
20
30
40
50
60
70
80
90
100
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 160756 m
Et
(m)
Figure 19 Estimation error of GA based adaptive 120572-120573-120574 filter
based adaptive 120572-120573-120574 filter is about 3 sec and the averageestimation error is about 160756m It concludes that theexperimental results in Case 5 have the same trend as inCase 4 therefore the simulated data can be used to observethe effect of shipmotion on the tracking performance of ship-borne phased array radar under different sea state conditionsThe convergence time and average estimation error of GAbased adaptive 120572-120573-120574 filter are better than themethod used in[6] where the tracking error of AEKF converges to less thanabout 20m at 550 iterations (seciteration) and the estimationerror will remain within the range of about 20m when thebeam pointing error is compensated by FBFN controller
Case 6 (estimating roll and pitch angles and tracking circularmoving target using GA based adaptive 120572-120573-120574 filter and
14 International Journal of Antennas and Propagation
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 187684 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 20 Estimation error of PSO based adaptive 120572-120573-120574 filter
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 199594 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 21 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
measured data (with beam pointing error)) As the scenariodescribed in Case 3 the measured roll and pitch signalsshown in Figure 7 are used to evaluate the performance ofGA based adaptive 120572-120573-120574 filter for estimating the roll andpitch angles and tracking the circular trajectory target Theestimation errors of 120572-120573-120574 filter with fixed gains of 119892
1199031=
1198921199011
= 1198922= 01 and 119892
1199031= 1198921199011
= 1198922= 075 are shown in
Figures 24 and 25 respectivelyThe average estimation errorsfor fixed filter gains of 01 and 075 are about 41638m and572386m respectively The optimal roll gain 119892
1199031 pitch gain
1198921199011 and target tracking gain 119892
2of adaptive 120572-120573-120574 filter
estimated by the GA method during the observation timeof 419 sec are listed in Table 8 The estimation error of GAbased adaptive 120572-120573-120574 filter for tracking a circular flight targetis shown in Figure 26 where the convergence time of GAbased adaptive 120572-120573-120574 filter is about 5 sec and the average
0
500
1000
1500
0 200 400 600 800 1000Time (s)
Average estimation error = 1080515 m
Et
(m)
Figure 22 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
0 200 400 600 800 1000Time (s)
Average estimation error = 4279449 m
Et
(m)
Figure 23 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
estimation error is about 36667m It concludes that the GAbased adaptive 120572-120573-120574 filter can greatly improve the trackingaccuracy and convergence time of the conventional 120572-120573-120574filter for tracking the circular trajectory target
5 Conclusions
This paper proposed an intelligent beam pointing error com-pensationmechanism for ship-borne phased array radarTheGA based adaptive 120572-120573-120574 filter estimates the roll and pitchangles of the ship moving in the sea and thus compensatesfor the antenna beam pointing error in order to enhance thetracking accuracy of phased array radar system
International Journal of Antennas and Propagation 15
Table 6 Estimated gain parameters for PSO based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0366 03645 0532 04519 05151 05002 04797 05254 04146 053511198921199011
0366 03647 03634 03644 03643 0363 03644 03684 03643 036421198922
0125 01271 02286 02166 01169 01799 01735 01564 00767 01346
Table 7 Estimation errors for different gains
Calculated by GA Calculated by PSO Fixed gain1198921199031
0459sim0583 03645sim05352 01 075 091198921199011
0498sim0667 0363sim03684 01 075 091198922
00586sim01697 00367sim02291 01 075 09Average 119864
119905(m) 160756 187684 199594 1080515 4279449
Table 8 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim42 43sim84 85sim126 127sim168 169sim210 211sim252 253sim294 295sim336 337sim378 379sim4191198921199031
0619 04996 05316 0644 05052 05568 05316 04459 05092 059241198921199011
0582 06727 07201 04913 04168 07169 05983 02036 05858 06211198922
0214 02215 01612 02474 01822 0157 02073 01211 02278 01074
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 41638 m
Figure 24 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
Six experimental cases are conducted to verify the perfor-mance of ship-borne phased array radar using the proposedGA based adaptive 120572-120573-120574 filter In Case 1 the GA basedadaptive 120572-120573-120574 filter is used to estimate the roll and pitchangles with the simulated and measured roll and pitchsignals respectively The simulation results show that themeasurement data has the minimum estimation error theestimation error in the sea state 2 is the second and theestimation error in the sea state 3 is the worst Cases 2 and 3show that when the GA based adaptive 120572-120573-120574 filter is appliedto estimate the beam pointing error and to track the targetin a ship-borne phased array radar the circular trajectory
0 50 100 150 200 250 300 350 400 450Time (s)
0
500
1000
1500
Average estimation error = 572386 m
Et
(m)
Figure 25 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
target will be tracked with the larger average estimation errorand longer convergence time than linear trajectory targetTheexperimental results ofCases 4 and 5demonstrate thatwhen alinear trajectory target is tracked by ship-borne phased arrayradar the average estimation error and convergence time ofship-borne phased array radar using the GA based adaptive120572-120573-120574 filter are better than PSO based adaptive 120572-120573-120574 filterand fixed filter gain 120572-120573-120574 filter The convergence time andtracking accuracy of ship-borne phased array radar usingthe proposed GA based adaptive 120572-120573-120574 filter are superiorto the AEKF significantly In conclusions the proposed GAbased adaptive 120572-120573-120574 filter is a real time applicable algorithm
16 International Journal of Antennas and Propagation
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 36667 m
Figure 26 Estimation error of GA based adaptive 120572-120573-120574 filter
whose accuracy is good with relatively less complexity Inaddition the roll and pitch signalsmeasured in real operationenvironments can be replaced with the simulated roll andpitch signals to test all the relevant properties of practical shipmotion compensation and air target tracking for ship-bornephased array radar
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
This work is supported in part by research grants fromNational Science Council China (NSC 102-2218-E-155-001)
References
[1] Z Fang and Q Rundong Ship-borne Phased Array RadarMotion Compensation System Engineering and ElectronicTechnologies 1998
[2] L J Love J F Jansen and F G Pin ldquoCompensation of waveinduced motion and force phenomenon for ship based highperformance robotic and human amplifying systemsrdquo OakRidge National Laboratory ORNLTM-2003233 2003
[3] R J Hill Motion Compensation for Ship Borne Radars andLidars US Department of Commerce National Oceanic andAtmospheric Administration Office of Oceanic and Atmo-spheric Research Earth System Research Laboratory PhysicalSciences Division 2005
[4] T I Fossen and T Perez ldquoKalman filtering for positioning andheading control of ships and offshore rigsrdquo IEEEControl SystemsMagazine vol 29 no 6 pp 32ndash46 2009
[5] S Kuchler C Pregizer J K Eberharter K Schneider and OSawodny ldquoReal-time estimation of a shiprsquos attituderdquo in Proceed-ings of theAmericanControl Conference (ACC rsquo11) pp 2411ndash2416San Francisco Calif USA July 2011
[6] J Mar K C Tsai Y T Wang and M B Basnet ldquoIntelligentmotion compensation for improving the tracking performanceof shipborne phased array radarrdquo International Journal ofAntennas and Propagation vol 2013 Article ID 384756 14pages 2013
[7] T Dirk and S Tarunraj ldquoOptimal design of 120572-120573-(120574) filtersrdquo inAmerican Control Conference vol 6 pp 4348ndash4352 ChicagoIll USA June 2000
[8] D Tenne and T Singh ldquoCharacterizing performance of 120572-120573-120574filtersrdquo IEEE Transactions on Aerospace and Electronic Systemsvol 38 no 3 pp 1072ndash1087 2002
[9] T E Lee J P Su and K W Yu ldquoParameter optimization fora third-order sampled-data trackerrdquo in Proceedings of the 2ndInternational Conference on Innovative Computing Informationand Control p 336 Kumamoto Japan September 2007
[10] Y H Lho and J H Painter ldquoA fuzzy tuned adaptive Kalmanfilterrdquo in Industrial FuzzyControl and Intelligent System pp 144ndash148 December 1993
[11] N Ramakoti A Vinay and R K Jatoth ldquoParticle swarm opti-mization aided Kalman filter for object trackingrdquo in Proceedingsof the International Conference on Advances in ComputingControl and Telecommunication Technologies (ACT rsquo09) pp 531ndash533 Kerela India December 2009
[12] T-E Lee J-P Su andK-WYu ldquoAparticle swarmoptimization-based state estimation scheme formoving objectsrdquoTransactionsof the Institute of Measurement and Control vol 34 no 2-3 pp236ndash254 2012
[13] D C Law S A McLaughlin M J Post et al ldquoAn electronicallystabilized phased array system for shipborne atmospheric windprofilingrdquo Journal of Atmospheric and Oceanic Technology vol19 no 6 pp 924ndash933 2002
[14] C A Balanis AntennaTheory Analysis and Design JohnWileyamp Sons New York NY USA 3rd edition 2005
[15] J Mar and Y R Lin ldquoImplementation of SDR digital beam-former for microsatellite SARrdquo IEEE Geoscience and RemoteSensing Letters vol 6 no 1 pp 92ndash96 2009
[16] R C Eberhart and Y Shi Computational Intelligence Conceptsto Implementations ElsevierMorgan Kaufmann 2007
[17] K Y Lee and J-B Park ldquoApplication of particle swarm optimi-zation to economic dispatch problem advantages anddisadvan-tagesrdquo in Proceedings of the IEEE PES Power Systems Conferenceand Exposition (PSCE rsquo06) pp 188ndash192 Atlanta Ga USANovember 2006
[18] R C Eberhart and J Kennedy ldquoNew optimizer using particleswarm theoryrdquo in Proceedings of the 6th International Sympo-sium onMicroMachine and Human Science pp 39ndash43 NagoyaJapan October 1995
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
International Journal of Antennas and Propagation 7
Initialize particles with random positions and random velocities
For each particle calculate the fitness values
Is fitness value ltbest fitness value
Particle exhausted
Set global best fitness value as best fitness value
Update particle positions and velocities using (18) and (20)
Maximum iterationreached
Give global best fitnessvalue and position
Set global best as local best fitness value
Yes
No
No
No
Yes
Yes
Figure 5 Flow chart of PSO method
where 119909119894119889(119899)
and V119894119889(119899)
are position and velocity of (119894119889)thparticle at (119899)th iteration respectively 119901
119894119889(119899)and 119901
119892119889(119899)are
the best position and global best position of (119894119889)th particle at(119899)th iteration respectively Rand( ) and rand( ) are randomnumber between 0 and 1 respectively Equation (15) updatesthe velocity of the particles and (17) updates the position ofparticles 119870 means inertia 119888
1and 1198882are correction factors
and swarm size means number of particles for an iterationas described in [16]
Particles in the PSO method sometimes go out of con-straints while changing their position Reinitialization of theparticlersquos position randomly to fall within the constraints isdone as soon as a particle moves out of constraints Thisensures full use of particles and also increases the chance ofdiscovering a better solution To reduce the chance of gettingtrapped in a local minimum the advantages of using PSOmethod are that it is derivative-free easy to implement andeasy to comprehend and the solution is independent of theinitial point [17] One of the drawbacks of the PSOmethod isthat it does not guarantee success that is the solution foundmay not be the best solution
35 GAPSO Based Adaptive 120572-120573-120574 Filter The flow chart ofthe adaptive 120572-120573-120574 filter algorithm is shown in Figure 6where 119873 is the number of sampling data to determine thefilter gain In this algorithm the parameters120572120573 and 120574 whichare related to the filter gain 119892 are optimized by minimizingthe RMSE using GA or PSO method The proposed adaptive120572-120573-120574 filter processes the current sampled data 119909
119894which
predicts the next sample data 119909119894+1
as iteration continuesThe adaptive GA or PSO process only occurs at certain timeintervalThe previous119873 data that is 119909
119901(119894minus119873) sdot sdot sdot 119909
119901(119894minus1) is
sent to the adaptiveGAor PSO algorithmwhere the optimumvalue of filter gain (119892) is determined for the 120572-120573-120574 filterThen 119909
119894+1is predicted and the process continues The value
of119873 determines running time and accuracy of the algorithmso 119873 must have small enough value to support real timeimplementation and large enough value to provide acceptableaccuracy
Since the roll and pitch angles of ship motion changeover time randomly in order to speed up the processingtime and to reach the objectives of minimum estimationerror the filter gain values are real time updated by using thepipelined architectureThe recorded roll and pitch signals are
8 International Journal of Antennas and Propagation
Collect previous N data using PSO or GA algorithm
Predicted value for120572-120573-120574 filterxi
xi+1
ximinus1 ximinusFind g for Nximinus1 ximinusN
that minimizes RMSE
Figure 6 Block diagram of adaptive 120572-120573-120574 filter
segmented into block of 1000 sampled data (100 sec) for lineartrajectory target and 420 sampled data (42 sec) for circulartrajectory target respectivelyThe gain value of 120572
2-1205732-1205742filter
in 101ndash200 sec is estimated by GA and PSO methods usingthe collected data during 1ndash100 sec the gain value of 120572
2-
1205732-1205742filter in 201ndash300 sec is estimated by the GA and PSO
methods using the collected data in 101ndash200 sec and so onSimilarly the measured flight target signals are segmentedinto block of 100 sampled data (100 sec) for linear trajectorytarget and 42 sampled data (42 sec) for circular trajectorytarget respectively The gain value of 120572
2-1205732-1205742filter in 101ndash
200 sec is estimated by the GA and PSO methods using thecollected data during 1ndash100 sec the gain value of 120572
2-1205732-1205742
filter in the interval of 201ndash300 sec is estimated by the GA andPSO methods using the collected data in 101ndash200 sec and soon
For the PSO process it was found that gain parameterconverged within 100 iterations so the number of iterationsfor an adaptive 120572-120573-120574 filter was kept at constant 100 iterationsper processing cycle swarm size = 30 inertia = 07298 andcorrection factors 119888
1= 21 119888
2= 2 For the GA process we
have considered population size = 8 string length = 12crossover rate = 08 and mutation rate = 005
If 120579119894119901 119894 = 1 2 119873 and
120579119894119901 119894 = 1 2 119873 are real
measurement pitchrsquos values and estimated pitchrsquos values ofadaptive 120572
1-1205731-1205741filter respectively then 120579
119894119903 119894 = 1 2 119873
and 120579119894119903 119894 = 1 2 119873 are realmeasurement rollrsquos values and
estimated rollrsquos values of adaptive 1205721-1205731-1205741filter respectively
where 119894 is the current iteration and 119873 is the total number ofsampling data The RMSEs of pitch and roll estimations aredefined as follows
RMSE119901= radic
1
119873
119873
sum
119894 = 1
(120579119894119901minus 120579119894119901)
2
(18)
RMSE119903= radic
1
119873
119873
sum
119894 = 1
(120579119894119903minus 120579119894119903)
2
(19)
where the main purpose of GA and PSO methods is to findthe optimum values of 119892
1119901and 119892
1119903which minimize the
RMSE119901and RMSE
119903 respectively
If X119894and X
119894 119894 = 1 2 119873 are real measurement target
position vector values and estimated target position vector
values of adaptive 1205722-1205732-1205742filter the RMSE of target tracking
is defined as
RMSE = radic1
119873
119873
sum
119894 = 1
((119909119894minus 119909119894)2+ (119910119894minus 119910119894)2+ (119911119894minus 119894)2) (20)
where 119909119894 119910119894 and 119911
119894are the components of target position
vector X119894in 119909 119910 and 119911 axes The main purpose of GA and
PSO algorithms is to find the optimum gain value of g2which
minimizes the RMSE
4 Experimental Results
Six scenarios are used to simulate the tracking performanceof ship-borne phased array radar using the proposed adaptive120572-120573-120574 filter for beam pointing error compensation and targettracking in three-dimensional space
Case 1 (roll and pitch estimations using GA based adaptive120572-120573-120574 filter and measured data) The roll and pitch datarecorded by a gyroscope of the ship were used in the exper-iments The total number of data is 1000 and the samplingtime is 01 sec The angle variation of roll and pitch signalsmeasured by ship gyroscope are shown in Figure 7 whichwillresult in the beam pointing errors of phased array antennaTheGA based adaptive 120572-120573-120574 filter is used to estimate the rolland pitch angles Figures 8(a) and 8(b) show that the conver-gent time of roll and pitch angles are 3 sec and 5 sec respec-tively under simulated sea states 2 and 3 and measured dataThe shiprsquos roll and pitch signals are simulated according to (3)and (4) with roll and pitch angle parameters of sea states 2and 3 listed in Table 1 and standard deviation of 001 Itconcludes that the proposed GA based adaptive 120572-120573-120574 filter issuitable to compensate the ship motion As shown in Table 1the period of sea state 2 is longer than sea state 3 Thereforethe convergent RMSE (about 019∘ for roll about 01∘ forpitch) of sea state 2 is less than sea state 3 (about 029∘ for rollabout 022∘ for pitch) because the roll and pitch signals withshorter period in sea state 3 have rapid oscillations which aremore difficult to be predicted precisely
Case 2 (tracking linear moving target in three-dimensionalspace using GA based adaptive 120572-120573-120574 filter (without beampointing error)) Assuming the ship is not affected by thesea waves (antenna beam pointing error is zero) the trackingperformance of GA based adaptive 120572-120573-120574 filter for a straightflight target trajectory is simulated The initial position ofradar is (0 0 0) The ship moves with a speed of 10msec in
International Journal of Antennas and Propagation 9
0 100 200 300 400 500 600 700 800 900 1000
0
02
04
06
08
Pitc
h an
gle (
deg)
Time (10minus1 s)
minus12
minus08
minus06
minus04
minus02
minus1
(a)
0
1
2
3
Roll
angl
e (de
g)
0 100 200 300 400 500 600 700 800 900 1000Time (10minus1 s)
minus1
minus2
minus3
minus4
(b)
Figure 7 Demonstrating angle variation of (a) pitch and (b) roll signals measured by gyroscope
0
005
01
015
02
025
03
035
04
RMSE
(deg
)
Sea 3Sea 2Measurement
0 100 200 300 400 500 600 700 800 900 1000Time (10minus1 s)
(a)
0
005
01
015
02
025
03
035
04RM
SE (d
eg)
Sea 3Sea 2Measurement
Time (10minus1 s)0 100 200 300 400 500 600 700 800 900 1000
(b)
Figure 8 RMSE of (a) roll and (b) pitch under simulated data of sea states 2 and 3 and measured data
the 119884 axial direction The linear path equation of the ship isgiven by
X119904(119909119905 119910119905 119911119905) = X0119904+ V119904119905 (21)
where 119905 = 1 2 3 1000 sec V119904is ship velocity vector
and X0119904is initial ship position vector in three-dimensional
spaceThe radar position is updated every secondThe initialposition of flying target is (minus74840 minus129620 9100)m whichis about 150 km from the radar The flight speed of target is300msec (119883 axial velocity of 150msec and 119884 axial velocityof 260msec) The target location is updated every secondThe linear target trajectory equation is
X119898(119909119905 119910119905 119911119905) = X0119898
+ V119898119905 (22)
where V119898
is the target velocity vector and X0119898
is theinitial target position vector in three-dimensional space Thetracking accuracy of GA based adaptive 120572-120573-120574 filter for astraight flight target is shown in Figure 9 where the trajectoryestimation error is calculated by
119864119905= radic(119909
119905minus 119909119905)2+ (119910119905minus 119910119905)2+ (119911119905minus 119905)2 (23)
where (119909119905 119910119905 119911119905) and (119909
119905 119910119905 119905) denote the real target location
and estimated target location respectively It shows that theGA based adaptive 120572-120573-120574 filter converges to less than about2m after 5 sec
10 International Journal of Antennas and Propagation
Table 2 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0574 06399 0652 06904 06537 0674 06286 06313 06044 053921198921199011
0629 06664 0622 07756 0664 07399 0663 06652 06816 064981198922
0285 05827 03282 03851 00054 03480 02916 03736 0547 04864
Table 3 Estimated gain parameters for PSO based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0557 05986 0357 05345 06142 05774 05544 0529 04592 056261198921199011
0612 05898 04129 05675 05286 03953 05565 06199 05214 062851198922
0326 007 01136 0146 0098 02365 021 02547 02474 01245
0 20 40 60 80 100 1200
20
40
60
80
100
120
140
160
180
Time (s)
Et
(m)
Average estimation error = 15176m
Figure 9 Estimation error for linear target trajectory without beampointing error
Case 3 (tracking circular moving target in three-dimensionalspace using GA based adaptive 120572-120573-120574 filter (without beampointing error)) Figures 10(a) and 10(b) show the simulationscenario of beam pointing error compensation for circularmaneuvering air target and linear moving ship The shiphas the same linear moving speed and path equation asCase 2The flying target is parallel to the119883119884 plane the initialposition of flying target is (8885 4590 9100)m the velocity is300msec and the angle (120579
0) between the initial position and
the119883-axis is 30∘The trajectory of the flight vehicle is a circlewhich has a radius of 10 km The center of circular trajectoryis 119884-axis The rotational cycle time of the flight target is 209seconds and the total observation time is 419 sec
The circular trajectory equation is
X119898(119909119905 119910119905 119911119905) = (119903 cos (120579
0minus 120596119905) 119903 sin (120579
0minus 120596119905) 9100) (24)
where 119905 = 1 2 3 418 419 sec 119903 is the circular flight radiusin meter 120579
0is the initial angle between the flight vehicle
and the 119883-axis 120596 is the angular speed of flight vehicle Theaccuracy of GA based adaptive 120572-120573-120574 filter for tracking acircular flight target is shown in Figure 11 where theGAbasedadaptive 120572-120573-120574 filter converges to less than 3m after 8 secTherefore using the GA based adaptive 120572-120573-120574 filter to trackthe circular trajectory target has the larger average estimationerror and longer convergent time than linear trajectory target
Case 4 (estimating roll and pitch angles and tracking lineartrajectory target using GAPSO based adaptive 120572-120573-120574 filterand measured data (with beam pointing error)) The mea-sured roll and pitch signals shown in Figure 7 are used toevaluate the performance of GAPSO based adaptive 120572-120573-120574filter for estimating the roll and pitch angles and tracking thelinear trajectory targetThe beamof 6times6 planar array antennais steered to track the linear trajectory target The samplingfrequency is set as 10Hz Therefore the shiprsquos roll and pitchsignals are sampled 10000 points within 1000 seconds Asthe scenario described in Case 2 the ship is along the 119884
axis of 119883119884 plane in the earth coordinates the ship speedis 10msec and the flying target speed is 300msec Theflying target is parallel to the 119883119884 plane 91 km above theground and the angle between the flight direction and the119884-axis is 120∘ The largest reconnaissance distance of ship-borne radar is 150 km When the air target is flying withinthe radar detection range the beampointing error estimationand linear trajectory target tracking of ship-borne phasedarray radar are simulated
Figure 12 is a simulation flow chart used to verifythe effect of beam pointing error compensation on thetracking accuracy of adaptive 120572-120573-120574 filter Tables 2 and 3demonstrate the optimal gain parameters of adaptive 120572-120573-120574 filter for roll 119892
1199031 pitch 119892
1199011 and target tracking 119892
2
obtained by the GA and PSO methods respectively thatminimizes the RMSE Figures 13ndash17 show the estimationerrors for five different gain values which are summarizedin Table 4 It demonstrates that the greater gain value(119892 = 01 075 and 09) will generate the greater error andneed the longer convergence time The estimation errors
International Journal of Antennas and Propagation 11
Table 4 Estimation errors for different gains
Calculated by GA Calculated by PSO Fixed gain1198921199031
05392sim06904 0357sim06142 01 075 091198921199011
0622sim07756 03953sim06285 01 075 091198922
0054sim05827 007sim03259 01 075 09Average 119864
119905(m) 14781 163224 169663 534804 3991188
Table 5 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0543 05653 05424 05455 05665 056 058 0532 05414 053851198921199011
0464 0484 04459 0422 04667 04654 04509 04459 04308 046511198922
0161 00929 01937 00865 00241 00609 01174 01897 01713 01727
Y
X
Z
(a)
Y
X10 km
2730∘
(b)
Figure 10 Simulation scenarios of Case 3 for (a) 3D and (b) top view diagrams
0 10 20 30 40 50 60 70 80 90 1000
50
100
150
200
250
Time (s)
Et
(m)
Average estimation error = 24905m
Figure 11 Estimation error for circular target trajectory without beam pointing error
12 International Journal of Antennas and Propagation
Generate the target detection data
Using GA or PSO
to track the target
Generate beam steering angle
Measure or simulate roll and pitch angles
Using GA or PSO
roll and pitch angles
Simulate beam pointing error compensation
Verify tracking accuracy ofGAPSO adaptive 120572-120573-120574 filter
120572-120573-120574 filter 120572-120573-120574 filter to estimate
Figure 12 Flow chart of tracking performance simulation forCase 4experiment
0 100 200 300 400 500 600 700 800 900 10000
50
100
150
200
250
300
Time (s)
Et
(m)
Average estimation error = 14781m
Figure 13 Estimation error of GA based adaptive 120572-120573-120574 filter
presented in Figures 16 and 17 are too large to be appli-cable when the gain values are equal to 075 and 09The GA and PSO have smaller errors compared with thefixed gain values and the GA obtains the best estimationaccuracy The estimation errors of GA and PSO meth-ods were found to be close From Figures 13 14 and 15we can observe that there is a peak value that occurredat the time duration about 510 to 515 seconds becausethe horizontal beam direction of ship-borne phased array
0
50
100
150
200
250
300
Et
(m)
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 163224 m
Figure 14 Estimation error of PSO based adaptive 120572-120573-120574 filter
0
20
40
60
80
100
120
140
160
180
200Et
(m)
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 169663 m
Figure 15 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
radar is steered according to the linear trajectory targetwhich is flying through sky just above the ship at thetime instant of 510 second As shown in Figure 18 thehorizontal beam steering angle of ship-borne phased arrayradar at about the time of 510 second is changing fromdescending into ascending When the GA based adaptive120572-120573-120574 filter tracks the linear trajectory target the esti-mation error values are less than 30 meters or less ifthe peak estimation errors during these 5 seconds areignored
Case 5 (estimating roll and pitch angles and tracking lineartrajectory target using GAPSO based adaptive 120572-120573-120574 filterand simulated data (with beam pointing error)) The shiprsquosroll and pitch signals are simulated according to (3) and (4)with parameters of sea state 2 listed in Table 1 and standard
International Journal of Antennas and Propagation 13
0 200 400 600 800 10000
500
1000
1500
Time (s)
Et
(m)
Average estimation error = 534804 m
Figure 16 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0 200 400 600 800 10000
1000
2000
3000
4000
5000
6000
7000
8000
9000
Time (s)
Et
(m)
Average estimation error = 3991188 m
Figure 17 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
deviation of 001 The simulated shiprsquos roll and pitch signalsare applied for five different methods to estimate the gain ofGAPSO based adaptive 120572-120573-120574 filter The total observationtime is 1000 sec and sampling time is 01 sec The simulationreplicates 10000 times The input samples are updated byone new sample for the next iteration The other simulationscenario and parameters are the same asCase 4 Tables 5 and 6demonstrate the optimal gain parameters of adaptive 120572-120573-120574filter for roll gain 119892
1199031 pitch gain 119892
1199011 and target tracking gain
vector g2obtained by the GA and PSOmethods respectively
that minimizes RMSE Figures 19 20 21 22 and 23 showthe estimation errors for five different gain values which aresummarized in Table 7 The estimation error of GA basedadaptive 120572-120573-120574 filter for tracking a linear trajectory targetis shown in Figure 19 where the convergence time of GA
0 100 200 300 400 500 600 700 800 900 1000Time (s)
0
05
1
15
2
25
Azi
mut
h ste
erin
g an
gle (
rad)
With beam pointing errorWithout beam pointing error
Figure 18 Azimuth beam steering angle of ship-borne phased arrayradar
0
10
20
30
40
50
60
70
80
90
100
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 160756 m
Et
(m)
Figure 19 Estimation error of GA based adaptive 120572-120573-120574 filter
based adaptive 120572-120573-120574 filter is about 3 sec and the averageestimation error is about 160756m It concludes that theexperimental results in Case 5 have the same trend as inCase 4 therefore the simulated data can be used to observethe effect of shipmotion on the tracking performance of ship-borne phased array radar under different sea state conditionsThe convergence time and average estimation error of GAbased adaptive 120572-120573-120574 filter are better than themethod used in[6] where the tracking error of AEKF converges to less thanabout 20m at 550 iterations (seciteration) and the estimationerror will remain within the range of about 20m when thebeam pointing error is compensated by FBFN controller
Case 6 (estimating roll and pitch angles and tracking circularmoving target using GA based adaptive 120572-120573-120574 filter and
14 International Journal of Antennas and Propagation
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 187684 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 20 Estimation error of PSO based adaptive 120572-120573-120574 filter
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 199594 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 21 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
measured data (with beam pointing error)) As the scenariodescribed in Case 3 the measured roll and pitch signalsshown in Figure 7 are used to evaluate the performance ofGA based adaptive 120572-120573-120574 filter for estimating the roll andpitch angles and tracking the circular trajectory target Theestimation errors of 120572-120573-120574 filter with fixed gains of 119892
1199031=
1198921199011
= 1198922= 01 and 119892
1199031= 1198921199011
= 1198922= 075 are shown in
Figures 24 and 25 respectivelyThe average estimation errorsfor fixed filter gains of 01 and 075 are about 41638m and572386m respectively The optimal roll gain 119892
1199031 pitch gain
1198921199011 and target tracking gain 119892
2of adaptive 120572-120573-120574 filter
estimated by the GA method during the observation timeof 419 sec are listed in Table 8 The estimation error of GAbased adaptive 120572-120573-120574 filter for tracking a circular flight targetis shown in Figure 26 where the convergence time of GAbased adaptive 120572-120573-120574 filter is about 5 sec and the average
0
500
1000
1500
0 200 400 600 800 1000Time (s)
Average estimation error = 1080515 m
Et
(m)
Figure 22 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
0 200 400 600 800 1000Time (s)
Average estimation error = 4279449 m
Et
(m)
Figure 23 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
estimation error is about 36667m It concludes that the GAbased adaptive 120572-120573-120574 filter can greatly improve the trackingaccuracy and convergence time of the conventional 120572-120573-120574filter for tracking the circular trajectory target
5 Conclusions
This paper proposed an intelligent beam pointing error com-pensationmechanism for ship-borne phased array radarTheGA based adaptive 120572-120573-120574 filter estimates the roll and pitchangles of the ship moving in the sea and thus compensatesfor the antenna beam pointing error in order to enhance thetracking accuracy of phased array radar system
International Journal of Antennas and Propagation 15
Table 6 Estimated gain parameters for PSO based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0366 03645 0532 04519 05151 05002 04797 05254 04146 053511198921199011
0366 03647 03634 03644 03643 0363 03644 03684 03643 036421198922
0125 01271 02286 02166 01169 01799 01735 01564 00767 01346
Table 7 Estimation errors for different gains
Calculated by GA Calculated by PSO Fixed gain1198921199031
0459sim0583 03645sim05352 01 075 091198921199011
0498sim0667 0363sim03684 01 075 091198922
00586sim01697 00367sim02291 01 075 09Average 119864
119905(m) 160756 187684 199594 1080515 4279449
Table 8 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim42 43sim84 85sim126 127sim168 169sim210 211sim252 253sim294 295sim336 337sim378 379sim4191198921199031
0619 04996 05316 0644 05052 05568 05316 04459 05092 059241198921199011
0582 06727 07201 04913 04168 07169 05983 02036 05858 06211198922
0214 02215 01612 02474 01822 0157 02073 01211 02278 01074
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 41638 m
Figure 24 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
Six experimental cases are conducted to verify the perfor-mance of ship-borne phased array radar using the proposedGA based adaptive 120572-120573-120574 filter In Case 1 the GA basedadaptive 120572-120573-120574 filter is used to estimate the roll and pitchangles with the simulated and measured roll and pitchsignals respectively The simulation results show that themeasurement data has the minimum estimation error theestimation error in the sea state 2 is the second and theestimation error in the sea state 3 is the worst Cases 2 and 3show that when the GA based adaptive 120572-120573-120574 filter is appliedto estimate the beam pointing error and to track the targetin a ship-borne phased array radar the circular trajectory
0 50 100 150 200 250 300 350 400 450Time (s)
0
500
1000
1500
Average estimation error = 572386 m
Et
(m)
Figure 25 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
target will be tracked with the larger average estimation errorand longer convergence time than linear trajectory targetTheexperimental results ofCases 4 and 5demonstrate thatwhen alinear trajectory target is tracked by ship-borne phased arrayradar the average estimation error and convergence time ofship-borne phased array radar using the GA based adaptive120572-120573-120574 filter are better than PSO based adaptive 120572-120573-120574 filterand fixed filter gain 120572-120573-120574 filter The convergence time andtracking accuracy of ship-borne phased array radar usingthe proposed GA based adaptive 120572-120573-120574 filter are superiorto the AEKF significantly In conclusions the proposed GAbased adaptive 120572-120573-120574 filter is a real time applicable algorithm
16 International Journal of Antennas and Propagation
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 36667 m
Figure 26 Estimation error of GA based adaptive 120572-120573-120574 filter
whose accuracy is good with relatively less complexity Inaddition the roll and pitch signalsmeasured in real operationenvironments can be replaced with the simulated roll andpitch signals to test all the relevant properties of practical shipmotion compensation and air target tracking for ship-bornephased array radar
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
This work is supported in part by research grants fromNational Science Council China (NSC 102-2218-E-155-001)
References
[1] Z Fang and Q Rundong Ship-borne Phased Array RadarMotion Compensation System Engineering and ElectronicTechnologies 1998
[2] L J Love J F Jansen and F G Pin ldquoCompensation of waveinduced motion and force phenomenon for ship based highperformance robotic and human amplifying systemsrdquo OakRidge National Laboratory ORNLTM-2003233 2003
[3] R J Hill Motion Compensation for Ship Borne Radars andLidars US Department of Commerce National Oceanic andAtmospheric Administration Office of Oceanic and Atmo-spheric Research Earth System Research Laboratory PhysicalSciences Division 2005
[4] T I Fossen and T Perez ldquoKalman filtering for positioning andheading control of ships and offshore rigsrdquo IEEEControl SystemsMagazine vol 29 no 6 pp 32ndash46 2009
[5] S Kuchler C Pregizer J K Eberharter K Schneider and OSawodny ldquoReal-time estimation of a shiprsquos attituderdquo in Proceed-ings of theAmericanControl Conference (ACC rsquo11) pp 2411ndash2416San Francisco Calif USA July 2011
[6] J Mar K C Tsai Y T Wang and M B Basnet ldquoIntelligentmotion compensation for improving the tracking performanceof shipborne phased array radarrdquo International Journal ofAntennas and Propagation vol 2013 Article ID 384756 14pages 2013
[7] T Dirk and S Tarunraj ldquoOptimal design of 120572-120573-(120574) filtersrdquo inAmerican Control Conference vol 6 pp 4348ndash4352 ChicagoIll USA June 2000
[8] D Tenne and T Singh ldquoCharacterizing performance of 120572-120573-120574filtersrdquo IEEE Transactions on Aerospace and Electronic Systemsvol 38 no 3 pp 1072ndash1087 2002
[9] T E Lee J P Su and K W Yu ldquoParameter optimization fora third-order sampled-data trackerrdquo in Proceedings of the 2ndInternational Conference on Innovative Computing Informationand Control p 336 Kumamoto Japan September 2007
[10] Y H Lho and J H Painter ldquoA fuzzy tuned adaptive Kalmanfilterrdquo in Industrial FuzzyControl and Intelligent System pp 144ndash148 December 1993
[11] N Ramakoti A Vinay and R K Jatoth ldquoParticle swarm opti-mization aided Kalman filter for object trackingrdquo in Proceedingsof the International Conference on Advances in ComputingControl and Telecommunication Technologies (ACT rsquo09) pp 531ndash533 Kerela India December 2009
[12] T-E Lee J-P Su andK-WYu ldquoAparticle swarmoptimization-based state estimation scheme formoving objectsrdquoTransactionsof the Institute of Measurement and Control vol 34 no 2-3 pp236ndash254 2012
[13] D C Law S A McLaughlin M J Post et al ldquoAn electronicallystabilized phased array system for shipborne atmospheric windprofilingrdquo Journal of Atmospheric and Oceanic Technology vol19 no 6 pp 924ndash933 2002
[14] C A Balanis AntennaTheory Analysis and Design JohnWileyamp Sons New York NY USA 3rd edition 2005
[15] J Mar and Y R Lin ldquoImplementation of SDR digital beam-former for microsatellite SARrdquo IEEE Geoscience and RemoteSensing Letters vol 6 no 1 pp 92ndash96 2009
[16] R C Eberhart and Y Shi Computational Intelligence Conceptsto Implementations ElsevierMorgan Kaufmann 2007
[17] K Y Lee and J-B Park ldquoApplication of particle swarm optimi-zation to economic dispatch problem advantages anddisadvan-tagesrdquo in Proceedings of the IEEE PES Power Systems Conferenceand Exposition (PSCE rsquo06) pp 188ndash192 Atlanta Ga USANovember 2006
[18] R C Eberhart and J Kennedy ldquoNew optimizer using particleswarm theoryrdquo in Proceedings of the 6th International Sympo-sium onMicroMachine and Human Science pp 39ndash43 NagoyaJapan October 1995
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
8 International Journal of Antennas and Propagation
Collect previous N data using PSO or GA algorithm
Predicted value for120572-120573-120574 filterxi
xi+1
ximinus1 ximinusFind g for Nximinus1 ximinusN
that minimizes RMSE
Figure 6 Block diagram of adaptive 120572-120573-120574 filter
segmented into block of 1000 sampled data (100 sec) for lineartrajectory target and 420 sampled data (42 sec) for circulartrajectory target respectivelyThe gain value of 120572
2-1205732-1205742filter
in 101ndash200 sec is estimated by GA and PSO methods usingthe collected data during 1ndash100 sec the gain value of 120572
2-
1205732-1205742filter in 201ndash300 sec is estimated by the GA and PSO
methods using the collected data in 101ndash200 sec and so onSimilarly the measured flight target signals are segmentedinto block of 100 sampled data (100 sec) for linear trajectorytarget and 42 sampled data (42 sec) for circular trajectorytarget respectively The gain value of 120572
2-1205732-1205742filter in 101ndash
200 sec is estimated by the GA and PSO methods using thecollected data during 1ndash100 sec the gain value of 120572
2-1205732-1205742
filter in the interval of 201ndash300 sec is estimated by the GA andPSO methods using the collected data in 101ndash200 sec and soon
For the PSO process it was found that gain parameterconverged within 100 iterations so the number of iterationsfor an adaptive 120572-120573-120574 filter was kept at constant 100 iterationsper processing cycle swarm size = 30 inertia = 07298 andcorrection factors 119888
1= 21 119888
2= 2 For the GA process we
have considered population size = 8 string length = 12crossover rate = 08 and mutation rate = 005
If 120579119894119901 119894 = 1 2 119873 and
120579119894119901 119894 = 1 2 119873 are real
measurement pitchrsquos values and estimated pitchrsquos values ofadaptive 120572
1-1205731-1205741filter respectively then 120579
119894119903 119894 = 1 2 119873
and 120579119894119903 119894 = 1 2 119873 are realmeasurement rollrsquos values and
estimated rollrsquos values of adaptive 1205721-1205731-1205741filter respectively
where 119894 is the current iteration and 119873 is the total number ofsampling data The RMSEs of pitch and roll estimations aredefined as follows
RMSE119901= radic
1
119873
119873
sum
119894 = 1
(120579119894119901minus 120579119894119901)
2
(18)
RMSE119903= radic
1
119873
119873
sum
119894 = 1
(120579119894119903minus 120579119894119903)
2
(19)
where the main purpose of GA and PSO methods is to findthe optimum values of 119892
1119901and 119892
1119903which minimize the
RMSE119901and RMSE
119903 respectively
If X119894and X
119894 119894 = 1 2 119873 are real measurement target
position vector values and estimated target position vector
values of adaptive 1205722-1205732-1205742filter the RMSE of target tracking
is defined as
RMSE = radic1
119873
119873
sum
119894 = 1
((119909119894minus 119909119894)2+ (119910119894minus 119910119894)2+ (119911119894minus 119894)2) (20)
where 119909119894 119910119894 and 119911
119894are the components of target position
vector X119894in 119909 119910 and 119911 axes The main purpose of GA and
PSO algorithms is to find the optimum gain value of g2which
minimizes the RMSE
4 Experimental Results
Six scenarios are used to simulate the tracking performanceof ship-borne phased array radar using the proposed adaptive120572-120573-120574 filter for beam pointing error compensation and targettracking in three-dimensional space
Case 1 (roll and pitch estimations using GA based adaptive120572-120573-120574 filter and measured data) The roll and pitch datarecorded by a gyroscope of the ship were used in the exper-iments The total number of data is 1000 and the samplingtime is 01 sec The angle variation of roll and pitch signalsmeasured by ship gyroscope are shown in Figure 7 whichwillresult in the beam pointing errors of phased array antennaTheGA based adaptive 120572-120573-120574 filter is used to estimate the rolland pitch angles Figures 8(a) and 8(b) show that the conver-gent time of roll and pitch angles are 3 sec and 5 sec respec-tively under simulated sea states 2 and 3 and measured dataThe shiprsquos roll and pitch signals are simulated according to (3)and (4) with roll and pitch angle parameters of sea states 2and 3 listed in Table 1 and standard deviation of 001 Itconcludes that the proposed GA based adaptive 120572-120573-120574 filter issuitable to compensate the ship motion As shown in Table 1the period of sea state 2 is longer than sea state 3 Thereforethe convergent RMSE (about 019∘ for roll about 01∘ forpitch) of sea state 2 is less than sea state 3 (about 029∘ for rollabout 022∘ for pitch) because the roll and pitch signals withshorter period in sea state 3 have rapid oscillations which aremore difficult to be predicted precisely
Case 2 (tracking linear moving target in three-dimensionalspace using GA based adaptive 120572-120573-120574 filter (without beampointing error)) Assuming the ship is not affected by thesea waves (antenna beam pointing error is zero) the trackingperformance of GA based adaptive 120572-120573-120574 filter for a straightflight target trajectory is simulated The initial position ofradar is (0 0 0) The ship moves with a speed of 10msec in
International Journal of Antennas and Propagation 9
0 100 200 300 400 500 600 700 800 900 1000
0
02
04
06
08
Pitc
h an
gle (
deg)
Time (10minus1 s)
minus12
minus08
minus06
minus04
minus02
minus1
(a)
0
1
2
3
Roll
angl
e (de
g)
0 100 200 300 400 500 600 700 800 900 1000Time (10minus1 s)
minus1
minus2
minus3
minus4
(b)
Figure 7 Demonstrating angle variation of (a) pitch and (b) roll signals measured by gyroscope
0
005
01
015
02
025
03
035
04
RMSE
(deg
)
Sea 3Sea 2Measurement
0 100 200 300 400 500 600 700 800 900 1000Time (10minus1 s)
(a)
0
005
01
015
02
025
03
035
04RM
SE (d
eg)
Sea 3Sea 2Measurement
Time (10minus1 s)0 100 200 300 400 500 600 700 800 900 1000
(b)
Figure 8 RMSE of (a) roll and (b) pitch under simulated data of sea states 2 and 3 and measured data
the 119884 axial direction The linear path equation of the ship isgiven by
X119904(119909119905 119910119905 119911119905) = X0119904+ V119904119905 (21)
where 119905 = 1 2 3 1000 sec V119904is ship velocity vector
and X0119904is initial ship position vector in three-dimensional
spaceThe radar position is updated every secondThe initialposition of flying target is (minus74840 minus129620 9100)m whichis about 150 km from the radar The flight speed of target is300msec (119883 axial velocity of 150msec and 119884 axial velocityof 260msec) The target location is updated every secondThe linear target trajectory equation is
X119898(119909119905 119910119905 119911119905) = X0119898
+ V119898119905 (22)
where V119898
is the target velocity vector and X0119898
is theinitial target position vector in three-dimensional space Thetracking accuracy of GA based adaptive 120572-120573-120574 filter for astraight flight target is shown in Figure 9 where the trajectoryestimation error is calculated by
119864119905= radic(119909
119905minus 119909119905)2+ (119910119905minus 119910119905)2+ (119911119905minus 119905)2 (23)
where (119909119905 119910119905 119911119905) and (119909
119905 119910119905 119905) denote the real target location
and estimated target location respectively It shows that theGA based adaptive 120572-120573-120574 filter converges to less than about2m after 5 sec
10 International Journal of Antennas and Propagation
Table 2 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0574 06399 0652 06904 06537 0674 06286 06313 06044 053921198921199011
0629 06664 0622 07756 0664 07399 0663 06652 06816 064981198922
0285 05827 03282 03851 00054 03480 02916 03736 0547 04864
Table 3 Estimated gain parameters for PSO based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0557 05986 0357 05345 06142 05774 05544 0529 04592 056261198921199011
0612 05898 04129 05675 05286 03953 05565 06199 05214 062851198922
0326 007 01136 0146 0098 02365 021 02547 02474 01245
0 20 40 60 80 100 1200
20
40
60
80
100
120
140
160
180
Time (s)
Et
(m)
Average estimation error = 15176m
Figure 9 Estimation error for linear target trajectory without beampointing error
Case 3 (tracking circular moving target in three-dimensionalspace using GA based adaptive 120572-120573-120574 filter (without beampointing error)) Figures 10(a) and 10(b) show the simulationscenario of beam pointing error compensation for circularmaneuvering air target and linear moving ship The shiphas the same linear moving speed and path equation asCase 2The flying target is parallel to the119883119884 plane the initialposition of flying target is (8885 4590 9100)m the velocity is300msec and the angle (120579
0) between the initial position and
the119883-axis is 30∘The trajectory of the flight vehicle is a circlewhich has a radius of 10 km The center of circular trajectoryis 119884-axis The rotational cycle time of the flight target is 209seconds and the total observation time is 419 sec
The circular trajectory equation is
X119898(119909119905 119910119905 119911119905) = (119903 cos (120579
0minus 120596119905) 119903 sin (120579
0minus 120596119905) 9100) (24)
where 119905 = 1 2 3 418 419 sec 119903 is the circular flight radiusin meter 120579
0is the initial angle between the flight vehicle
and the 119883-axis 120596 is the angular speed of flight vehicle Theaccuracy of GA based adaptive 120572-120573-120574 filter for tracking acircular flight target is shown in Figure 11 where theGAbasedadaptive 120572-120573-120574 filter converges to less than 3m after 8 secTherefore using the GA based adaptive 120572-120573-120574 filter to trackthe circular trajectory target has the larger average estimationerror and longer convergent time than linear trajectory target
Case 4 (estimating roll and pitch angles and tracking lineartrajectory target using GAPSO based adaptive 120572-120573-120574 filterand measured data (with beam pointing error)) The mea-sured roll and pitch signals shown in Figure 7 are used toevaluate the performance of GAPSO based adaptive 120572-120573-120574filter for estimating the roll and pitch angles and tracking thelinear trajectory targetThe beamof 6times6 planar array antennais steered to track the linear trajectory target The samplingfrequency is set as 10Hz Therefore the shiprsquos roll and pitchsignals are sampled 10000 points within 1000 seconds Asthe scenario described in Case 2 the ship is along the 119884
axis of 119883119884 plane in the earth coordinates the ship speedis 10msec and the flying target speed is 300msec Theflying target is parallel to the 119883119884 plane 91 km above theground and the angle between the flight direction and the119884-axis is 120∘ The largest reconnaissance distance of ship-borne radar is 150 km When the air target is flying withinthe radar detection range the beampointing error estimationand linear trajectory target tracking of ship-borne phasedarray radar are simulated
Figure 12 is a simulation flow chart used to verifythe effect of beam pointing error compensation on thetracking accuracy of adaptive 120572-120573-120574 filter Tables 2 and 3demonstrate the optimal gain parameters of adaptive 120572-120573-120574 filter for roll 119892
1199031 pitch 119892
1199011 and target tracking 119892
2
obtained by the GA and PSO methods respectively thatminimizes the RMSE Figures 13ndash17 show the estimationerrors for five different gain values which are summarizedin Table 4 It demonstrates that the greater gain value(119892 = 01 075 and 09) will generate the greater error andneed the longer convergence time The estimation errors
International Journal of Antennas and Propagation 11
Table 4 Estimation errors for different gains
Calculated by GA Calculated by PSO Fixed gain1198921199031
05392sim06904 0357sim06142 01 075 091198921199011
0622sim07756 03953sim06285 01 075 091198922
0054sim05827 007sim03259 01 075 09Average 119864
119905(m) 14781 163224 169663 534804 3991188
Table 5 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0543 05653 05424 05455 05665 056 058 0532 05414 053851198921199011
0464 0484 04459 0422 04667 04654 04509 04459 04308 046511198922
0161 00929 01937 00865 00241 00609 01174 01897 01713 01727
Y
X
Z
(a)
Y
X10 km
2730∘
(b)
Figure 10 Simulation scenarios of Case 3 for (a) 3D and (b) top view diagrams
0 10 20 30 40 50 60 70 80 90 1000
50
100
150
200
250
Time (s)
Et
(m)
Average estimation error = 24905m
Figure 11 Estimation error for circular target trajectory without beam pointing error
12 International Journal of Antennas and Propagation
Generate the target detection data
Using GA or PSO
to track the target
Generate beam steering angle
Measure or simulate roll and pitch angles
Using GA or PSO
roll and pitch angles
Simulate beam pointing error compensation
Verify tracking accuracy ofGAPSO adaptive 120572-120573-120574 filter
120572-120573-120574 filter 120572-120573-120574 filter to estimate
Figure 12 Flow chart of tracking performance simulation forCase 4experiment
0 100 200 300 400 500 600 700 800 900 10000
50
100
150
200
250
300
Time (s)
Et
(m)
Average estimation error = 14781m
Figure 13 Estimation error of GA based adaptive 120572-120573-120574 filter
presented in Figures 16 and 17 are too large to be appli-cable when the gain values are equal to 075 and 09The GA and PSO have smaller errors compared with thefixed gain values and the GA obtains the best estimationaccuracy The estimation errors of GA and PSO meth-ods were found to be close From Figures 13 14 and 15we can observe that there is a peak value that occurredat the time duration about 510 to 515 seconds becausethe horizontal beam direction of ship-borne phased array
0
50
100
150
200
250
300
Et
(m)
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 163224 m
Figure 14 Estimation error of PSO based adaptive 120572-120573-120574 filter
0
20
40
60
80
100
120
140
160
180
200Et
(m)
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 169663 m
Figure 15 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
radar is steered according to the linear trajectory targetwhich is flying through sky just above the ship at thetime instant of 510 second As shown in Figure 18 thehorizontal beam steering angle of ship-borne phased arrayradar at about the time of 510 second is changing fromdescending into ascending When the GA based adaptive120572-120573-120574 filter tracks the linear trajectory target the esti-mation error values are less than 30 meters or less ifthe peak estimation errors during these 5 seconds areignored
Case 5 (estimating roll and pitch angles and tracking lineartrajectory target using GAPSO based adaptive 120572-120573-120574 filterand simulated data (with beam pointing error)) The shiprsquosroll and pitch signals are simulated according to (3) and (4)with parameters of sea state 2 listed in Table 1 and standard
International Journal of Antennas and Propagation 13
0 200 400 600 800 10000
500
1000
1500
Time (s)
Et
(m)
Average estimation error = 534804 m
Figure 16 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0 200 400 600 800 10000
1000
2000
3000
4000
5000
6000
7000
8000
9000
Time (s)
Et
(m)
Average estimation error = 3991188 m
Figure 17 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
deviation of 001 The simulated shiprsquos roll and pitch signalsare applied for five different methods to estimate the gain ofGAPSO based adaptive 120572-120573-120574 filter The total observationtime is 1000 sec and sampling time is 01 sec The simulationreplicates 10000 times The input samples are updated byone new sample for the next iteration The other simulationscenario and parameters are the same asCase 4 Tables 5 and 6demonstrate the optimal gain parameters of adaptive 120572-120573-120574filter for roll gain 119892
1199031 pitch gain 119892
1199011 and target tracking gain
vector g2obtained by the GA and PSOmethods respectively
that minimizes RMSE Figures 19 20 21 22 and 23 showthe estimation errors for five different gain values which aresummarized in Table 7 The estimation error of GA basedadaptive 120572-120573-120574 filter for tracking a linear trajectory targetis shown in Figure 19 where the convergence time of GA
0 100 200 300 400 500 600 700 800 900 1000Time (s)
0
05
1
15
2
25
Azi
mut
h ste
erin
g an
gle (
rad)
With beam pointing errorWithout beam pointing error
Figure 18 Azimuth beam steering angle of ship-borne phased arrayradar
0
10
20
30
40
50
60
70
80
90
100
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 160756 m
Et
(m)
Figure 19 Estimation error of GA based adaptive 120572-120573-120574 filter
based adaptive 120572-120573-120574 filter is about 3 sec and the averageestimation error is about 160756m It concludes that theexperimental results in Case 5 have the same trend as inCase 4 therefore the simulated data can be used to observethe effect of shipmotion on the tracking performance of ship-borne phased array radar under different sea state conditionsThe convergence time and average estimation error of GAbased adaptive 120572-120573-120574 filter are better than themethod used in[6] where the tracking error of AEKF converges to less thanabout 20m at 550 iterations (seciteration) and the estimationerror will remain within the range of about 20m when thebeam pointing error is compensated by FBFN controller
Case 6 (estimating roll and pitch angles and tracking circularmoving target using GA based adaptive 120572-120573-120574 filter and
14 International Journal of Antennas and Propagation
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 187684 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 20 Estimation error of PSO based adaptive 120572-120573-120574 filter
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 199594 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 21 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
measured data (with beam pointing error)) As the scenariodescribed in Case 3 the measured roll and pitch signalsshown in Figure 7 are used to evaluate the performance ofGA based adaptive 120572-120573-120574 filter for estimating the roll andpitch angles and tracking the circular trajectory target Theestimation errors of 120572-120573-120574 filter with fixed gains of 119892
1199031=
1198921199011
= 1198922= 01 and 119892
1199031= 1198921199011
= 1198922= 075 are shown in
Figures 24 and 25 respectivelyThe average estimation errorsfor fixed filter gains of 01 and 075 are about 41638m and572386m respectively The optimal roll gain 119892
1199031 pitch gain
1198921199011 and target tracking gain 119892
2of adaptive 120572-120573-120574 filter
estimated by the GA method during the observation timeof 419 sec are listed in Table 8 The estimation error of GAbased adaptive 120572-120573-120574 filter for tracking a circular flight targetis shown in Figure 26 where the convergence time of GAbased adaptive 120572-120573-120574 filter is about 5 sec and the average
0
500
1000
1500
0 200 400 600 800 1000Time (s)
Average estimation error = 1080515 m
Et
(m)
Figure 22 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
0 200 400 600 800 1000Time (s)
Average estimation error = 4279449 m
Et
(m)
Figure 23 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
estimation error is about 36667m It concludes that the GAbased adaptive 120572-120573-120574 filter can greatly improve the trackingaccuracy and convergence time of the conventional 120572-120573-120574filter for tracking the circular trajectory target
5 Conclusions
This paper proposed an intelligent beam pointing error com-pensationmechanism for ship-borne phased array radarTheGA based adaptive 120572-120573-120574 filter estimates the roll and pitchangles of the ship moving in the sea and thus compensatesfor the antenna beam pointing error in order to enhance thetracking accuracy of phased array radar system
International Journal of Antennas and Propagation 15
Table 6 Estimated gain parameters for PSO based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0366 03645 0532 04519 05151 05002 04797 05254 04146 053511198921199011
0366 03647 03634 03644 03643 0363 03644 03684 03643 036421198922
0125 01271 02286 02166 01169 01799 01735 01564 00767 01346
Table 7 Estimation errors for different gains
Calculated by GA Calculated by PSO Fixed gain1198921199031
0459sim0583 03645sim05352 01 075 091198921199011
0498sim0667 0363sim03684 01 075 091198922
00586sim01697 00367sim02291 01 075 09Average 119864
119905(m) 160756 187684 199594 1080515 4279449
Table 8 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim42 43sim84 85sim126 127sim168 169sim210 211sim252 253sim294 295sim336 337sim378 379sim4191198921199031
0619 04996 05316 0644 05052 05568 05316 04459 05092 059241198921199011
0582 06727 07201 04913 04168 07169 05983 02036 05858 06211198922
0214 02215 01612 02474 01822 0157 02073 01211 02278 01074
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 41638 m
Figure 24 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
Six experimental cases are conducted to verify the perfor-mance of ship-borne phased array radar using the proposedGA based adaptive 120572-120573-120574 filter In Case 1 the GA basedadaptive 120572-120573-120574 filter is used to estimate the roll and pitchangles with the simulated and measured roll and pitchsignals respectively The simulation results show that themeasurement data has the minimum estimation error theestimation error in the sea state 2 is the second and theestimation error in the sea state 3 is the worst Cases 2 and 3show that when the GA based adaptive 120572-120573-120574 filter is appliedto estimate the beam pointing error and to track the targetin a ship-borne phased array radar the circular trajectory
0 50 100 150 200 250 300 350 400 450Time (s)
0
500
1000
1500
Average estimation error = 572386 m
Et
(m)
Figure 25 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
target will be tracked with the larger average estimation errorand longer convergence time than linear trajectory targetTheexperimental results ofCases 4 and 5demonstrate thatwhen alinear trajectory target is tracked by ship-borne phased arrayradar the average estimation error and convergence time ofship-borne phased array radar using the GA based adaptive120572-120573-120574 filter are better than PSO based adaptive 120572-120573-120574 filterand fixed filter gain 120572-120573-120574 filter The convergence time andtracking accuracy of ship-borne phased array radar usingthe proposed GA based adaptive 120572-120573-120574 filter are superiorto the AEKF significantly In conclusions the proposed GAbased adaptive 120572-120573-120574 filter is a real time applicable algorithm
16 International Journal of Antennas and Propagation
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 36667 m
Figure 26 Estimation error of GA based adaptive 120572-120573-120574 filter
whose accuracy is good with relatively less complexity Inaddition the roll and pitch signalsmeasured in real operationenvironments can be replaced with the simulated roll andpitch signals to test all the relevant properties of practical shipmotion compensation and air target tracking for ship-bornephased array radar
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
This work is supported in part by research grants fromNational Science Council China (NSC 102-2218-E-155-001)
References
[1] Z Fang and Q Rundong Ship-borne Phased Array RadarMotion Compensation System Engineering and ElectronicTechnologies 1998
[2] L J Love J F Jansen and F G Pin ldquoCompensation of waveinduced motion and force phenomenon for ship based highperformance robotic and human amplifying systemsrdquo OakRidge National Laboratory ORNLTM-2003233 2003
[3] R J Hill Motion Compensation for Ship Borne Radars andLidars US Department of Commerce National Oceanic andAtmospheric Administration Office of Oceanic and Atmo-spheric Research Earth System Research Laboratory PhysicalSciences Division 2005
[4] T I Fossen and T Perez ldquoKalman filtering for positioning andheading control of ships and offshore rigsrdquo IEEEControl SystemsMagazine vol 29 no 6 pp 32ndash46 2009
[5] S Kuchler C Pregizer J K Eberharter K Schneider and OSawodny ldquoReal-time estimation of a shiprsquos attituderdquo in Proceed-ings of theAmericanControl Conference (ACC rsquo11) pp 2411ndash2416San Francisco Calif USA July 2011
[6] J Mar K C Tsai Y T Wang and M B Basnet ldquoIntelligentmotion compensation for improving the tracking performanceof shipborne phased array radarrdquo International Journal ofAntennas and Propagation vol 2013 Article ID 384756 14pages 2013
[7] T Dirk and S Tarunraj ldquoOptimal design of 120572-120573-(120574) filtersrdquo inAmerican Control Conference vol 6 pp 4348ndash4352 ChicagoIll USA June 2000
[8] D Tenne and T Singh ldquoCharacterizing performance of 120572-120573-120574filtersrdquo IEEE Transactions on Aerospace and Electronic Systemsvol 38 no 3 pp 1072ndash1087 2002
[9] T E Lee J P Su and K W Yu ldquoParameter optimization fora third-order sampled-data trackerrdquo in Proceedings of the 2ndInternational Conference on Innovative Computing Informationand Control p 336 Kumamoto Japan September 2007
[10] Y H Lho and J H Painter ldquoA fuzzy tuned adaptive Kalmanfilterrdquo in Industrial FuzzyControl and Intelligent System pp 144ndash148 December 1993
[11] N Ramakoti A Vinay and R K Jatoth ldquoParticle swarm opti-mization aided Kalman filter for object trackingrdquo in Proceedingsof the International Conference on Advances in ComputingControl and Telecommunication Technologies (ACT rsquo09) pp 531ndash533 Kerela India December 2009
[12] T-E Lee J-P Su andK-WYu ldquoAparticle swarmoptimization-based state estimation scheme formoving objectsrdquoTransactionsof the Institute of Measurement and Control vol 34 no 2-3 pp236ndash254 2012
[13] D C Law S A McLaughlin M J Post et al ldquoAn electronicallystabilized phased array system for shipborne atmospheric windprofilingrdquo Journal of Atmospheric and Oceanic Technology vol19 no 6 pp 924ndash933 2002
[14] C A Balanis AntennaTheory Analysis and Design JohnWileyamp Sons New York NY USA 3rd edition 2005
[15] J Mar and Y R Lin ldquoImplementation of SDR digital beam-former for microsatellite SARrdquo IEEE Geoscience and RemoteSensing Letters vol 6 no 1 pp 92ndash96 2009
[16] R C Eberhart and Y Shi Computational Intelligence Conceptsto Implementations ElsevierMorgan Kaufmann 2007
[17] K Y Lee and J-B Park ldquoApplication of particle swarm optimi-zation to economic dispatch problem advantages anddisadvan-tagesrdquo in Proceedings of the IEEE PES Power Systems Conferenceand Exposition (PSCE rsquo06) pp 188ndash192 Atlanta Ga USANovember 2006
[18] R C Eberhart and J Kennedy ldquoNew optimizer using particleswarm theoryrdquo in Proceedings of the 6th International Sympo-sium onMicroMachine and Human Science pp 39ndash43 NagoyaJapan October 1995
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
International Journal of Antennas and Propagation 9
0 100 200 300 400 500 600 700 800 900 1000
0
02
04
06
08
Pitc
h an
gle (
deg)
Time (10minus1 s)
minus12
minus08
minus06
minus04
minus02
minus1
(a)
0
1
2
3
Roll
angl
e (de
g)
0 100 200 300 400 500 600 700 800 900 1000Time (10minus1 s)
minus1
minus2
minus3
minus4
(b)
Figure 7 Demonstrating angle variation of (a) pitch and (b) roll signals measured by gyroscope
0
005
01
015
02
025
03
035
04
RMSE
(deg
)
Sea 3Sea 2Measurement
0 100 200 300 400 500 600 700 800 900 1000Time (10minus1 s)
(a)
0
005
01
015
02
025
03
035
04RM
SE (d
eg)
Sea 3Sea 2Measurement
Time (10minus1 s)0 100 200 300 400 500 600 700 800 900 1000
(b)
Figure 8 RMSE of (a) roll and (b) pitch under simulated data of sea states 2 and 3 and measured data
the 119884 axial direction The linear path equation of the ship isgiven by
X119904(119909119905 119910119905 119911119905) = X0119904+ V119904119905 (21)
where 119905 = 1 2 3 1000 sec V119904is ship velocity vector
and X0119904is initial ship position vector in three-dimensional
spaceThe radar position is updated every secondThe initialposition of flying target is (minus74840 minus129620 9100)m whichis about 150 km from the radar The flight speed of target is300msec (119883 axial velocity of 150msec and 119884 axial velocityof 260msec) The target location is updated every secondThe linear target trajectory equation is
X119898(119909119905 119910119905 119911119905) = X0119898
+ V119898119905 (22)
where V119898
is the target velocity vector and X0119898
is theinitial target position vector in three-dimensional space Thetracking accuracy of GA based adaptive 120572-120573-120574 filter for astraight flight target is shown in Figure 9 where the trajectoryestimation error is calculated by
119864119905= radic(119909
119905minus 119909119905)2+ (119910119905minus 119910119905)2+ (119911119905minus 119905)2 (23)
where (119909119905 119910119905 119911119905) and (119909
119905 119910119905 119905) denote the real target location
and estimated target location respectively It shows that theGA based adaptive 120572-120573-120574 filter converges to less than about2m after 5 sec
10 International Journal of Antennas and Propagation
Table 2 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0574 06399 0652 06904 06537 0674 06286 06313 06044 053921198921199011
0629 06664 0622 07756 0664 07399 0663 06652 06816 064981198922
0285 05827 03282 03851 00054 03480 02916 03736 0547 04864
Table 3 Estimated gain parameters for PSO based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0557 05986 0357 05345 06142 05774 05544 0529 04592 056261198921199011
0612 05898 04129 05675 05286 03953 05565 06199 05214 062851198922
0326 007 01136 0146 0098 02365 021 02547 02474 01245
0 20 40 60 80 100 1200
20
40
60
80
100
120
140
160
180
Time (s)
Et
(m)
Average estimation error = 15176m
Figure 9 Estimation error for linear target trajectory without beampointing error
Case 3 (tracking circular moving target in three-dimensionalspace using GA based adaptive 120572-120573-120574 filter (without beampointing error)) Figures 10(a) and 10(b) show the simulationscenario of beam pointing error compensation for circularmaneuvering air target and linear moving ship The shiphas the same linear moving speed and path equation asCase 2The flying target is parallel to the119883119884 plane the initialposition of flying target is (8885 4590 9100)m the velocity is300msec and the angle (120579
0) between the initial position and
the119883-axis is 30∘The trajectory of the flight vehicle is a circlewhich has a radius of 10 km The center of circular trajectoryis 119884-axis The rotational cycle time of the flight target is 209seconds and the total observation time is 419 sec
The circular trajectory equation is
X119898(119909119905 119910119905 119911119905) = (119903 cos (120579
0minus 120596119905) 119903 sin (120579
0minus 120596119905) 9100) (24)
where 119905 = 1 2 3 418 419 sec 119903 is the circular flight radiusin meter 120579
0is the initial angle between the flight vehicle
and the 119883-axis 120596 is the angular speed of flight vehicle Theaccuracy of GA based adaptive 120572-120573-120574 filter for tracking acircular flight target is shown in Figure 11 where theGAbasedadaptive 120572-120573-120574 filter converges to less than 3m after 8 secTherefore using the GA based adaptive 120572-120573-120574 filter to trackthe circular trajectory target has the larger average estimationerror and longer convergent time than linear trajectory target
Case 4 (estimating roll and pitch angles and tracking lineartrajectory target using GAPSO based adaptive 120572-120573-120574 filterand measured data (with beam pointing error)) The mea-sured roll and pitch signals shown in Figure 7 are used toevaluate the performance of GAPSO based adaptive 120572-120573-120574filter for estimating the roll and pitch angles and tracking thelinear trajectory targetThe beamof 6times6 planar array antennais steered to track the linear trajectory target The samplingfrequency is set as 10Hz Therefore the shiprsquos roll and pitchsignals are sampled 10000 points within 1000 seconds Asthe scenario described in Case 2 the ship is along the 119884
axis of 119883119884 plane in the earth coordinates the ship speedis 10msec and the flying target speed is 300msec Theflying target is parallel to the 119883119884 plane 91 km above theground and the angle between the flight direction and the119884-axis is 120∘ The largest reconnaissance distance of ship-borne radar is 150 km When the air target is flying withinthe radar detection range the beampointing error estimationand linear trajectory target tracking of ship-borne phasedarray radar are simulated
Figure 12 is a simulation flow chart used to verifythe effect of beam pointing error compensation on thetracking accuracy of adaptive 120572-120573-120574 filter Tables 2 and 3demonstrate the optimal gain parameters of adaptive 120572-120573-120574 filter for roll 119892
1199031 pitch 119892
1199011 and target tracking 119892
2
obtained by the GA and PSO methods respectively thatminimizes the RMSE Figures 13ndash17 show the estimationerrors for five different gain values which are summarizedin Table 4 It demonstrates that the greater gain value(119892 = 01 075 and 09) will generate the greater error andneed the longer convergence time The estimation errors
International Journal of Antennas and Propagation 11
Table 4 Estimation errors for different gains
Calculated by GA Calculated by PSO Fixed gain1198921199031
05392sim06904 0357sim06142 01 075 091198921199011
0622sim07756 03953sim06285 01 075 091198922
0054sim05827 007sim03259 01 075 09Average 119864
119905(m) 14781 163224 169663 534804 3991188
Table 5 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0543 05653 05424 05455 05665 056 058 0532 05414 053851198921199011
0464 0484 04459 0422 04667 04654 04509 04459 04308 046511198922
0161 00929 01937 00865 00241 00609 01174 01897 01713 01727
Y
X
Z
(a)
Y
X10 km
2730∘
(b)
Figure 10 Simulation scenarios of Case 3 for (a) 3D and (b) top view diagrams
0 10 20 30 40 50 60 70 80 90 1000
50
100
150
200
250
Time (s)
Et
(m)
Average estimation error = 24905m
Figure 11 Estimation error for circular target trajectory without beam pointing error
12 International Journal of Antennas and Propagation
Generate the target detection data
Using GA or PSO
to track the target
Generate beam steering angle
Measure or simulate roll and pitch angles
Using GA or PSO
roll and pitch angles
Simulate beam pointing error compensation
Verify tracking accuracy ofGAPSO adaptive 120572-120573-120574 filter
120572-120573-120574 filter 120572-120573-120574 filter to estimate
Figure 12 Flow chart of tracking performance simulation forCase 4experiment
0 100 200 300 400 500 600 700 800 900 10000
50
100
150
200
250
300
Time (s)
Et
(m)
Average estimation error = 14781m
Figure 13 Estimation error of GA based adaptive 120572-120573-120574 filter
presented in Figures 16 and 17 are too large to be appli-cable when the gain values are equal to 075 and 09The GA and PSO have smaller errors compared with thefixed gain values and the GA obtains the best estimationaccuracy The estimation errors of GA and PSO meth-ods were found to be close From Figures 13 14 and 15we can observe that there is a peak value that occurredat the time duration about 510 to 515 seconds becausethe horizontal beam direction of ship-borne phased array
0
50
100
150
200
250
300
Et
(m)
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 163224 m
Figure 14 Estimation error of PSO based adaptive 120572-120573-120574 filter
0
20
40
60
80
100
120
140
160
180
200Et
(m)
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 169663 m
Figure 15 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
radar is steered according to the linear trajectory targetwhich is flying through sky just above the ship at thetime instant of 510 second As shown in Figure 18 thehorizontal beam steering angle of ship-borne phased arrayradar at about the time of 510 second is changing fromdescending into ascending When the GA based adaptive120572-120573-120574 filter tracks the linear trajectory target the esti-mation error values are less than 30 meters or less ifthe peak estimation errors during these 5 seconds areignored
Case 5 (estimating roll and pitch angles and tracking lineartrajectory target using GAPSO based adaptive 120572-120573-120574 filterand simulated data (with beam pointing error)) The shiprsquosroll and pitch signals are simulated according to (3) and (4)with parameters of sea state 2 listed in Table 1 and standard
International Journal of Antennas and Propagation 13
0 200 400 600 800 10000
500
1000
1500
Time (s)
Et
(m)
Average estimation error = 534804 m
Figure 16 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0 200 400 600 800 10000
1000
2000
3000
4000
5000
6000
7000
8000
9000
Time (s)
Et
(m)
Average estimation error = 3991188 m
Figure 17 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
deviation of 001 The simulated shiprsquos roll and pitch signalsare applied for five different methods to estimate the gain ofGAPSO based adaptive 120572-120573-120574 filter The total observationtime is 1000 sec and sampling time is 01 sec The simulationreplicates 10000 times The input samples are updated byone new sample for the next iteration The other simulationscenario and parameters are the same asCase 4 Tables 5 and 6demonstrate the optimal gain parameters of adaptive 120572-120573-120574filter for roll gain 119892
1199031 pitch gain 119892
1199011 and target tracking gain
vector g2obtained by the GA and PSOmethods respectively
that minimizes RMSE Figures 19 20 21 22 and 23 showthe estimation errors for five different gain values which aresummarized in Table 7 The estimation error of GA basedadaptive 120572-120573-120574 filter for tracking a linear trajectory targetis shown in Figure 19 where the convergence time of GA
0 100 200 300 400 500 600 700 800 900 1000Time (s)
0
05
1
15
2
25
Azi
mut
h ste
erin
g an
gle (
rad)
With beam pointing errorWithout beam pointing error
Figure 18 Azimuth beam steering angle of ship-borne phased arrayradar
0
10
20
30
40
50
60
70
80
90
100
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 160756 m
Et
(m)
Figure 19 Estimation error of GA based adaptive 120572-120573-120574 filter
based adaptive 120572-120573-120574 filter is about 3 sec and the averageestimation error is about 160756m It concludes that theexperimental results in Case 5 have the same trend as inCase 4 therefore the simulated data can be used to observethe effect of shipmotion on the tracking performance of ship-borne phased array radar under different sea state conditionsThe convergence time and average estimation error of GAbased adaptive 120572-120573-120574 filter are better than themethod used in[6] where the tracking error of AEKF converges to less thanabout 20m at 550 iterations (seciteration) and the estimationerror will remain within the range of about 20m when thebeam pointing error is compensated by FBFN controller
Case 6 (estimating roll and pitch angles and tracking circularmoving target using GA based adaptive 120572-120573-120574 filter and
14 International Journal of Antennas and Propagation
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 187684 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 20 Estimation error of PSO based adaptive 120572-120573-120574 filter
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 199594 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 21 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
measured data (with beam pointing error)) As the scenariodescribed in Case 3 the measured roll and pitch signalsshown in Figure 7 are used to evaluate the performance ofGA based adaptive 120572-120573-120574 filter for estimating the roll andpitch angles and tracking the circular trajectory target Theestimation errors of 120572-120573-120574 filter with fixed gains of 119892
1199031=
1198921199011
= 1198922= 01 and 119892
1199031= 1198921199011
= 1198922= 075 are shown in
Figures 24 and 25 respectivelyThe average estimation errorsfor fixed filter gains of 01 and 075 are about 41638m and572386m respectively The optimal roll gain 119892
1199031 pitch gain
1198921199011 and target tracking gain 119892
2of adaptive 120572-120573-120574 filter
estimated by the GA method during the observation timeof 419 sec are listed in Table 8 The estimation error of GAbased adaptive 120572-120573-120574 filter for tracking a circular flight targetis shown in Figure 26 where the convergence time of GAbased adaptive 120572-120573-120574 filter is about 5 sec and the average
0
500
1000
1500
0 200 400 600 800 1000Time (s)
Average estimation error = 1080515 m
Et
(m)
Figure 22 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
0 200 400 600 800 1000Time (s)
Average estimation error = 4279449 m
Et
(m)
Figure 23 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
estimation error is about 36667m It concludes that the GAbased adaptive 120572-120573-120574 filter can greatly improve the trackingaccuracy and convergence time of the conventional 120572-120573-120574filter for tracking the circular trajectory target
5 Conclusions
This paper proposed an intelligent beam pointing error com-pensationmechanism for ship-borne phased array radarTheGA based adaptive 120572-120573-120574 filter estimates the roll and pitchangles of the ship moving in the sea and thus compensatesfor the antenna beam pointing error in order to enhance thetracking accuracy of phased array radar system
International Journal of Antennas and Propagation 15
Table 6 Estimated gain parameters for PSO based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0366 03645 0532 04519 05151 05002 04797 05254 04146 053511198921199011
0366 03647 03634 03644 03643 0363 03644 03684 03643 036421198922
0125 01271 02286 02166 01169 01799 01735 01564 00767 01346
Table 7 Estimation errors for different gains
Calculated by GA Calculated by PSO Fixed gain1198921199031
0459sim0583 03645sim05352 01 075 091198921199011
0498sim0667 0363sim03684 01 075 091198922
00586sim01697 00367sim02291 01 075 09Average 119864
119905(m) 160756 187684 199594 1080515 4279449
Table 8 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim42 43sim84 85sim126 127sim168 169sim210 211sim252 253sim294 295sim336 337sim378 379sim4191198921199031
0619 04996 05316 0644 05052 05568 05316 04459 05092 059241198921199011
0582 06727 07201 04913 04168 07169 05983 02036 05858 06211198922
0214 02215 01612 02474 01822 0157 02073 01211 02278 01074
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 41638 m
Figure 24 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
Six experimental cases are conducted to verify the perfor-mance of ship-borne phased array radar using the proposedGA based adaptive 120572-120573-120574 filter In Case 1 the GA basedadaptive 120572-120573-120574 filter is used to estimate the roll and pitchangles with the simulated and measured roll and pitchsignals respectively The simulation results show that themeasurement data has the minimum estimation error theestimation error in the sea state 2 is the second and theestimation error in the sea state 3 is the worst Cases 2 and 3show that when the GA based adaptive 120572-120573-120574 filter is appliedto estimate the beam pointing error and to track the targetin a ship-borne phased array radar the circular trajectory
0 50 100 150 200 250 300 350 400 450Time (s)
0
500
1000
1500
Average estimation error = 572386 m
Et
(m)
Figure 25 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
target will be tracked with the larger average estimation errorand longer convergence time than linear trajectory targetTheexperimental results ofCases 4 and 5demonstrate thatwhen alinear trajectory target is tracked by ship-borne phased arrayradar the average estimation error and convergence time ofship-borne phased array radar using the GA based adaptive120572-120573-120574 filter are better than PSO based adaptive 120572-120573-120574 filterand fixed filter gain 120572-120573-120574 filter The convergence time andtracking accuracy of ship-borne phased array radar usingthe proposed GA based adaptive 120572-120573-120574 filter are superiorto the AEKF significantly In conclusions the proposed GAbased adaptive 120572-120573-120574 filter is a real time applicable algorithm
16 International Journal of Antennas and Propagation
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 36667 m
Figure 26 Estimation error of GA based adaptive 120572-120573-120574 filter
whose accuracy is good with relatively less complexity Inaddition the roll and pitch signalsmeasured in real operationenvironments can be replaced with the simulated roll andpitch signals to test all the relevant properties of practical shipmotion compensation and air target tracking for ship-bornephased array radar
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
This work is supported in part by research grants fromNational Science Council China (NSC 102-2218-E-155-001)
References
[1] Z Fang and Q Rundong Ship-borne Phased Array RadarMotion Compensation System Engineering and ElectronicTechnologies 1998
[2] L J Love J F Jansen and F G Pin ldquoCompensation of waveinduced motion and force phenomenon for ship based highperformance robotic and human amplifying systemsrdquo OakRidge National Laboratory ORNLTM-2003233 2003
[3] R J Hill Motion Compensation for Ship Borne Radars andLidars US Department of Commerce National Oceanic andAtmospheric Administration Office of Oceanic and Atmo-spheric Research Earth System Research Laboratory PhysicalSciences Division 2005
[4] T I Fossen and T Perez ldquoKalman filtering for positioning andheading control of ships and offshore rigsrdquo IEEEControl SystemsMagazine vol 29 no 6 pp 32ndash46 2009
[5] S Kuchler C Pregizer J K Eberharter K Schneider and OSawodny ldquoReal-time estimation of a shiprsquos attituderdquo in Proceed-ings of theAmericanControl Conference (ACC rsquo11) pp 2411ndash2416San Francisco Calif USA July 2011
[6] J Mar K C Tsai Y T Wang and M B Basnet ldquoIntelligentmotion compensation for improving the tracking performanceof shipborne phased array radarrdquo International Journal ofAntennas and Propagation vol 2013 Article ID 384756 14pages 2013
[7] T Dirk and S Tarunraj ldquoOptimal design of 120572-120573-(120574) filtersrdquo inAmerican Control Conference vol 6 pp 4348ndash4352 ChicagoIll USA June 2000
[8] D Tenne and T Singh ldquoCharacterizing performance of 120572-120573-120574filtersrdquo IEEE Transactions on Aerospace and Electronic Systemsvol 38 no 3 pp 1072ndash1087 2002
[9] T E Lee J P Su and K W Yu ldquoParameter optimization fora third-order sampled-data trackerrdquo in Proceedings of the 2ndInternational Conference on Innovative Computing Informationand Control p 336 Kumamoto Japan September 2007
[10] Y H Lho and J H Painter ldquoA fuzzy tuned adaptive Kalmanfilterrdquo in Industrial FuzzyControl and Intelligent System pp 144ndash148 December 1993
[11] N Ramakoti A Vinay and R K Jatoth ldquoParticle swarm opti-mization aided Kalman filter for object trackingrdquo in Proceedingsof the International Conference on Advances in ComputingControl and Telecommunication Technologies (ACT rsquo09) pp 531ndash533 Kerela India December 2009
[12] T-E Lee J-P Su andK-WYu ldquoAparticle swarmoptimization-based state estimation scheme formoving objectsrdquoTransactionsof the Institute of Measurement and Control vol 34 no 2-3 pp236ndash254 2012
[13] D C Law S A McLaughlin M J Post et al ldquoAn electronicallystabilized phased array system for shipborne atmospheric windprofilingrdquo Journal of Atmospheric and Oceanic Technology vol19 no 6 pp 924ndash933 2002
[14] C A Balanis AntennaTheory Analysis and Design JohnWileyamp Sons New York NY USA 3rd edition 2005
[15] J Mar and Y R Lin ldquoImplementation of SDR digital beam-former for microsatellite SARrdquo IEEE Geoscience and RemoteSensing Letters vol 6 no 1 pp 92ndash96 2009
[16] R C Eberhart and Y Shi Computational Intelligence Conceptsto Implementations ElsevierMorgan Kaufmann 2007
[17] K Y Lee and J-B Park ldquoApplication of particle swarm optimi-zation to economic dispatch problem advantages anddisadvan-tagesrdquo in Proceedings of the IEEE PES Power Systems Conferenceand Exposition (PSCE rsquo06) pp 188ndash192 Atlanta Ga USANovember 2006
[18] R C Eberhart and J Kennedy ldquoNew optimizer using particleswarm theoryrdquo in Proceedings of the 6th International Sympo-sium onMicroMachine and Human Science pp 39ndash43 NagoyaJapan October 1995
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
10 International Journal of Antennas and Propagation
Table 2 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0574 06399 0652 06904 06537 0674 06286 06313 06044 053921198921199011
0629 06664 0622 07756 0664 07399 0663 06652 06816 064981198922
0285 05827 03282 03851 00054 03480 02916 03736 0547 04864
Table 3 Estimated gain parameters for PSO based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0557 05986 0357 05345 06142 05774 05544 0529 04592 056261198921199011
0612 05898 04129 05675 05286 03953 05565 06199 05214 062851198922
0326 007 01136 0146 0098 02365 021 02547 02474 01245
0 20 40 60 80 100 1200
20
40
60
80
100
120
140
160
180
Time (s)
Et
(m)
Average estimation error = 15176m
Figure 9 Estimation error for linear target trajectory without beampointing error
Case 3 (tracking circular moving target in three-dimensionalspace using GA based adaptive 120572-120573-120574 filter (without beampointing error)) Figures 10(a) and 10(b) show the simulationscenario of beam pointing error compensation for circularmaneuvering air target and linear moving ship The shiphas the same linear moving speed and path equation asCase 2The flying target is parallel to the119883119884 plane the initialposition of flying target is (8885 4590 9100)m the velocity is300msec and the angle (120579
0) between the initial position and
the119883-axis is 30∘The trajectory of the flight vehicle is a circlewhich has a radius of 10 km The center of circular trajectoryis 119884-axis The rotational cycle time of the flight target is 209seconds and the total observation time is 419 sec
The circular trajectory equation is
X119898(119909119905 119910119905 119911119905) = (119903 cos (120579
0minus 120596119905) 119903 sin (120579
0minus 120596119905) 9100) (24)
where 119905 = 1 2 3 418 419 sec 119903 is the circular flight radiusin meter 120579
0is the initial angle between the flight vehicle
and the 119883-axis 120596 is the angular speed of flight vehicle Theaccuracy of GA based adaptive 120572-120573-120574 filter for tracking acircular flight target is shown in Figure 11 where theGAbasedadaptive 120572-120573-120574 filter converges to less than 3m after 8 secTherefore using the GA based adaptive 120572-120573-120574 filter to trackthe circular trajectory target has the larger average estimationerror and longer convergent time than linear trajectory target
Case 4 (estimating roll and pitch angles and tracking lineartrajectory target using GAPSO based adaptive 120572-120573-120574 filterand measured data (with beam pointing error)) The mea-sured roll and pitch signals shown in Figure 7 are used toevaluate the performance of GAPSO based adaptive 120572-120573-120574filter for estimating the roll and pitch angles and tracking thelinear trajectory targetThe beamof 6times6 planar array antennais steered to track the linear trajectory target The samplingfrequency is set as 10Hz Therefore the shiprsquos roll and pitchsignals are sampled 10000 points within 1000 seconds Asthe scenario described in Case 2 the ship is along the 119884
axis of 119883119884 plane in the earth coordinates the ship speedis 10msec and the flying target speed is 300msec Theflying target is parallel to the 119883119884 plane 91 km above theground and the angle between the flight direction and the119884-axis is 120∘ The largest reconnaissance distance of ship-borne radar is 150 km When the air target is flying withinthe radar detection range the beampointing error estimationand linear trajectory target tracking of ship-borne phasedarray radar are simulated
Figure 12 is a simulation flow chart used to verifythe effect of beam pointing error compensation on thetracking accuracy of adaptive 120572-120573-120574 filter Tables 2 and 3demonstrate the optimal gain parameters of adaptive 120572-120573-120574 filter for roll 119892
1199031 pitch 119892
1199011 and target tracking 119892
2
obtained by the GA and PSO methods respectively thatminimizes the RMSE Figures 13ndash17 show the estimationerrors for five different gain values which are summarizedin Table 4 It demonstrates that the greater gain value(119892 = 01 075 and 09) will generate the greater error andneed the longer convergence time The estimation errors
International Journal of Antennas and Propagation 11
Table 4 Estimation errors for different gains
Calculated by GA Calculated by PSO Fixed gain1198921199031
05392sim06904 0357sim06142 01 075 091198921199011
0622sim07756 03953sim06285 01 075 091198922
0054sim05827 007sim03259 01 075 09Average 119864
119905(m) 14781 163224 169663 534804 3991188
Table 5 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0543 05653 05424 05455 05665 056 058 0532 05414 053851198921199011
0464 0484 04459 0422 04667 04654 04509 04459 04308 046511198922
0161 00929 01937 00865 00241 00609 01174 01897 01713 01727
Y
X
Z
(a)
Y
X10 km
2730∘
(b)
Figure 10 Simulation scenarios of Case 3 for (a) 3D and (b) top view diagrams
0 10 20 30 40 50 60 70 80 90 1000
50
100
150
200
250
Time (s)
Et
(m)
Average estimation error = 24905m
Figure 11 Estimation error for circular target trajectory without beam pointing error
12 International Journal of Antennas and Propagation
Generate the target detection data
Using GA or PSO
to track the target
Generate beam steering angle
Measure or simulate roll and pitch angles
Using GA or PSO
roll and pitch angles
Simulate beam pointing error compensation
Verify tracking accuracy ofGAPSO adaptive 120572-120573-120574 filter
120572-120573-120574 filter 120572-120573-120574 filter to estimate
Figure 12 Flow chart of tracking performance simulation forCase 4experiment
0 100 200 300 400 500 600 700 800 900 10000
50
100
150
200
250
300
Time (s)
Et
(m)
Average estimation error = 14781m
Figure 13 Estimation error of GA based adaptive 120572-120573-120574 filter
presented in Figures 16 and 17 are too large to be appli-cable when the gain values are equal to 075 and 09The GA and PSO have smaller errors compared with thefixed gain values and the GA obtains the best estimationaccuracy The estimation errors of GA and PSO meth-ods were found to be close From Figures 13 14 and 15we can observe that there is a peak value that occurredat the time duration about 510 to 515 seconds becausethe horizontal beam direction of ship-borne phased array
0
50
100
150
200
250
300
Et
(m)
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 163224 m
Figure 14 Estimation error of PSO based adaptive 120572-120573-120574 filter
0
20
40
60
80
100
120
140
160
180
200Et
(m)
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 169663 m
Figure 15 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
radar is steered according to the linear trajectory targetwhich is flying through sky just above the ship at thetime instant of 510 second As shown in Figure 18 thehorizontal beam steering angle of ship-borne phased arrayradar at about the time of 510 second is changing fromdescending into ascending When the GA based adaptive120572-120573-120574 filter tracks the linear trajectory target the esti-mation error values are less than 30 meters or less ifthe peak estimation errors during these 5 seconds areignored
Case 5 (estimating roll and pitch angles and tracking lineartrajectory target using GAPSO based adaptive 120572-120573-120574 filterand simulated data (with beam pointing error)) The shiprsquosroll and pitch signals are simulated according to (3) and (4)with parameters of sea state 2 listed in Table 1 and standard
International Journal of Antennas and Propagation 13
0 200 400 600 800 10000
500
1000
1500
Time (s)
Et
(m)
Average estimation error = 534804 m
Figure 16 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0 200 400 600 800 10000
1000
2000
3000
4000
5000
6000
7000
8000
9000
Time (s)
Et
(m)
Average estimation error = 3991188 m
Figure 17 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
deviation of 001 The simulated shiprsquos roll and pitch signalsare applied for five different methods to estimate the gain ofGAPSO based adaptive 120572-120573-120574 filter The total observationtime is 1000 sec and sampling time is 01 sec The simulationreplicates 10000 times The input samples are updated byone new sample for the next iteration The other simulationscenario and parameters are the same asCase 4 Tables 5 and 6demonstrate the optimal gain parameters of adaptive 120572-120573-120574filter for roll gain 119892
1199031 pitch gain 119892
1199011 and target tracking gain
vector g2obtained by the GA and PSOmethods respectively
that minimizes RMSE Figures 19 20 21 22 and 23 showthe estimation errors for five different gain values which aresummarized in Table 7 The estimation error of GA basedadaptive 120572-120573-120574 filter for tracking a linear trajectory targetis shown in Figure 19 where the convergence time of GA
0 100 200 300 400 500 600 700 800 900 1000Time (s)
0
05
1
15
2
25
Azi
mut
h ste
erin
g an
gle (
rad)
With beam pointing errorWithout beam pointing error
Figure 18 Azimuth beam steering angle of ship-borne phased arrayradar
0
10
20
30
40
50
60
70
80
90
100
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 160756 m
Et
(m)
Figure 19 Estimation error of GA based adaptive 120572-120573-120574 filter
based adaptive 120572-120573-120574 filter is about 3 sec and the averageestimation error is about 160756m It concludes that theexperimental results in Case 5 have the same trend as inCase 4 therefore the simulated data can be used to observethe effect of shipmotion on the tracking performance of ship-borne phased array radar under different sea state conditionsThe convergence time and average estimation error of GAbased adaptive 120572-120573-120574 filter are better than themethod used in[6] where the tracking error of AEKF converges to less thanabout 20m at 550 iterations (seciteration) and the estimationerror will remain within the range of about 20m when thebeam pointing error is compensated by FBFN controller
Case 6 (estimating roll and pitch angles and tracking circularmoving target using GA based adaptive 120572-120573-120574 filter and
14 International Journal of Antennas and Propagation
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 187684 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 20 Estimation error of PSO based adaptive 120572-120573-120574 filter
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 199594 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 21 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
measured data (with beam pointing error)) As the scenariodescribed in Case 3 the measured roll and pitch signalsshown in Figure 7 are used to evaluate the performance ofGA based adaptive 120572-120573-120574 filter for estimating the roll andpitch angles and tracking the circular trajectory target Theestimation errors of 120572-120573-120574 filter with fixed gains of 119892
1199031=
1198921199011
= 1198922= 01 and 119892
1199031= 1198921199011
= 1198922= 075 are shown in
Figures 24 and 25 respectivelyThe average estimation errorsfor fixed filter gains of 01 and 075 are about 41638m and572386m respectively The optimal roll gain 119892
1199031 pitch gain
1198921199011 and target tracking gain 119892
2of adaptive 120572-120573-120574 filter
estimated by the GA method during the observation timeof 419 sec are listed in Table 8 The estimation error of GAbased adaptive 120572-120573-120574 filter for tracking a circular flight targetis shown in Figure 26 where the convergence time of GAbased adaptive 120572-120573-120574 filter is about 5 sec and the average
0
500
1000
1500
0 200 400 600 800 1000Time (s)
Average estimation error = 1080515 m
Et
(m)
Figure 22 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
0 200 400 600 800 1000Time (s)
Average estimation error = 4279449 m
Et
(m)
Figure 23 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
estimation error is about 36667m It concludes that the GAbased adaptive 120572-120573-120574 filter can greatly improve the trackingaccuracy and convergence time of the conventional 120572-120573-120574filter for tracking the circular trajectory target
5 Conclusions
This paper proposed an intelligent beam pointing error com-pensationmechanism for ship-borne phased array radarTheGA based adaptive 120572-120573-120574 filter estimates the roll and pitchangles of the ship moving in the sea and thus compensatesfor the antenna beam pointing error in order to enhance thetracking accuracy of phased array radar system
International Journal of Antennas and Propagation 15
Table 6 Estimated gain parameters for PSO based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0366 03645 0532 04519 05151 05002 04797 05254 04146 053511198921199011
0366 03647 03634 03644 03643 0363 03644 03684 03643 036421198922
0125 01271 02286 02166 01169 01799 01735 01564 00767 01346
Table 7 Estimation errors for different gains
Calculated by GA Calculated by PSO Fixed gain1198921199031
0459sim0583 03645sim05352 01 075 091198921199011
0498sim0667 0363sim03684 01 075 091198922
00586sim01697 00367sim02291 01 075 09Average 119864
119905(m) 160756 187684 199594 1080515 4279449
Table 8 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim42 43sim84 85sim126 127sim168 169sim210 211sim252 253sim294 295sim336 337sim378 379sim4191198921199031
0619 04996 05316 0644 05052 05568 05316 04459 05092 059241198921199011
0582 06727 07201 04913 04168 07169 05983 02036 05858 06211198922
0214 02215 01612 02474 01822 0157 02073 01211 02278 01074
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 41638 m
Figure 24 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
Six experimental cases are conducted to verify the perfor-mance of ship-borne phased array radar using the proposedGA based adaptive 120572-120573-120574 filter In Case 1 the GA basedadaptive 120572-120573-120574 filter is used to estimate the roll and pitchangles with the simulated and measured roll and pitchsignals respectively The simulation results show that themeasurement data has the minimum estimation error theestimation error in the sea state 2 is the second and theestimation error in the sea state 3 is the worst Cases 2 and 3show that when the GA based adaptive 120572-120573-120574 filter is appliedto estimate the beam pointing error and to track the targetin a ship-borne phased array radar the circular trajectory
0 50 100 150 200 250 300 350 400 450Time (s)
0
500
1000
1500
Average estimation error = 572386 m
Et
(m)
Figure 25 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
target will be tracked with the larger average estimation errorand longer convergence time than linear trajectory targetTheexperimental results ofCases 4 and 5demonstrate thatwhen alinear trajectory target is tracked by ship-borne phased arrayradar the average estimation error and convergence time ofship-borne phased array radar using the GA based adaptive120572-120573-120574 filter are better than PSO based adaptive 120572-120573-120574 filterand fixed filter gain 120572-120573-120574 filter The convergence time andtracking accuracy of ship-borne phased array radar usingthe proposed GA based adaptive 120572-120573-120574 filter are superiorto the AEKF significantly In conclusions the proposed GAbased adaptive 120572-120573-120574 filter is a real time applicable algorithm
16 International Journal of Antennas and Propagation
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 36667 m
Figure 26 Estimation error of GA based adaptive 120572-120573-120574 filter
whose accuracy is good with relatively less complexity Inaddition the roll and pitch signalsmeasured in real operationenvironments can be replaced with the simulated roll andpitch signals to test all the relevant properties of practical shipmotion compensation and air target tracking for ship-bornephased array radar
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
This work is supported in part by research grants fromNational Science Council China (NSC 102-2218-E-155-001)
References
[1] Z Fang and Q Rundong Ship-borne Phased Array RadarMotion Compensation System Engineering and ElectronicTechnologies 1998
[2] L J Love J F Jansen and F G Pin ldquoCompensation of waveinduced motion and force phenomenon for ship based highperformance robotic and human amplifying systemsrdquo OakRidge National Laboratory ORNLTM-2003233 2003
[3] R J Hill Motion Compensation for Ship Borne Radars andLidars US Department of Commerce National Oceanic andAtmospheric Administration Office of Oceanic and Atmo-spheric Research Earth System Research Laboratory PhysicalSciences Division 2005
[4] T I Fossen and T Perez ldquoKalman filtering for positioning andheading control of ships and offshore rigsrdquo IEEEControl SystemsMagazine vol 29 no 6 pp 32ndash46 2009
[5] S Kuchler C Pregizer J K Eberharter K Schneider and OSawodny ldquoReal-time estimation of a shiprsquos attituderdquo in Proceed-ings of theAmericanControl Conference (ACC rsquo11) pp 2411ndash2416San Francisco Calif USA July 2011
[6] J Mar K C Tsai Y T Wang and M B Basnet ldquoIntelligentmotion compensation for improving the tracking performanceof shipborne phased array radarrdquo International Journal ofAntennas and Propagation vol 2013 Article ID 384756 14pages 2013
[7] T Dirk and S Tarunraj ldquoOptimal design of 120572-120573-(120574) filtersrdquo inAmerican Control Conference vol 6 pp 4348ndash4352 ChicagoIll USA June 2000
[8] D Tenne and T Singh ldquoCharacterizing performance of 120572-120573-120574filtersrdquo IEEE Transactions on Aerospace and Electronic Systemsvol 38 no 3 pp 1072ndash1087 2002
[9] T E Lee J P Su and K W Yu ldquoParameter optimization fora third-order sampled-data trackerrdquo in Proceedings of the 2ndInternational Conference on Innovative Computing Informationand Control p 336 Kumamoto Japan September 2007
[10] Y H Lho and J H Painter ldquoA fuzzy tuned adaptive Kalmanfilterrdquo in Industrial FuzzyControl and Intelligent System pp 144ndash148 December 1993
[11] N Ramakoti A Vinay and R K Jatoth ldquoParticle swarm opti-mization aided Kalman filter for object trackingrdquo in Proceedingsof the International Conference on Advances in ComputingControl and Telecommunication Technologies (ACT rsquo09) pp 531ndash533 Kerela India December 2009
[12] T-E Lee J-P Su andK-WYu ldquoAparticle swarmoptimization-based state estimation scheme formoving objectsrdquoTransactionsof the Institute of Measurement and Control vol 34 no 2-3 pp236ndash254 2012
[13] D C Law S A McLaughlin M J Post et al ldquoAn electronicallystabilized phased array system for shipborne atmospheric windprofilingrdquo Journal of Atmospheric and Oceanic Technology vol19 no 6 pp 924ndash933 2002
[14] C A Balanis AntennaTheory Analysis and Design JohnWileyamp Sons New York NY USA 3rd edition 2005
[15] J Mar and Y R Lin ldquoImplementation of SDR digital beam-former for microsatellite SARrdquo IEEE Geoscience and RemoteSensing Letters vol 6 no 1 pp 92ndash96 2009
[16] R C Eberhart and Y Shi Computational Intelligence Conceptsto Implementations ElsevierMorgan Kaufmann 2007
[17] K Y Lee and J-B Park ldquoApplication of particle swarm optimi-zation to economic dispatch problem advantages anddisadvan-tagesrdquo in Proceedings of the IEEE PES Power Systems Conferenceand Exposition (PSCE rsquo06) pp 188ndash192 Atlanta Ga USANovember 2006
[18] R C Eberhart and J Kennedy ldquoNew optimizer using particleswarm theoryrdquo in Proceedings of the 6th International Sympo-sium onMicroMachine and Human Science pp 39ndash43 NagoyaJapan October 1995
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
International Journal of Antennas and Propagation 11
Table 4 Estimation errors for different gains
Calculated by GA Calculated by PSO Fixed gain1198921199031
05392sim06904 0357sim06142 01 075 091198921199011
0622sim07756 03953sim06285 01 075 091198922
0054sim05827 007sim03259 01 075 09Average 119864
119905(m) 14781 163224 169663 534804 3991188
Table 5 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0543 05653 05424 05455 05665 056 058 0532 05414 053851198921199011
0464 0484 04459 0422 04667 04654 04509 04459 04308 046511198922
0161 00929 01937 00865 00241 00609 01174 01897 01713 01727
Y
X
Z
(a)
Y
X10 km
2730∘
(b)
Figure 10 Simulation scenarios of Case 3 for (a) 3D and (b) top view diagrams
0 10 20 30 40 50 60 70 80 90 1000
50
100
150
200
250
Time (s)
Et
(m)
Average estimation error = 24905m
Figure 11 Estimation error for circular target trajectory without beam pointing error
12 International Journal of Antennas and Propagation
Generate the target detection data
Using GA or PSO
to track the target
Generate beam steering angle
Measure or simulate roll and pitch angles
Using GA or PSO
roll and pitch angles
Simulate beam pointing error compensation
Verify tracking accuracy ofGAPSO adaptive 120572-120573-120574 filter
120572-120573-120574 filter 120572-120573-120574 filter to estimate
Figure 12 Flow chart of tracking performance simulation forCase 4experiment
0 100 200 300 400 500 600 700 800 900 10000
50
100
150
200
250
300
Time (s)
Et
(m)
Average estimation error = 14781m
Figure 13 Estimation error of GA based adaptive 120572-120573-120574 filter
presented in Figures 16 and 17 are too large to be appli-cable when the gain values are equal to 075 and 09The GA and PSO have smaller errors compared with thefixed gain values and the GA obtains the best estimationaccuracy The estimation errors of GA and PSO meth-ods were found to be close From Figures 13 14 and 15we can observe that there is a peak value that occurredat the time duration about 510 to 515 seconds becausethe horizontal beam direction of ship-borne phased array
0
50
100
150
200
250
300
Et
(m)
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 163224 m
Figure 14 Estimation error of PSO based adaptive 120572-120573-120574 filter
0
20
40
60
80
100
120
140
160
180
200Et
(m)
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 169663 m
Figure 15 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
radar is steered according to the linear trajectory targetwhich is flying through sky just above the ship at thetime instant of 510 second As shown in Figure 18 thehorizontal beam steering angle of ship-borne phased arrayradar at about the time of 510 second is changing fromdescending into ascending When the GA based adaptive120572-120573-120574 filter tracks the linear trajectory target the esti-mation error values are less than 30 meters or less ifthe peak estimation errors during these 5 seconds areignored
Case 5 (estimating roll and pitch angles and tracking lineartrajectory target using GAPSO based adaptive 120572-120573-120574 filterand simulated data (with beam pointing error)) The shiprsquosroll and pitch signals are simulated according to (3) and (4)with parameters of sea state 2 listed in Table 1 and standard
International Journal of Antennas and Propagation 13
0 200 400 600 800 10000
500
1000
1500
Time (s)
Et
(m)
Average estimation error = 534804 m
Figure 16 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0 200 400 600 800 10000
1000
2000
3000
4000
5000
6000
7000
8000
9000
Time (s)
Et
(m)
Average estimation error = 3991188 m
Figure 17 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
deviation of 001 The simulated shiprsquos roll and pitch signalsare applied for five different methods to estimate the gain ofGAPSO based adaptive 120572-120573-120574 filter The total observationtime is 1000 sec and sampling time is 01 sec The simulationreplicates 10000 times The input samples are updated byone new sample for the next iteration The other simulationscenario and parameters are the same asCase 4 Tables 5 and 6demonstrate the optimal gain parameters of adaptive 120572-120573-120574filter for roll gain 119892
1199031 pitch gain 119892
1199011 and target tracking gain
vector g2obtained by the GA and PSOmethods respectively
that minimizes RMSE Figures 19 20 21 22 and 23 showthe estimation errors for five different gain values which aresummarized in Table 7 The estimation error of GA basedadaptive 120572-120573-120574 filter for tracking a linear trajectory targetis shown in Figure 19 where the convergence time of GA
0 100 200 300 400 500 600 700 800 900 1000Time (s)
0
05
1
15
2
25
Azi
mut
h ste
erin
g an
gle (
rad)
With beam pointing errorWithout beam pointing error
Figure 18 Azimuth beam steering angle of ship-borne phased arrayradar
0
10
20
30
40
50
60
70
80
90
100
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 160756 m
Et
(m)
Figure 19 Estimation error of GA based adaptive 120572-120573-120574 filter
based adaptive 120572-120573-120574 filter is about 3 sec and the averageestimation error is about 160756m It concludes that theexperimental results in Case 5 have the same trend as inCase 4 therefore the simulated data can be used to observethe effect of shipmotion on the tracking performance of ship-borne phased array radar under different sea state conditionsThe convergence time and average estimation error of GAbased adaptive 120572-120573-120574 filter are better than themethod used in[6] where the tracking error of AEKF converges to less thanabout 20m at 550 iterations (seciteration) and the estimationerror will remain within the range of about 20m when thebeam pointing error is compensated by FBFN controller
Case 6 (estimating roll and pitch angles and tracking circularmoving target using GA based adaptive 120572-120573-120574 filter and
14 International Journal of Antennas and Propagation
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 187684 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 20 Estimation error of PSO based adaptive 120572-120573-120574 filter
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 199594 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 21 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
measured data (with beam pointing error)) As the scenariodescribed in Case 3 the measured roll and pitch signalsshown in Figure 7 are used to evaluate the performance ofGA based adaptive 120572-120573-120574 filter for estimating the roll andpitch angles and tracking the circular trajectory target Theestimation errors of 120572-120573-120574 filter with fixed gains of 119892
1199031=
1198921199011
= 1198922= 01 and 119892
1199031= 1198921199011
= 1198922= 075 are shown in
Figures 24 and 25 respectivelyThe average estimation errorsfor fixed filter gains of 01 and 075 are about 41638m and572386m respectively The optimal roll gain 119892
1199031 pitch gain
1198921199011 and target tracking gain 119892
2of adaptive 120572-120573-120574 filter
estimated by the GA method during the observation timeof 419 sec are listed in Table 8 The estimation error of GAbased adaptive 120572-120573-120574 filter for tracking a circular flight targetis shown in Figure 26 where the convergence time of GAbased adaptive 120572-120573-120574 filter is about 5 sec and the average
0
500
1000
1500
0 200 400 600 800 1000Time (s)
Average estimation error = 1080515 m
Et
(m)
Figure 22 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
0 200 400 600 800 1000Time (s)
Average estimation error = 4279449 m
Et
(m)
Figure 23 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
estimation error is about 36667m It concludes that the GAbased adaptive 120572-120573-120574 filter can greatly improve the trackingaccuracy and convergence time of the conventional 120572-120573-120574filter for tracking the circular trajectory target
5 Conclusions
This paper proposed an intelligent beam pointing error com-pensationmechanism for ship-borne phased array radarTheGA based adaptive 120572-120573-120574 filter estimates the roll and pitchangles of the ship moving in the sea and thus compensatesfor the antenna beam pointing error in order to enhance thetracking accuracy of phased array radar system
International Journal of Antennas and Propagation 15
Table 6 Estimated gain parameters for PSO based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0366 03645 0532 04519 05151 05002 04797 05254 04146 053511198921199011
0366 03647 03634 03644 03643 0363 03644 03684 03643 036421198922
0125 01271 02286 02166 01169 01799 01735 01564 00767 01346
Table 7 Estimation errors for different gains
Calculated by GA Calculated by PSO Fixed gain1198921199031
0459sim0583 03645sim05352 01 075 091198921199011
0498sim0667 0363sim03684 01 075 091198922
00586sim01697 00367sim02291 01 075 09Average 119864
119905(m) 160756 187684 199594 1080515 4279449
Table 8 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim42 43sim84 85sim126 127sim168 169sim210 211sim252 253sim294 295sim336 337sim378 379sim4191198921199031
0619 04996 05316 0644 05052 05568 05316 04459 05092 059241198921199011
0582 06727 07201 04913 04168 07169 05983 02036 05858 06211198922
0214 02215 01612 02474 01822 0157 02073 01211 02278 01074
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 41638 m
Figure 24 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
Six experimental cases are conducted to verify the perfor-mance of ship-borne phased array radar using the proposedGA based adaptive 120572-120573-120574 filter In Case 1 the GA basedadaptive 120572-120573-120574 filter is used to estimate the roll and pitchangles with the simulated and measured roll and pitchsignals respectively The simulation results show that themeasurement data has the minimum estimation error theestimation error in the sea state 2 is the second and theestimation error in the sea state 3 is the worst Cases 2 and 3show that when the GA based adaptive 120572-120573-120574 filter is appliedto estimate the beam pointing error and to track the targetin a ship-borne phased array radar the circular trajectory
0 50 100 150 200 250 300 350 400 450Time (s)
0
500
1000
1500
Average estimation error = 572386 m
Et
(m)
Figure 25 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
target will be tracked with the larger average estimation errorand longer convergence time than linear trajectory targetTheexperimental results ofCases 4 and 5demonstrate thatwhen alinear trajectory target is tracked by ship-borne phased arrayradar the average estimation error and convergence time ofship-borne phased array radar using the GA based adaptive120572-120573-120574 filter are better than PSO based adaptive 120572-120573-120574 filterand fixed filter gain 120572-120573-120574 filter The convergence time andtracking accuracy of ship-borne phased array radar usingthe proposed GA based adaptive 120572-120573-120574 filter are superiorto the AEKF significantly In conclusions the proposed GAbased adaptive 120572-120573-120574 filter is a real time applicable algorithm
16 International Journal of Antennas and Propagation
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 36667 m
Figure 26 Estimation error of GA based adaptive 120572-120573-120574 filter
whose accuracy is good with relatively less complexity Inaddition the roll and pitch signalsmeasured in real operationenvironments can be replaced with the simulated roll andpitch signals to test all the relevant properties of practical shipmotion compensation and air target tracking for ship-bornephased array radar
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
This work is supported in part by research grants fromNational Science Council China (NSC 102-2218-E-155-001)
References
[1] Z Fang and Q Rundong Ship-borne Phased Array RadarMotion Compensation System Engineering and ElectronicTechnologies 1998
[2] L J Love J F Jansen and F G Pin ldquoCompensation of waveinduced motion and force phenomenon for ship based highperformance robotic and human amplifying systemsrdquo OakRidge National Laboratory ORNLTM-2003233 2003
[3] R J Hill Motion Compensation for Ship Borne Radars andLidars US Department of Commerce National Oceanic andAtmospheric Administration Office of Oceanic and Atmo-spheric Research Earth System Research Laboratory PhysicalSciences Division 2005
[4] T I Fossen and T Perez ldquoKalman filtering for positioning andheading control of ships and offshore rigsrdquo IEEEControl SystemsMagazine vol 29 no 6 pp 32ndash46 2009
[5] S Kuchler C Pregizer J K Eberharter K Schneider and OSawodny ldquoReal-time estimation of a shiprsquos attituderdquo in Proceed-ings of theAmericanControl Conference (ACC rsquo11) pp 2411ndash2416San Francisco Calif USA July 2011
[6] J Mar K C Tsai Y T Wang and M B Basnet ldquoIntelligentmotion compensation for improving the tracking performanceof shipborne phased array radarrdquo International Journal ofAntennas and Propagation vol 2013 Article ID 384756 14pages 2013
[7] T Dirk and S Tarunraj ldquoOptimal design of 120572-120573-(120574) filtersrdquo inAmerican Control Conference vol 6 pp 4348ndash4352 ChicagoIll USA June 2000
[8] D Tenne and T Singh ldquoCharacterizing performance of 120572-120573-120574filtersrdquo IEEE Transactions on Aerospace and Electronic Systemsvol 38 no 3 pp 1072ndash1087 2002
[9] T E Lee J P Su and K W Yu ldquoParameter optimization fora third-order sampled-data trackerrdquo in Proceedings of the 2ndInternational Conference on Innovative Computing Informationand Control p 336 Kumamoto Japan September 2007
[10] Y H Lho and J H Painter ldquoA fuzzy tuned adaptive Kalmanfilterrdquo in Industrial FuzzyControl and Intelligent System pp 144ndash148 December 1993
[11] N Ramakoti A Vinay and R K Jatoth ldquoParticle swarm opti-mization aided Kalman filter for object trackingrdquo in Proceedingsof the International Conference on Advances in ComputingControl and Telecommunication Technologies (ACT rsquo09) pp 531ndash533 Kerela India December 2009
[12] T-E Lee J-P Su andK-WYu ldquoAparticle swarmoptimization-based state estimation scheme formoving objectsrdquoTransactionsof the Institute of Measurement and Control vol 34 no 2-3 pp236ndash254 2012
[13] D C Law S A McLaughlin M J Post et al ldquoAn electronicallystabilized phased array system for shipborne atmospheric windprofilingrdquo Journal of Atmospheric and Oceanic Technology vol19 no 6 pp 924ndash933 2002
[14] C A Balanis AntennaTheory Analysis and Design JohnWileyamp Sons New York NY USA 3rd edition 2005
[15] J Mar and Y R Lin ldquoImplementation of SDR digital beam-former for microsatellite SARrdquo IEEE Geoscience and RemoteSensing Letters vol 6 no 1 pp 92ndash96 2009
[16] R C Eberhart and Y Shi Computational Intelligence Conceptsto Implementations ElsevierMorgan Kaufmann 2007
[17] K Y Lee and J-B Park ldquoApplication of particle swarm optimi-zation to economic dispatch problem advantages anddisadvan-tagesrdquo in Proceedings of the IEEE PES Power Systems Conferenceand Exposition (PSCE rsquo06) pp 188ndash192 Atlanta Ga USANovember 2006
[18] R C Eberhart and J Kennedy ldquoNew optimizer using particleswarm theoryrdquo in Proceedings of the 6th International Sympo-sium onMicroMachine and Human Science pp 39ndash43 NagoyaJapan October 1995
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
12 International Journal of Antennas and Propagation
Generate the target detection data
Using GA or PSO
to track the target
Generate beam steering angle
Measure or simulate roll and pitch angles
Using GA or PSO
roll and pitch angles
Simulate beam pointing error compensation
Verify tracking accuracy ofGAPSO adaptive 120572-120573-120574 filter
120572-120573-120574 filter 120572-120573-120574 filter to estimate
Figure 12 Flow chart of tracking performance simulation forCase 4experiment
0 100 200 300 400 500 600 700 800 900 10000
50
100
150
200
250
300
Time (s)
Et
(m)
Average estimation error = 14781m
Figure 13 Estimation error of GA based adaptive 120572-120573-120574 filter
presented in Figures 16 and 17 are too large to be appli-cable when the gain values are equal to 075 and 09The GA and PSO have smaller errors compared with thefixed gain values and the GA obtains the best estimationaccuracy The estimation errors of GA and PSO meth-ods were found to be close From Figures 13 14 and 15we can observe that there is a peak value that occurredat the time duration about 510 to 515 seconds becausethe horizontal beam direction of ship-borne phased array
0
50
100
150
200
250
300
Et
(m)
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 163224 m
Figure 14 Estimation error of PSO based adaptive 120572-120573-120574 filter
0
20
40
60
80
100
120
140
160
180
200Et
(m)
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 169663 m
Figure 15 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
radar is steered according to the linear trajectory targetwhich is flying through sky just above the ship at thetime instant of 510 second As shown in Figure 18 thehorizontal beam steering angle of ship-borne phased arrayradar at about the time of 510 second is changing fromdescending into ascending When the GA based adaptive120572-120573-120574 filter tracks the linear trajectory target the esti-mation error values are less than 30 meters or less ifthe peak estimation errors during these 5 seconds areignored
Case 5 (estimating roll and pitch angles and tracking lineartrajectory target using GAPSO based adaptive 120572-120573-120574 filterand simulated data (with beam pointing error)) The shiprsquosroll and pitch signals are simulated according to (3) and (4)with parameters of sea state 2 listed in Table 1 and standard
International Journal of Antennas and Propagation 13
0 200 400 600 800 10000
500
1000
1500
Time (s)
Et
(m)
Average estimation error = 534804 m
Figure 16 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0 200 400 600 800 10000
1000
2000
3000
4000
5000
6000
7000
8000
9000
Time (s)
Et
(m)
Average estimation error = 3991188 m
Figure 17 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
deviation of 001 The simulated shiprsquos roll and pitch signalsare applied for five different methods to estimate the gain ofGAPSO based adaptive 120572-120573-120574 filter The total observationtime is 1000 sec and sampling time is 01 sec The simulationreplicates 10000 times The input samples are updated byone new sample for the next iteration The other simulationscenario and parameters are the same asCase 4 Tables 5 and 6demonstrate the optimal gain parameters of adaptive 120572-120573-120574filter for roll gain 119892
1199031 pitch gain 119892
1199011 and target tracking gain
vector g2obtained by the GA and PSOmethods respectively
that minimizes RMSE Figures 19 20 21 22 and 23 showthe estimation errors for five different gain values which aresummarized in Table 7 The estimation error of GA basedadaptive 120572-120573-120574 filter for tracking a linear trajectory targetis shown in Figure 19 where the convergence time of GA
0 100 200 300 400 500 600 700 800 900 1000Time (s)
0
05
1
15
2
25
Azi
mut
h ste
erin
g an
gle (
rad)
With beam pointing errorWithout beam pointing error
Figure 18 Azimuth beam steering angle of ship-borne phased arrayradar
0
10
20
30
40
50
60
70
80
90
100
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 160756 m
Et
(m)
Figure 19 Estimation error of GA based adaptive 120572-120573-120574 filter
based adaptive 120572-120573-120574 filter is about 3 sec and the averageestimation error is about 160756m It concludes that theexperimental results in Case 5 have the same trend as inCase 4 therefore the simulated data can be used to observethe effect of shipmotion on the tracking performance of ship-borne phased array radar under different sea state conditionsThe convergence time and average estimation error of GAbased adaptive 120572-120573-120574 filter are better than themethod used in[6] where the tracking error of AEKF converges to less thanabout 20m at 550 iterations (seciteration) and the estimationerror will remain within the range of about 20m when thebeam pointing error is compensated by FBFN controller
Case 6 (estimating roll and pitch angles and tracking circularmoving target using GA based adaptive 120572-120573-120574 filter and
14 International Journal of Antennas and Propagation
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 187684 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 20 Estimation error of PSO based adaptive 120572-120573-120574 filter
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 199594 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 21 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
measured data (with beam pointing error)) As the scenariodescribed in Case 3 the measured roll and pitch signalsshown in Figure 7 are used to evaluate the performance ofGA based adaptive 120572-120573-120574 filter for estimating the roll andpitch angles and tracking the circular trajectory target Theestimation errors of 120572-120573-120574 filter with fixed gains of 119892
1199031=
1198921199011
= 1198922= 01 and 119892
1199031= 1198921199011
= 1198922= 075 are shown in
Figures 24 and 25 respectivelyThe average estimation errorsfor fixed filter gains of 01 and 075 are about 41638m and572386m respectively The optimal roll gain 119892
1199031 pitch gain
1198921199011 and target tracking gain 119892
2of adaptive 120572-120573-120574 filter
estimated by the GA method during the observation timeof 419 sec are listed in Table 8 The estimation error of GAbased adaptive 120572-120573-120574 filter for tracking a circular flight targetis shown in Figure 26 where the convergence time of GAbased adaptive 120572-120573-120574 filter is about 5 sec and the average
0
500
1000
1500
0 200 400 600 800 1000Time (s)
Average estimation error = 1080515 m
Et
(m)
Figure 22 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
0 200 400 600 800 1000Time (s)
Average estimation error = 4279449 m
Et
(m)
Figure 23 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
estimation error is about 36667m It concludes that the GAbased adaptive 120572-120573-120574 filter can greatly improve the trackingaccuracy and convergence time of the conventional 120572-120573-120574filter for tracking the circular trajectory target
5 Conclusions
This paper proposed an intelligent beam pointing error com-pensationmechanism for ship-borne phased array radarTheGA based adaptive 120572-120573-120574 filter estimates the roll and pitchangles of the ship moving in the sea and thus compensatesfor the antenna beam pointing error in order to enhance thetracking accuracy of phased array radar system
International Journal of Antennas and Propagation 15
Table 6 Estimated gain parameters for PSO based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0366 03645 0532 04519 05151 05002 04797 05254 04146 053511198921199011
0366 03647 03634 03644 03643 0363 03644 03684 03643 036421198922
0125 01271 02286 02166 01169 01799 01735 01564 00767 01346
Table 7 Estimation errors for different gains
Calculated by GA Calculated by PSO Fixed gain1198921199031
0459sim0583 03645sim05352 01 075 091198921199011
0498sim0667 0363sim03684 01 075 091198922
00586sim01697 00367sim02291 01 075 09Average 119864
119905(m) 160756 187684 199594 1080515 4279449
Table 8 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim42 43sim84 85sim126 127sim168 169sim210 211sim252 253sim294 295sim336 337sim378 379sim4191198921199031
0619 04996 05316 0644 05052 05568 05316 04459 05092 059241198921199011
0582 06727 07201 04913 04168 07169 05983 02036 05858 06211198922
0214 02215 01612 02474 01822 0157 02073 01211 02278 01074
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 41638 m
Figure 24 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
Six experimental cases are conducted to verify the perfor-mance of ship-borne phased array radar using the proposedGA based adaptive 120572-120573-120574 filter In Case 1 the GA basedadaptive 120572-120573-120574 filter is used to estimate the roll and pitchangles with the simulated and measured roll and pitchsignals respectively The simulation results show that themeasurement data has the minimum estimation error theestimation error in the sea state 2 is the second and theestimation error in the sea state 3 is the worst Cases 2 and 3show that when the GA based adaptive 120572-120573-120574 filter is appliedto estimate the beam pointing error and to track the targetin a ship-borne phased array radar the circular trajectory
0 50 100 150 200 250 300 350 400 450Time (s)
0
500
1000
1500
Average estimation error = 572386 m
Et
(m)
Figure 25 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
target will be tracked with the larger average estimation errorand longer convergence time than linear trajectory targetTheexperimental results ofCases 4 and 5demonstrate thatwhen alinear trajectory target is tracked by ship-borne phased arrayradar the average estimation error and convergence time ofship-borne phased array radar using the GA based adaptive120572-120573-120574 filter are better than PSO based adaptive 120572-120573-120574 filterand fixed filter gain 120572-120573-120574 filter The convergence time andtracking accuracy of ship-borne phased array radar usingthe proposed GA based adaptive 120572-120573-120574 filter are superiorto the AEKF significantly In conclusions the proposed GAbased adaptive 120572-120573-120574 filter is a real time applicable algorithm
16 International Journal of Antennas and Propagation
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 36667 m
Figure 26 Estimation error of GA based adaptive 120572-120573-120574 filter
whose accuracy is good with relatively less complexity Inaddition the roll and pitch signalsmeasured in real operationenvironments can be replaced with the simulated roll andpitch signals to test all the relevant properties of practical shipmotion compensation and air target tracking for ship-bornephased array radar
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
This work is supported in part by research grants fromNational Science Council China (NSC 102-2218-E-155-001)
References
[1] Z Fang and Q Rundong Ship-borne Phased Array RadarMotion Compensation System Engineering and ElectronicTechnologies 1998
[2] L J Love J F Jansen and F G Pin ldquoCompensation of waveinduced motion and force phenomenon for ship based highperformance robotic and human amplifying systemsrdquo OakRidge National Laboratory ORNLTM-2003233 2003
[3] R J Hill Motion Compensation for Ship Borne Radars andLidars US Department of Commerce National Oceanic andAtmospheric Administration Office of Oceanic and Atmo-spheric Research Earth System Research Laboratory PhysicalSciences Division 2005
[4] T I Fossen and T Perez ldquoKalman filtering for positioning andheading control of ships and offshore rigsrdquo IEEEControl SystemsMagazine vol 29 no 6 pp 32ndash46 2009
[5] S Kuchler C Pregizer J K Eberharter K Schneider and OSawodny ldquoReal-time estimation of a shiprsquos attituderdquo in Proceed-ings of theAmericanControl Conference (ACC rsquo11) pp 2411ndash2416San Francisco Calif USA July 2011
[6] J Mar K C Tsai Y T Wang and M B Basnet ldquoIntelligentmotion compensation for improving the tracking performanceof shipborne phased array radarrdquo International Journal ofAntennas and Propagation vol 2013 Article ID 384756 14pages 2013
[7] T Dirk and S Tarunraj ldquoOptimal design of 120572-120573-(120574) filtersrdquo inAmerican Control Conference vol 6 pp 4348ndash4352 ChicagoIll USA June 2000
[8] D Tenne and T Singh ldquoCharacterizing performance of 120572-120573-120574filtersrdquo IEEE Transactions on Aerospace and Electronic Systemsvol 38 no 3 pp 1072ndash1087 2002
[9] T E Lee J P Su and K W Yu ldquoParameter optimization fora third-order sampled-data trackerrdquo in Proceedings of the 2ndInternational Conference on Innovative Computing Informationand Control p 336 Kumamoto Japan September 2007
[10] Y H Lho and J H Painter ldquoA fuzzy tuned adaptive Kalmanfilterrdquo in Industrial FuzzyControl and Intelligent System pp 144ndash148 December 1993
[11] N Ramakoti A Vinay and R K Jatoth ldquoParticle swarm opti-mization aided Kalman filter for object trackingrdquo in Proceedingsof the International Conference on Advances in ComputingControl and Telecommunication Technologies (ACT rsquo09) pp 531ndash533 Kerela India December 2009
[12] T-E Lee J-P Su andK-WYu ldquoAparticle swarmoptimization-based state estimation scheme formoving objectsrdquoTransactionsof the Institute of Measurement and Control vol 34 no 2-3 pp236ndash254 2012
[13] D C Law S A McLaughlin M J Post et al ldquoAn electronicallystabilized phased array system for shipborne atmospheric windprofilingrdquo Journal of Atmospheric and Oceanic Technology vol19 no 6 pp 924ndash933 2002
[14] C A Balanis AntennaTheory Analysis and Design JohnWileyamp Sons New York NY USA 3rd edition 2005
[15] J Mar and Y R Lin ldquoImplementation of SDR digital beam-former for microsatellite SARrdquo IEEE Geoscience and RemoteSensing Letters vol 6 no 1 pp 92ndash96 2009
[16] R C Eberhart and Y Shi Computational Intelligence Conceptsto Implementations ElsevierMorgan Kaufmann 2007
[17] K Y Lee and J-B Park ldquoApplication of particle swarm optimi-zation to economic dispatch problem advantages anddisadvan-tagesrdquo in Proceedings of the IEEE PES Power Systems Conferenceand Exposition (PSCE rsquo06) pp 188ndash192 Atlanta Ga USANovember 2006
[18] R C Eberhart and J Kennedy ldquoNew optimizer using particleswarm theoryrdquo in Proceedings of the 6th International Sympo-sium onMicroMachine and Human Science pp 39ndash43 NagoyaJapan October 1995
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
International Journal of Antennas and Propagation 13
0 200 400 600 800 10000
500
1000
1500
Time (s)
Et
(m)
Average estimation error = 534804 m
Figure 16 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0 200 400 600 800 10000
1000
2000
3000
4000
5000
6000
7000
8000
9000
Time (s)
Et
(m)
Average estimation error = 3991188 m
Figure 17 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
deviation of 001 The simulated shiprsquos roll and pitch signalsare applied for five different methods to estimate the gain ofGAPSO based adaptive 120572-120573-120574 filter The total observationtime is 1000 sec and sampling time is 01 sec The simulationreplicates 10000 times The input samples are updated byone new sample for the next iteration The other simulationscenario and parameters are the same asCase 4 Tables 5 and 6demonstrate the optimal gain parameters of adaptive 120572-120573-120574filter for roll gain 119892
1199031 pitch gain 119892
1199011 and target tracking gain
vector g2obtained by the GA and PSOmethods respectively
that minimizes RMSE Figures 19 20 21 22 and 23 showthe estimation errors for five different gain values which aresummarized in Table 7 The estimation error of GA basedadaptive 120572-120573-120574 filter for tracking a linear trajectory targetis shown in Figure 19 where the convergence time of GA
0 100 200 300 400 500 600 700 800 900 1000Time (s)
0
05
1
15
2
25
Azi
mut
h ste
erin
g an
gle (
rad)
With beam pointing errorWithout beam pointing error
Figure 18 Azimuth beam steering angle of ship-borne phased arrayradar
0
10
20
30
40
50
60
70
80
90
100
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 160756 m
Et
(m)
Figure 19 Estimation error of GA based adaptive 120572-120573-120574 filter
based adaptive 120572-120573-120574 filter is about 3 sec and the averageestimation error is about 160756m It concludes that theexperimental results in Case 5 have the same trend as inCase 4 therefore the simulated data can be used to observethe effect of shipmotion on the tracking performance of ship-borne phased array radar under different sea state conditionsThe convergence time and average estimation error of GAbased adaptive 120572-120573-120574 filter are better than themethod used in[6] where the tracking error of AEKF converges to less thanabout 20m at 550 iterations (seciteration) and the estimationerror will remain within the range of about 20m when thebeam pointing error is compensated by FBFN controller
Case 6 (estimating roll and pitch angles and tracking circularmoving target using GA based adaptive 120572-120573-120574 filter and
14 International Journal of Antennas and Propagation
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 187684 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 20 Estimation error of PSO based adaptive 120572-120573-120574 filter
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 199594 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 21 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
measured data (with beam pointing error)) As the scenariodescribed in Case 3 the measured roll and pitch signalsshown in Figure 7 are used to evaluate the performance ofGA based adaptive 120572-120573-120574 filter for estimating the roll andpitch angles and tracking the circular trajectory target Theestimation errors of 120572-120573-120574 filter with fixed gains of 119892
1199031=
1198921199011
= 1198922= 01 and 119892
1199031= 1198921199011
= 1198922= 075 are shown in
Figures 24 and 25 respectivelyThe average estimation errorsfor fixed filter gains of 01 and 075 are about 41638m and572386m respectively The optimal roll gain 119892
1199031 pitch gain
1198921199011 and target tracking gain 119892
2of adaptive 120572-120573-120574 filter
estimated by the GA method during the observation timeof 419 sec are listed in Table 8 The estimation error of GAbased adaptive 120572-120573-120574 filter for tracking a circular flight targetis shown in Figure 26 where the convergence time of GAbased adaptive 120572-120573-120574 filter is about 5 sec and the average
0
500
1000
1500
0 200 400 600 800 1000Time (s)
Average estimation error = 1080515 m
Et
(m)
Figure 22 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
0 200 400 600 800 1000Time (s)
Average estimation error = 4279449 m
Et
(m)
Figure 23 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
estimation error is about 36667m It concludes that the GAbased adaptive 120572-120573-120574 filter can greatly improve the trackingaccuracy and convergence time of the conventional 120572-120573-120574filter for tracking the circular trajectory target
5 Conclusions
This paper proposed an intelligent beam pointing error com-pensationmechanism for ship-borne phased array radarTheGA based adaptive 120572-120573-120574 filter estimates the roll and pitchangles of the ship moving in the sea and thus compensatesfor the antenna beam pointing error in order to enhance thetracking accuracy of phased array radar system
International Journal of Antennas and Propagation 15
Table 6 Estimated gain parameters for PSO based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0366 03645 0532 04519 05151 05002 04797 05254 04146 053511198921199011
0366 03647 03634 03644 03643 0363 03644 03684 03643 036421198922
0125 01271 02286 02166 01169 01799 01735 01564 00767 01346
Table 7 Estimation errors for different gains
Calculated by GA Calculated by PSO Fixed gain1198921199031
0459sim0583 03645sim05352 01 075 091198921199011
0498sim0667 0363sim03684 01 075 091198922
00586sim01697 00367sim02291 01 075 09Average 119864
119905(m) 160756 187684 199594 1080515 4279449
Table 8 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim42 43sim84 85sim126 127sim168 169sim210 211sim252 253sim294 295sim336 337sim378 379sim4191198921199031
0619 04996 05316 0644 05052 05568 05316 04459 05092 059241198921199011
0582 06727 07201 04913 04168 07169 05983 02036 05858 06211198922
0214 02215 01612 02474 01822 0157 02073 01211 02278 01074
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 41638 m
Figure 24 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
Six experimental cases are conducted to verify the perfor-mance of ship-borne phased array radar using the proposedGA based adaptive 120572-120573-120574 filter In Case 1 the GA basedadaptive 120572-120573-120574 filter is used to estimate the roll and pitchangles with the simulated and measured roll and pitchsignals respectively The simulation results show that themeasurement data has the minimum estimation error theestimation error in the sea state 2 is the second and theestimation error in the sea state 3 is the worst Cases 2 and 3show that when the GA based adaptive 120572-120573-120574 filter is appliedto estimate the beam pointing error and to track the targetin a ship-borne phased array radar the circular trajectory
0 50 100 150 200 250 300 350 400 450Time (s)
0
500
1000
1500
Average estimation error = 572386 m
Et
(m)
Figure 25 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
target will be tracked with the larger average estimation errorand longer convergence time than linear trajectory targetTheexperimental results ofCases 4 and 5demonstrate thatwhen alinear trajectory target is tracked by ship-borne phased arrayradar the average estimation error and convergence time ofship-borne phased array radar using the GA based adaptive120572-120573-120574 filter are better than PSO based adaptive 120572-120573-120574 filterand fixed filter gain 120572-120573-120574 filter The convergence time andtracking accuracy of ship-borne phased array radar usingthe proposed GA based adaptive 120572-120573-120574 filter are superiorto the AEKF significantly In conclusions the proposed GAbased adaptive 120572-120573-120574 filter is a real time applicable algorithm
16 International Journal of Antennas and Propagation
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 36667 m
Figure 26 Estimation error of GA based adaptive 120572-120573-120574 filter
whose accuracy is good with relatively less complexity Inaddition the roll and pitch signalsmeasured in real operationenvironments can be replaced with the simulated roll andpitch signals to test all the relevant properties of practical shipmotion compensation and air target tracking for ship-bornephased array radar
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
This work is supported in part by research grants fromNational Science Council China (NSC 102-2218-E-155-001)
References
[1] Z Fang and Q Rundong Ship-borne Phased Array RadarMotion Compensation System Engineering and ElectronicTechnologies 1998
[2] L J Love J F Jansen and F G Pin ldquoCompensation of waveinduced motion and force phenomenon for ship based highperformance robotic and human amplifying systemsrdquo OakRidge National Laboratory ORNLTM-2003233 2003
[3] R J Hill Motion Compensation for Ship Borne Radars andLidars US Department of Commerce National Oceanic andAtmospheric Administration Office of Oceanic and Atmo-spheric Research Earth System Research Laboratory PhysicalSciences Division 2005
[4] T I Fossen and T Perez ldquoKalman filtering for positioning andheading control of ships and offshore rigsrdquo IEEEControl SystemsMagazine vol 29 no 6 pp 32ndash46 2009
[5] S Kuchler C Pregizer J K Eberharter K Schneider and OSawodny ldquoReal-time estimation of a shiprsquos attituderdquo in Proceed-ings of theAmericanControl Conference (ACC rsquo11) pp 2411ndash2416San Francisco Calif USA July 2011
[6] J Mar K C Tsai Y T Wang and M B Basnet ldquoIntelligentmotion compensation for improving the tracking performanceof shipborne phased array radarrdquo International Journal ofAntennas and Propagation vol 2013 Article ID 384756 14pages 2013
[7] T Dirk and S Tarunraj ldquoOptimal design of 120572-120573-(120574) filtersrdquo inAmerican Control Conference vol 6 pp 4348ndash4352 ChicagoIll USA June 2000
[8] D Tenne and T Singh ldquoCharacterizing performance of 120572-120573-120574filtersrdquo IEEE Transactions on Aerospace and Electronic Systemsvol 38 no 3 pp 1072ndash1087 2002
[9] T E Lee J P Su and K W Yu ldquoParameter optimization fora third-order sampled-data trackerrdquo in Proceedings of the 2ndInternational Conference on Innovative Computing Informationand Control p 336 Kumamoto Japan September 2007
[10] Y H Lho and J H Painter ldquoA fuzzy tuned adaptive Kalmanfilterrdquo in Industrial FuzzyControl and Intelligent System pp 144ndash148 December 1993
[11] N Ramakoti A Vinay and R K Jatoth ldquoParticle swarm opti-mization aided Kalman filter for object trackingrdquo in Proceedingsof the International Conference on Advances in ComputingControl and Telecommunication Technologies (ACT rsquo09) pp 531ndash533 Kerela India December 2009
[12] T-E Lee J-P Su andK-WYu ldquoAparticle swarmoptimization-based state estimation scheme formoving objectsrdquoTransactionsof the Institute of Measurement and Control vol 34 no 2-3 pp236ndash254 2012
[13] D C Law S A McLaughlin M J Post et al ldquoAn electronicallystabilized phased array system for shipborne atmospheric windprofilingrdquo Journal of Atmospheric and Oceanic Technology vol19 no 6 pp 924ndash933 2002
[14] C A Balanis AntennaTheory Analysis and Design JohnWileyamp Sons New York NY USA 3rd edition 2005
[15] J Mar and Y R Lin ldquoImplementation of SDR digital beam-former for microsatellite SARrdquo IEEE Geoscience and RemoteSensing Letters vol 6 no 1 pp 92ndash96 2009
[16] R C Eberhart and Y Shi Computational Intelligence Conceptsto Implementations ElsevierMorgan Kaufmann 2007
[17] K Y Lee and J-B Park ldquoApplication of particle swarm optimi-zation to economic dispatch problem advantages anddisadvan-tagesrdquo in Proceedings of the IEEE PES Power Systems Conferenceand Exposition (PSCE rsquo06) pp 188ndash192 Atlanta Ga USANovember 2006
[18] R C Eberhart and J Kennedy ldquoNew optimizer using particleswarm theoryrdquo in Proceedings of the 6th International Sympo-sium onMicroMachine and Human Science pp 39ndash43 NagoyaJapan October 1995
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
14 International Journal of Antennas and Propagation
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 187684 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 20 Estimation error of PSO based adaptive 120572-120573-120574 filter
0 100 200 300 400 500 600 700 800 900 1000Time (s)
Average estimation error = 199594 m
0
10
20
30
40
50
60
70
80
90
100
Et
(m)
Figure 21 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
measured data (with beam pointing error)) As the scenariodescribed in Case 3 the measured roll and pitch signalsshown in Figure 7 are used to evaluate the performance ofGA based adaptive 120572-120573-120574 filter for estimating the roll andpitch angles and tracking the circular trajectory target Theestimation errors of 120572-120573-120574 filter with fixed gains of 119892
1199031=
1198921199011
= 1198922= 01 and 119892
1199031= 1198921199011
= 1198922= 075 are shown in
Figures 24 and 25 respectivelyThe average estimation errorsfor fixed filter gains of 01 and 075 are about 41638m and572386m respectively The optimal roll gain 119892
1199031 pitch gain
1198921199011 and target tracking gain 119892
2of adaptive 120572-120573-120574 filter
estimated by the GA method during the observation timeof 419 sec are listed in Table 8 The estimation error of GAbased adaptive 120572-120573-120574 filter for tracking a circular flight targetis shown in Figure 26 where the convergence time of GAbased adaptive 120572-120573-120574 filter is about 5 sec and the average
0
500
1000
1500
0 200 400 600 800 1000Time (s)
Average estimation error = 1080515 m
Et
(m)
Figure 22 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
0 200 400 600 800 1000Time (s)
Average estimation error = 4279449 m
Et
(m)
Figure 23 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
09
estimation error is about 36667m It concludes that the GAbased adaptive 120572-120573-120574 filter can greatly improve the trackingaccuracy and convergence time of the conventional 120572-120573-120574filter for tracking the circular trajectory target
5 Conclusions
This paper proposed an intelligent beam pointing error com-pensationmechanism for ship-borne phased array radarTheGA based adaptive 120572-120573-120574 filter estimates the roll and pitchangles of the ship moving in the sea and thus compensatesfor the antenna beam pointing error in order to enhance thetracking accuracy of phased array radar system
International Journal of Antennas and Propagation 15
Table 6 Estimated gain parameters for PSO based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0366 03645 0532 04519 05151 05002 04797 05254 04146 053511198921199011
0366 03647 03634 03644 03643 0363 03644 03684 03643 036421198922
0125 01271 02286 02166 01169 01799 01735 01564 00767 01346
Table 7 Estimation errors for different gains
Calculated by GA Calculated by PSO Fixed gain1198921199031
0459sim0583 03645sim05352 01 075 091198921199011
0498sim0667 0363sim03684 01 075 091198922
00586sim01697 00367sim02291 01 075 09Average 119864
119905(m) 160756 187684 199594 1080515 4279449
Table 8 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim42 43sim84 85sim126 127sim168 169sim210 211sim252 253sim294 295sim336 337sim378 379sim4191198921199031
0619 04996 05316 0644 05052 05568 05316 04459 05092 059241198921199011
0582 06727 07201 04913 04168 07169 05983 02036 05858 06211198922
0214 02215 01612 02474 01822 0157 02073 01211 02278 01074
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 41638 m
Figure 24 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
Six experimental cases are conducted to verify the perfor-mance of ship-borne phased array radar using the proposedGA based adaptive 120572-120573-120574 filter In Case 1 the GA basedadaptive 120572-120573-120574 filter is used to estimate the roll and pitchangles with the simulated and measured roll and pitchsignals respectively The simulation results show that themeasurement data has the minimum estimation error theestimation error in the sea state 2 is the second and theestimation error in the sea state 3 is the worst Cases 2 and 3show that when the GA based adaptive 120572-120573-120574 filter is appliedto estimate the beam pointing error and to track the targetin a ship-borne phased array radar the circular trajectory
0 50 100 150 200 250 300 350 400 450Time (s)
0
500
1000
1500
Average estimation error = 572386 m
Et
(m)
Figure 25 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
target will be tracked with the larger average estimation errorand longer convergence time than linear trajectory targetTheexperimental results ofCases 4 and 5demonstrate thatwhen alinear trajectory target is tracked by ship-borne phased arrayradar the average estimation error and convergence time ofship-borne phased array radar using the GA based adaptive120572-120573-120574 filter are better than PSO based adaptive 120572-120573-120574 filterand fixed filter gain 120572-120573-120574 filter The convergence time andtracking accuracy of ship-borne phased array radar usingthe proposed GA based adaptive 120572-120573-120574 filter are superiorto the AEKF significantly In conclusions the proposed GAbased adaptive 120572-120573-120574 filter is a real time applicable algorithm
16 International Journal of Antennas and Propagation
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 36667 m
Figure 26 Estimation error of GA based adaptive 120572-120573-120574 filter
whose accuracy is good with relatively less complexity Inaddition the roll and pitch signalsmeasured in real operationenvironments can be replaced with the simulated roll andpitch signals to test all the relevant properties of practical shipmotion compensation and air target tracking for ship-bornephased array radar
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
This work is supported in part by research grants fromNational Science Council China (NSC 102-2218-E-155-001)
References
[1] Z Fang and Q Rundong Ship-borne Phased Array RadarMotion Compensation System Engineering and ElectronicTechnologies 1998
[2] L J Love J F Jansen and F G Pin ldquoCompensation of waveinduced motion and force phenomenon for ship based highperformance robotic and human amplifying systemsrdquo OakRidge National Laboratory ORNLTM-2003233 2003
[3] R J Hill Motion Compensation for Ship Borne Radars andLidars US Department of Commerce National Oceanic andAtmospheric Administration Office of Oceanic and Atmo-spheric Research Earth System Research Laboratory PhysicalSciences Division 2005
[4] T I Fossen and T Perez ldquoKalman filtering for positioning andheading control of ships and offshore rigsrdquo IEEEControl SystemsMagazine vol 29 no 6 pp 32ndash46 2009
[5] S Kuchler C Pregizer J K Eberharter K Schneider and OSawodny ldquoReal-time estimation of a shiprsquos attituderdquo in Proceed-ings of theAmericanControl Conference (ACC rsquo11) pp 2411ndash2416San Francisco Calif USA July 2011
[6] J Mar K C Tsai Y T Wang and M B Basnet ldquoIntelligentmotion compensation for improving the tracking performanceof shipborne phased array radarrdquo International Journal ofAntennas and Propagation vol 2013 Article ID 384756 14pages 2013
[7] T Dirk and S Tarunraj ldquoOptimal design of 120572-120573-(120574) filtersrdquo inAmerican Control Conference vol 6 pp 4348ndash4352 ChicagoIll USA June 2000
[8] D Tenne and T Singh ldquoCharacterizing performance of 120572-120573-120574filtersrdquo IEEE Transactions on Aerospace and Electronic Systemsvol 38 no 3 pp 1072ndash1087 2002
[9] T E Lee J P Su and K W Yu ldquoParameter optimization fora third-order sampled-data trackerrdquo in Proceedings of the 2ndInternational Conference on Innovative Computing Informationand Control p 336 Kumamoto Japan September 2007
[10] Y H Lho and J H Painter ldquoA fuzzy tuned adaptive Kalmanfilterrdquo in Industrial FuzzyControl and Intelligent System pp 144ndash148 December 1993
[11] N Ramakoti A Vinay and R K Jatoth ldquoParticle swarm opti-mization aided Kalman filter for object trackingrdquo in Proceedingsof the International Conference on Advances in ComputingControl and Telecommunication Technologies (ACT rsquo09) pp 531ndash533 Kerela India December 2009
[12] T-E Lee J-P Su andK-WYu ldquoAparticle swarmoptimization-based state estimation scheme formoving objectsrdquoTransactionsof the Institute of Measurement and Control vol 34 no 2-3 pp236ndash254 2012
[13] D C Law S A McLaughlin M J Post et al ldquoAn electronicallystabilized phased array system for shipborne atmospheric windprofilingrdquo Journal of Atmospheric and Oceanic Technology vol19 no 6 pp 924ndash933 2002
[14] C A Balanis AntennaTheory Analysis and Design JohnWileyamp Sons New York NY USA 3rd edition 2005
[15] J Mar and Y R Lin ldquoImplementation of SDR digital beam-former for microsatellite SARrdquo IEEE Geoscience and RemoteSensing Letters vol 6 no 1 pp 92ndash96 2009
[16] R C Eberhart and Y Shi Computational Intelligence Conceptsto Implementations ElsevierMorgan Kaufmann 2007
[17] K Y Lee and J-B Park ldquoApplication of particle swarm optimi-zation to economic dispatch problem advantages anddisadvan-tagesrdquo in Proceedings of the IEEE PES Power Systems Conferenceand Exposition (PSCE rsquo06) pp 188ndash192 Atlanta Ga USANovember 2006
[18] R C Eberhart and J Kennedy ldquoNew optimizer using particleswarm theoryrdquo in Proceedings of the 6th International Sympo-sium onMicroMachine and Human Science pp 39ndash43 NagoyaJapan October 1995
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
International Journal of Antennas and Propagation 15
Table 6 Estimated gain parameters for PSO based adaptive 120572-120573-120574 filter
Time (sec) 1sim100 101sim200 201sim300 301sim400 401sim500 501sim600 601sim700 701sim800 801sim900 901sim10001198921199031
0366 03645 0532 04519 05151 05002 04797 05254 04146 053511198921199011
0366 03647 03634 03644 03643 0363 03644 03684 03643 036421198922
0125 01271 02286 02166 01169 01799 01735 01564 00767 01346
Table 7 Estimation errors for different gains
Calculated by GA Calculated by PSO Fixed gain1198921199031
0459sim0583 03645sim05352 01 075 091198921199011
0498sim0667 0363sim03684 01 075 091198922
00586sim01697 00367sim02291 01 075 09Average 119864
119905(m) 160756 187684 199594 1080515 4279449
Table 8 Estimated gain parameters for GA based adaptive 120572-120573-120574 filter
Time (sec) 1sim42 43sim84 85sim126 127sim168 169sim210 211sim252 253sim294 295sim336 337sim378 379sim4191198921199031
0619 04996 05316 0644 05052 05568 05316 04459 05092 059241198921199011
0582 06727 07201 04913 04168 07169 05983 02036 05858 06211198922
0214 02215 01612 02474 01822 0157 02073 01211 02278 01074
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 41638 m
Figure 24 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
01
Six experimental cases are conducted to verify the perfor-mance of ship-borne phased array radar using the proposedGA based adaptive 120572-120573-120574 filter In Case 1 the GA basedadaptive 120572-120573-120574 filter is used to estimate the roll and pitchangles with the simulated and measured roll and pitchsignals respectively The simulation results show that themeasurement data has the minimum estimation error theestimation error in the sea state 2 is the second and theestimation error in the sea state 3 is the worst Cases 2 and 3show that when the GA based adaptive 120572-120573-120574 filter is appliedto estimate the beam pointing error and to track the targetin a ship-borne phased array radar the circular trajectory
0 50 100 150 200 250 300 350 400 450Time (s)
0
500
1000
1500
Average estimation error = 572386 m
Et
(m)
Figure 25 Estimation error of 120572-120573-120574 filter with 1198921199031
= 1198921199011
= 1198922=
075
target will be tracked with the larger average estimation errorand longer convergence time than linear trajectory targetTheexperimental results ofCases 4 and 5demonstrate thatwhen alinear trajectory target is tracked by ship-borne phased arrayradar the average estimation error and convergence time ofship-borne phased array radar using the GA based adaptive120572-120573-120574 filter are better than PSO based adaptive 120572-120573-120574 filterand fixed filter gain 120572-120573-120574 filter The convergence time andtracking accuracy of ship-borne phased array radar usingthe proposed GA based adaptive 120572-120573-120574 filter are superiorto the AEKF significantly In conclusions the proposed GAbased adaptive 120572-120573-120574 filter is a real time applicable algorithm
16 International Journal of Antennas and Propagation
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 36667 m
Figure 26 Estimation error of GA based adaptive 120572-120573-120574 filter
whose accuracy is good with relatively less complexity Inaddition the roll and pitch signalsmeasured in real operationenvironments can be replaced with the simulated roll andpitch signals to test all the relevant properties of practical shipmotion compensation and air target tracking for ship-bornephased array radar
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
This work is supported in part by research grants fromNational Science Council China (NSC 102-2218-E-155-001)
References
[1] Z Fang and Q Rundong Ship-borne Phased Array RadarMotion Compensation System Engineering and ElectronicTechnologies 1998
[2] L J Love J F Jansen and F G Pin ldquoCompensation of waveinduced motion and force phenomenon for ship based highperformance robotic and human amplifying systemsrdquo OakRidge National Laboratory ORNLTM-2003233 2003
[3] R J Hill Motion Compensation for Ship Borne Radars andLidars US Department of Commerce National Oceanic andAtmospheric Administration Office of Oceanic and Atmo-spheric Research Earth System Research Laboratory PhysicalSciences Division 2005
[4] T I Fossen and T Perez ldquoKalman filtering for positioning andheading control of ships and offshore rigsrdquo IEEEControl SystemsMagazine vol 29 no 6 pp 32ndash46 2009
[5] S Kuchler C Pregizer J K Eberharter K Schneider and OSawodny ldquoReal-time estimation of a shiprsquos attituderdquo in Proceed-ings of theAmericanControl Conference (ACC rsquo11) pp 2411ndash2416San Francisco Calif USA July 2011
[6] J Mar K C Tsai Y T Wang and M B Basnet ldquoIntelligentmotion compensation for improving the tracking performanceof shipborne phased array radarrdquo International Journal ofAntennas and Propagation vol 2013 Article ID 384756 14pages 2013
[7] T Dirk and S Tarunraj ldquoOptimal design of 120572-120573-(120574) filtersrdquo inAmerican Control Conference vol 6 pp 4348ndash4352 ChicagoIll USA June 2000
[8] D Tenne and T Singh ldquoCharacterizing performance of 120572-120573-120574filtersrdquo IEEE Transactions on Aerospace and Electronic Systemsvol 38 no 3 pp 1072ndash1087 2002
[9] T E Lee J P Su and K W Yu ldquoParameter optimization fora third-order sampled-data trackerrdquo in Proceedings of the 2ndInternational Conference on Innovative Computing Informationand Control p 336 Kumamoto Japan September 2007
[10] Y H Lho and J H Painter ldquoA fuzzy tuned adaptive Kalmanfilterrdquo in Industrial FuzzyControl and Intelligent System pp 144ndash148 December 1993
[11] N Ramakoti A Vinay and R K Jatoth ldquoParticle swarm opti-mization aided Kalman filter for object trackingrdquo in Proceedingsof the International Conference on Advances in ComputingControl and Telecommunication Technologies (ACT rsquo09) pp 531ndash533 Kerela India December 2009
[12] T-E Lee J-P Su andK-WYu ldquoAparticle swarmoptimization-based state estimation scheme formoving objectsrdquoTransactionsof the Institute of Measurement and Control vol 34 no 2-3 pp236ndash254 2012
[13] D C Law S A McLaughlin M J Post et al ldquoAn electronicallystabilized phased array system for shipborne atmospheric windprofilingrdquo Journal of Atmospheric and Oceanic Technology vol19 no 6 pp 924ndash933 2002
[14] C A Balanis AntennaTheory Analysis and Design JohnWileyamp Sons New York NY USA 3rd edition 2005
[15] J Mar and Y R Lin ldquoImplementation of SDR digital beam-former for microsatellite SARrdquo IEEE Geoscience and RemoteSensing Letters vol 6 no 1 pp 92ndash96 2009
[16] R C Eberhart and Y Shi Computational Intelligence Conceptsto Implementations ElsevierMorgan Kaufmann 2007
[17] K Y Lee and J-B Park ldquoApplication of particle swarm optimi-zation to economic dispatch problem advantages anddisadvan-tagesrdquo in Proceedings of the IEEE PES Power Systems Conferenceand Exposition (PSCE rsquo06) pp 188ndash192 Atlanta Ga USANovember 2006
[18] R C Eberhart and J Kennedy ldquoNew optimizer using particleswarm theoryrdquo in Proceedings of the 6th International Sympo-sium onMicroMachine and Human Science pp 39ndash43 NagoyaJapan October 1995
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
16 International Journal of Antennas and Propagation
0
20
40
60
80
100
120
140
160
180
200
Et
(m)
0 50 100 150 200 250 300 350 400 450Time (s)
Average estimation error = 36667 m
Figure 26 Estimation error of GA based adaptive 120572-120573-120574 filter
whose accuracy is good with relatively less complexity Inaddition the roll and pitch signalsmeasured in real operationenvironments can be replaced with the simulated roll andpitch signals to test all the relevant properties of practical shipmotion compensation and air target tracking for ship-bornephased array radar
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
This work is supported in part by research grants fromNational Science Council China (NSC 102-2218-E-155-001)
References
[1] Z Fang and Q Rundong Ship-borne Phased Array RadarMotion Compensation System Engineering and ElectronicTechnologies 1998
[2] L J Love J F Jansen and F G Pin ldquoCompensation of waveinduced motion and force phenomenon for ship based highperformance robotic and human amplifying systemsrdquo OakRidge National Laboratory ORNLTM-2003233 2003
[3] R J Hill Motion Compensation for Ship Borne Radars andLidars US Department of Commerce National Oceanic andAtmospheric Administration Office of Oceanic and Atmo-spheric Research Earth System Research Laboratory PhysicalSciences Division 2005
[4] T I Fossen and T Perez ldquoKalman filtering for positioning andheading control of ships and offshore rigsrdquo IEEEControl SystemsMagazine vol 29 no 6 pp 32ndash46 2009
[5] S Kuchler C Pregizer J K Eberharter K Schneider and OSawodny ldquoReal-time estimation of a shiprsquos attituderdquo in Proceed-ings of theAmericanControl Conference (ACC rsquo11) pp 2411ndash2416San Francisco Calif USA July 2011
[6] J Mar K C Tsai Y T Wang and M B Basnet ldquoIntelligentmotion compensation for improving the tracking performanceof shipborne phased array radarrdquo International Journal ofAntennas and Propagation vol 2013 Article ID 384756 14pages 2013
[7] T Dirk and S Tarunraj ldquoOptimal design of 120572-120573-(120574) filtersrdquo inAmerican Control Conference vol 6 pp 4348ndash4352 ChicagoIll USA June 2000
[8] D Tenne and T Singh ldquoCharacterizing performance of 120572-120573-120574filtersrdquo IEEE Transactions on Aerospace and Electronic Systemsvol 38 no 3 pp 1072ndash1087 2002
[9] T E Lee J P Su and K W Yu ldquoParameter optimization fora third-order sampled-data trackerrdquo in Proceedings of the 2ndInternational Conference on Innovative Computing Informationand Control p 336 Kumamoto Japan September 2007
[10] Y H Lho and J H Painter ldquoA fuzzy tuned adaptive Kalmanfilterrdquo in Industrial FuzzyControl and Intelligent System pp 144ndash148 December 1993
[11] N Ramakoti A Vinay and R K Jatoth ldquoParticle swarm opti-mization aided Kalman filter for object trackingrdquo in Proceedingsof the International Conference on Advances in ComputingControl and Telecommunication Technologies (ACT rsquo09) pp 531ndash533 Kerela India December 2009
[12] T-E Lee J-P Su andK-WYu ldquoAparticle swarmoptimization-based state estimation scheme formoving objectsrdquoTransactionsof the Institute of Measurement and Control vol 34 no 2-3 pp236ndash254 2012
[13] D C Law S A McLaughlin M J Post et al ldquoAn electronicallystabilized phased array system for shipborne atmospheric windprofilingrdquo Journal of Atmospheric and Oceanic Technology vol19 no 6 pp 924ndash933 2002
[14] C A Balanis AntennaTheory Analysis and Design JohnWileyamp Sons New York NY USA 3rd edition 2005
[15] J Mar and Y R Lin ldquoImplementation of SDR digital beam-former for microsatellite SARrdquo IEEE Geoscience and RemoteSensing Letters vol 6 no 1 pp 92ndash96 2009
[16] R C Eberhart and Y Shi Computational Intelligence Conceptsto Implementations ElsevierMorgan Kaufmann 2007
[17] K Y Lee and J-B Park ldquoApplication of particle swarm optimi-zation to economic dispatch problem advantages anddisadvan-tagesrdquo in Proceedings of the IEEE PES Power Systems Conferenceand Exposition (PSCE rsquo06) pp 188ndash192 Atlanta Ga USANovember 2006
[18] R C Eberhart and J Kennedy ldquoNew optimizer using particleswarm theoryrdquo in Proceedings of the 6th International Sympo-sium onMicroMachine and Human Science pp 39ndash43 NagoyaJapan October 1995
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of