Short-Range Positioning Systems: Fundamentals and Advanced … Positioning... · The purpose of any localization algorithm is, given a set of measurements, to find the locations of

University of Bologna, Engineering Faculty, July 5-8, 2010

Short Course for Doctoral Students

Short-Range Positioning Systems: Short Range Positioning Systems: Fundamentals and Advanced Research

Results with Case StudiesResults with Case Studies

Davide Dardari and Andrea Conti (*)( )

WiLAB –University of Bologna, ItalyDepartment of Electronics, Computer Science and Systemsp p y

(*) also with ENDIF, University of Ferrara, Italy

([email protected], [email protected])

2 Summary (1/2)

• Introduction– Motivations

L li i b i– Localization basics

• First part (D. Dardari): Distance estimation (ranging)– Basic concepts on estimation theory

– Performance limits in Time-of-Arrival (TOA) estimation( )

– Ultra-wide bandwidth (UWB) signals

– Ranging with UWB signals

– Main sources of error in TOA estimationMain sources of error in TOA estimation

– Practical TOA estimators

– The IEEE 802.15.4a standard

– Advanced issues: – Advanced issues: – Interference mitigation – Cognitive ranging – Secure Ranging

– NLOS condition - Passive Ranging & Loc.

D. Dardari, A. Conti, WiLAB c/o University of Bologna and ENDIF at University of Ferrara

3 Summary (2/2)

• Second part (A. Conti): Localization– Ingredients for localization

– A possible classification

– Localization schemes

– Cooperative localization

– Tracking

– Experimental resultsp

• In addition:– Case studiesCase studies


4 Introduction

Locating is the process used to determine the locationg pof one position relative to other defined positions

Since ever it is a fundamental need of the human being

I h h l i

? Cartography

In the pre-technologic era:

? Cartography

Signalingg g

Stars observation


5 Introduction

In the technologic era it is possible to localize persons and objectsg p p jin real-time

For example: GPS (Global Positioning System)


6 Introduction

But in many cases:

• nodes are GPS-denied (e.g., indoor, urban canyon)

• GPS is too expensive (e.g., wireless sensor networks) and only a small fraction of nodes (anchors) have positioning information due to cost, ( ) gsize and power constraints

• higher accuracy than GPS is required

• no infrastructure available (anchor-free) only relative coordinates are no infrastructure available (anchor free), only relative coordinates are estimated (ah hoc networks)

Sh i l i i i Short-range terrestrial positioning systems


7 Why short-range localization is important

• new context aware applications (e.g., real-time location systems, RTLS)• new market opportunities ($6 Billion in 2017*)

* R Das and P Harrop ”RFID Forecast Players and Opportunities 2007 2017” 2007 http://www idtechex com

Some examples:i t / l t ki (RFID)

* R. Das and P. Harrop RFID Forecast, Players and Opportunities 2007-2017 , 2007, http://www.idtechex.com.

• inventory/people tracking (RFID)• surveillance/security• wireless sensor networks (WSN)• virtual immersive and augmented reality applications• virtual immersive and augmented reality applications

[VICOM Project


[VICOM Project www.vicom-project.com]

8 For example in WSNs……

Sensed data without position and time information is often meaninglessSensed data without position and time information is often meaningless

For example, in habitat environments monitored sensed events must be ordered both in time and space to permit a correct interpretation.


9 In addition…

In many schemes proper time synchronization is required to achieve high ranging accuracies in positioning techniques

Positioning and time synchronization are also essential for basic mechanisms composing the wireless network to work efficiently:

•MAC scheduling algorithms can reduce packet collisions•power-saving strategies (wake-up sleeping times)•networking protocols to improve performance of routing algorithms (geo-routing)•enabling interference avoidance techniques in future cognitive radios


10 Needs

• high localization accuracy (<1m) in harsh environments (e.g., indoor)

• low device complexity (low cost, low size)

• standards (e.g., IEEE802.15.4a)


11

Localization basics


12 Localization Basics

The purpose of any localization algorithm is, given a set of The purpose of any localization algorithm is, given a set of measurements, to find the locations of target nodes with unknown positions.

P iti i i t i tPositioning occurs in two main steps:

1. Selected measurements are conducted between nodes1. Selected measurements are conducted between nodes2. Measurements are combined to determine the locations

of target nodes (localization algorithm).


13 Position estimation techniques classification (1/2)

•Anchor-based - Using GPS or using predefined coordinates, a subset of Anchor based Using GPS or using predefined coordinates, a subset of nodes in the network know their position a priori (these nodes are typically named anchors or beacons). Nodes with unknown positions (targets) use positioning information from anchors to determine their (targets) use positioning information from anchors to determine their location.

R i b t h d th d b bt i d bRanging between anchors and other nodes can be obtained by:- direct interaction (Single-Hop), - indirectly by means of intermediate nodes (Multi-Hop).

•Anchor-free - No nodes in the network have knowledge of their position a priori. Only relative coordinates can be found in such networks.


14 Position estimation techniques classification (2/2)

•Range-based - Measurements provide distance information among nodes (*)

•Angle-based - Measurements provide angle information among nodes

•Range-free - Only connectivity information is used (*)

Other techniques

•Interferometric – (low complexity, problems with multipath) (*)Interferometric (low complexity, problems with multipath) ( )•Scene analysis – (e.g., based on receiver power “signature”)•Inertial – (error accumulation)•DC magnetic tracker – (accurate but expensive and low range)•DC magnetic tracker – (accurate but expensive and low range)•Optical – (laser ranging systems, very accurate but expensive)•Hybrid – (e.g., Range-free and Inertial) (*)


(*) suitable for low cost devices (e.g., WSNs)

15

Basic Concepts on

Estimation Theory


16 Estimation theory basics

Problem statement

To obtain an estimate θ of the unknown parameter θpbased on the observation vector r = [r1, r2, ...., rN ]

T

estimator: θ = θ(r)

Th bl fi d h θ( )The problem is to find the estimator θ(r)which maximizes some performance metric(or minimizes a desired cost function)

Note : r could eventually represent a continuos time function r(t)

( f )

Note : r could eventually represent a continuos-time function r(t)through a suitable orthonormal base


17 Data model (observation model)

Data model

θ random, unknown: p(r, θ) = p(r|θ)p(θ), p(θ) prior information

Bayesian estimation approach

( ) ( | ) ( ) ( )

Bayesian estimation approach

θ deterministic unknown: p(r; θ) probability density function (PDF)

Classical estimation approach

These PDFs are the laws that govern the effect of θ on the observation r


18 Design Criteria for Bayesian Estimators (1/3)

RecallRecall

θ random, unknown: p(r, θ) = p(r|θ)p(θ), p(θ) prior information

Square Error (SE) cost function C(²) = ²2

Selection of a Cost Function

q ( ) ( )

where ² = θ(r)− θ is the estimation error

“Hit-or-Miss” cost functionC(²) =

½0 |²| ≤ ∆/21 |²| > ∆/2

½1 |²| > ∆/2

Estimator design

Our goal is to find an estimate that minimizes

the expected value of the cost


the expected value of the cost

19

S ( S )

Design Criteria for Bayesian Estimators (2/3)

Minimum Mean Square Error (MMSE) Estimator

Minimizes the MSE MSE = Eθ©²2ª

Minimizes the MSE MSE = Eθ,r©²ª

θ(r) = Eθ|r{θ|r} =Rθ p(θ|r) dθ posteriori mean

where the expectation is with respect to the posteriori PDF

( ) θ|r{ | }Rp( | ) p

p(θ|r) =p(r|θ)p(θ)Rp(r|θ)p(θ) dθ

=p(r|θ)p(θ)

p(r)

The estimation error has zero mean

Difficult to implement in the non-Gaussian case


p

20

( )

Design Criteria for Bayesian Estimators (3/3)

Maximum A Posteriori (MAP) Estimator

Minimizes the “hit-or-miss” cost function

θ(r) = argmaxθp(θ|r)

p(θ|r)

N l f l i il bl ˆ θNo general formula is available

If r and θ are jointly Gaussian, it is equivalent to the MMSE estimator

θ θ


21 Design Criteria for Classical Estimators (1/6)

Recall: Recall:

θ deterministic unknown: p(r; θ) PDF

Minimum Mean Square Error (MMSE)

MSE = Ern(θ(r)− θ)2

o= Var (²) + Er{²}

Similar to the Bayesian case (here the expectation is only over r), but in general not realizable since the estimator depends on the parameter to be estimated! parameter to be estimated!

W d th f lit f ti ti


We need other measures of quality of estimation process


Minimum Variance Unbiased (MVU) Estimator

1) Impose zero bias (unbiased estimator) Er{²} = 0

2) Minimize the variance of estimation error

θ(r) = argminθ(r)

En(θ(r)− θ)2

o= argmin

θ(r)Var (²)

The MVU estimator does not always existWhen exists no straightforward procedure are available to find the When exists, no straightforward procedure are available to find the estimator


23

C (C )

Design Criteria for Classical Estimators (3/6)

The Cramer-Rao Lower Bound (CRB)

1

CRB =³−E

n∂2 log p(r;θ)

∂θ2

o´−1

The CRB provides a minimum bound on the variance of any unbiased estimator:unbiased estimator:

Var (²) = En(θ(r)− θ)2

o≥ CRB( )

n( ( ) )

o≥

Def. Efficient estimator: if it attains the CRB for all θ



Maximum Likelihood (ML) Estimator

ˆ

p(r; θ)

θ(r) = argmaxθp(r; θ)

θθ

Not optimal in general, but straightforward

It is asymptotically efficient, i.e., asymptotically it is the MVU estimatorand the MSE tends to the CRB

If an efficient estimator exists, the MLE will produce it


25

C G

Design Criteria for Classical Estimators: example

Estimation of DC level in AWGN noise

rk = θ + nk for k = 1, 2, . . . , N E©n2kª= σ2Model: k k , , ,

p(r; θ) = 1 expn PN

k=1(rk−θ)2o

©k

ª

PDF: p(r; θ) = √2πσ2

expn− 2σ2

oPDF:

θ(r) = 1N

PNk=1 rk ³

ˆ´

2

ML estimator:

Var (²) = Var³θ(r)

´= σ2

N

CRB = σ2

Performance:

CRLB:

E {²} = 0

This is a MVU estimator, the best we can do!

CRB = NCRLB:


,

26

S ( S)

Design Criteria for Classical Estimators (6/6)

Least Squares (LS) Estimator

r = s(θ) + nObservation model ( )

No probabilistic assumptions are made about the observation

θ(r) = argminθ(r− s(θ))T ((r− s(θ))

Not optimal in general, but straightforward

Minimizing the LS error does not in general translate into minimizing the estimation error

If n is Gaussian, then the LS estimator is equivalent to the ML estimator


27

Distance estimation:

Ranging


28 Distance Estimation: Ranging

Ranging between two nodes is the technique employed by two nodes in g g q p y ythe network to determine the physical distance between them.

Node A Node B

Possible technologies

• Received Signal Strength (RSS)– direct RSS-distance mapping

f– interferometric

• Time-Based– e.m. waves

– ultrasound

• Near field (phase difference between the E & H fields in the near range) (Q trace corporation)


g ) (Q p )

29 Ranging based on RSS measurements

PR PRPR(dBm)

PR(dBm)

Unrealisticdeterministicchannel

Randomchannel

• Theoretical and empirical models are used to translate the difference between the

channel

d d

• Theoretical and empirical models are used to translate the difference between thetransmitted signal strength and the RSS into a range estimate.

• Propagation effects (refraction, reflection, shadowing, and multipath) cause theattenuation to poorly correlate with distanceattenuation to poorly correlate with distance

-1000,00 5,00 10,00 15,00 20,00 25,00 30,00

Inaccurate distance estimates

Ti h b d i i d 70-60-50-40-30-20

RSS

I (db

m)

Time synch between nodes is not required

-100-90-80-70

Distance (m)


30 RSS ranging: theoretical bound

Received power v.s. distance model (in dBm):p ( )

Pr(d) = P0 − 10 γ log10 d+ S

P0 received power (in dBm) at a reference distance of 1 meterγ: path-loss exponent (typiacl values between 2 and 5)S: Gaussian random variable with zero mean and standard deviation σS (shad-

) ( )owing spread). It accounts for large-scale random fluctuations (shadowing).

Cramer-Rao bound (CRB) for the distance estimation MSE

Var³d´≥

µln 10 σS

d

¶2•The MSE bound does not depend on signal structure

Var³d´≥

µ10 γ

d

¶.

•RSS ranging does not require time synchronization between nodes•No dedicated HW


Note: the CRB provides a minimum bound on the variance of any unbiased estimator.

31 Interferometric Ranging (1/2)

•Transmitters A and B transmit sinusoids at slightly different frequencies

•The envelope of the received composite signal, after band-pass filtering, will vary slowly over time.

•The phase offset of this envelope can be measured using cheap low-precision RF chips, and it contains information about the difference in distance of the two links.

M. Maroti, P. Völgyesi, S. Dora, B. Kusy, A. Nadas, A. Ledeczi, G. Balogh, and K. Molnar, “Radio interferometric l ti ” i P ACM C f E b dd d N t k d S S t S Di USA N 2005 1 12


geolocation,” in Proc. ACM Conference on Embedded Networked Sensor Systems, San Diego, USA, Nov. 2005, pp. 1–12.

32 Interferometric Ranging (2/2)

•To solve the unknown initial phase of the two transmitted sinusoids (no time synch is present), a similar measurement is made by node D in a different location

•The difference in phase offset measured at the two receivers only depends on the four distances. Hence, given a number of phase-offset-difference measurements, the unknown l f h b f dlocations of the two receivers may be inferred.

Comments:

•Sub-meter precision can be achieved in outdoor environments with simple hardware•Problems with severe multi-path (e.g., indoor)

•Infrastructure required


Infrastructure required.

33 Time-Based Ranging

The distance information between a pair of nodes A and B (ranging) can be( g g)obtained using the measurement of the propagation delay or Time-of-Flight(TOF) τf = d/c, where d is the actual distance between A and B.

• One-way Time-of Arrival (TOA) ranging

Node BNode A t1 t2 τf = t2 − t1

• Two-way TOA ranging

Node A Node B

τf =(t2−t1)+(t4−t3)

2


34 Time-Difference-of-Arrival (TDOA) (1/3)

TDOA scheme A TDOA scheme BO O

Networks with infrastructure (anchor nodes) and accurate anchors synchronization (e.g., through cable connections) are required



A typical approach uses a geometric interpretation to calculate the intersection of two or more hyperbolas: each sensor pair gives a hyperbola which represents of two or more hyperbolas: each sensor pair gives a hyperbola which represents the set of points at a constant range difference (time-difference) from two sensors



Networks with infrastructure (anchor nodes) and accurate anchors synchronization (e g through cable connections) are required synchronization (e.g., through cable connections) are required

TDOA h B it bl f t l l l it t t d TDOA scheme B suitable for extremely low complexity targets nodes (e.g., TAGs in UWB-RFID systems)


37

All i b d i h i i All time-based ranging techniques require

time-of-arrival (TOA) estimation of the received signal

For example, τf = 1 ns means a distance d ≈ 30 cm

Hence accurate time measurement Hence, accurate time measurement and TOA estimation are required!


38

Only an estimation of the real time t can be obtained due to

Time Measurements (1/2)

Only an estimation of the real time t can be obtained due to node local oscillator frequency drift

Reasonable model:

T(t) = (1 + δ)t+ µT(t) (1 + δ)t+ µ

δ: clock driftrelative to the correct raterelative to the correct rateµ: clock offset.

The rate of a perfect clock, dT(t)/dt, would equal 1 (i.e., δ = 0).p , ( )/ , q ( , )

The clock performance is often expressed in terms of part-per-million (ppm)defined as the maximum number of extra (or missed) clock counts over a total


of 106 counts, i.e., δ · 106.

39 Time Measurements (2/2)

Suppose that a node intends to generate a time delay of τd seconds,the effective generated delay τd in the presence of a clock drift δ would be

τd =τd1 + δ

In case a node has to measure a time interval of true durationτ = t2 − t1 seconds, the corresponding estimated value τ would be

τ = T(t2)− T(t1) = τ (1 + δ)

In both cases there is no dependence on the clock offset µ.


40 One-Way TOA Ranging: effect of clock drifts

According to node A’s local time, the packet is transmitted at time t(A)1 = TA(t1)

(included as a timestamp in the packet) and it is received at node B’s local time

t(B)2 = TB(t2). Node B calculates the estimated propagation delay as2 B( 2) p p g y

τf = t(B)2 − t

(A)1 = τf · (1 + δA) + t2 · (δB − δA) + µB − µA

clock offsets may be in the order of us, one a raging tied to net ork s nchroni ationone-way raging tied to network synchronization

One-way ranging requires stringent network synchronization constraints whichare often not feasible in low-cost systems.


41 Possible application of one-way TOA ranging

Ultrasound devicesUltrasound devices•propagation speed of acoustic waves (340 m/s) is much lower than light-speed •synchronization errors can be several orders of magnitude smaller than the typical propagation delay valuespropagation delay values

•high positioning accuracy (3 cm)

Ultrasound technology has several disadvantages

•Hybrid technology•Hybrid technology•Propagation limited by walls (coverage)•Effect of temperature and pressure (calibration)•Power hungry•Power-hungry

A. Harter, A. Hopper, P. Steggles, A. Ward and P. Webster ”The anatomy of a Context-Aware Application”, In Wireless

Active Bat localization system


Networks, Vol. 8, pp. 187-197, Feb. 2002.

42 Two-Way TOA Ranging: effect of clock drifts (1/2)

The effective response delay introduced by node B is τd/(1 + δB), whereas theestimated RTT denoted by τRT, according to node A’s time scale, is

τRT = 2 τf(1 + δA) +τd(1 + δA)

(1 + δB)

In absence of other information, node A derives the estimation of the propaga-tion time τ by equating the previous equation with the supposed round tription time, τf, by equating the previous equation with the supposed round-triptime 2 τf + τd leading to an error

τf − τf = τf δA +ε τd

≈ε τd

τf τf = τf δA +2(1 + δA − ε)

≈2(1 + δA − ε)

where ε , δA − δB .

τf: in the order of nanosecondsτd: in the order of microseconds/milliseconds (includes the time for packet


acquisition, sync., channel estimation, etc.)

43 Two-Way TOA Ranging: effect of clock drifts (2/2)

10−7

=1 s

10−8

)

τd=1 µs

τd=10 µs

τd=100 µs

τd=1 ms

τd=10 ms

10−10

10−9

Ran

ging

err

or (

s)

d=10 ms

10−11

R

10−7

10−6

10−5

10−12

ε

Low response delay must be ensured for high ranging accuracy

The ranging protocol must be implemented at PHY/MAC level


44 Mitigation of clock frequency offsets

Example:

• each node locally measures the (known) received packet duration and exchange its measure with the other node

• from the 2 measurements node A can evaluate a correction factor which applies to its estimated propagation time.

τf − τf = τf δA

pp p p g

• In this case the error becomes


45 Remarks

•Two-way ranging requires less stringent synchronization constraints compared to one-way ranging (relative clock drifts still affect ranging

)accuracy)

•Suitable for networks without infrastructure (e.g., ad hoc networks)

•Main scheme adopted in the IEEE 802.15.4a standard


46

Time-of-Arrival (TOA) estimation


47 Time-of-Arrival Estimation (TOA)

Consider the transmission of a single pulse p(t) in AWGN channel in the absence of

Problem statement in an ideal scenario

Consider the transmission of a single pulse p(t) in AWGN channel in the absence of other sources of error. The received signal is

r(t) =pEp p(t− τ) + n(t)

The problem is to obtain the best estimate for the TOA, τ , based on the receivedsignal r(t) observed over the interval [0, Tob).signal r(t) observed over the interval [0, Tob).

Classical non-linear parameter estimation problem


48 Maximum likelihood (ML) TOA estimator

matchedfilter

The ML TOA estimator is asymptotically efficient in AWGN


49 Theoretical limits in AWGN

The estimation Mean Square Error (MSE) of any unbiased estimation τ of τcan be bounded by the Cramer-Rao bound (CRB)

V (ˆ) E©(ˆ )2

ª≥ CRBVar (τ ) = E

©(τ − τ )2

ª≥ CRB

Th CRB i i bThe CRB is given by

CRB =N0/2

(2π)2E β2=

1

8π2 β2 SNR(2π)2Ep β2 8π2 β2 SNR

where− SNR , Ep/N0

2− β2 represents the second moment of the spectrum P (f) of p(t) defined by

β2 ,R∞−∞ f

2|P (f)|2 dfβ R∞

−∞ |P (f)|2 df

β ff ti b d idth


− β: effective bandwidth

50 Bound on ranging MSE

Lower bound on ranging MSELower bound on ranging MSE

V³d´ c2

Var³d´≥

c

8π2 β2 SNR.

How to improve the accuracy?

Increase the SNR towards very narrowband systems

To mitigate the effect of multipath in indoor environments, low frequency bands must be used not practical for small size devices due to large antennas

Increase the bandwidth towards ultra wideband (UWB) systems

Higher frequencies bands can be used. Multipath can be resolvable.Technique chosen by the IEEE 802.15.4a standard


51

Ultra-wide Bandwidth (UWB) Signals


52 Definition of a UWB signal

UWB signal is formally defined as any signal that (FCC):y g ( )

• occupies more than 500 MH f t MHz of spectrum or

• has a fractional bandwidth in excess of 20%

H L

H L

f fBF 0.2(f f ) / 2

−= ≥

+H L( ) /


53 Brief history of UWB

• The Radio was born as UWB: spark gap transmission designs of Marconi in the late 1890s

• After the Marconi’s “4 seven” patent (1900) the radio became “tuned”After the Marconi s 4 seven patent (1900) the radio became tuned

• Time-domain electromagnetic theory in early 1960s

• Short-pulse generators using tunnel diodes in early 1970s

• Applications to radar in 1970-80s

• The term “UWB” originated with the Defense Advanced Research Projects (DARPA) in a radar study undertaken in 1990

Fi di f ibl li i f UWB i i d i 1990• First studies of possible application of UWB to communication systems during 1990s(Win-Scholtz)

• 2/2002 - FCC adopted the First Report and Order (RO) that permits the marketing and / ( )operation of certain types of UWB products

• 2/2007 – Implementation of an UWB radio regulatory framework for the EU


54 Where do we find such a huge available bandwidth?


55

Better to utilize already allocated bands using extremely Better to utilize already allocated bands using extremely

low power spectral densities (PSD)

FCC limits ensure that UWB emission levels are exceedingly small• at or below spurious emission limits for all radios• at or below unintentional emitter limits• lowest limits ever applied by FCC to any systempp y y y

Part 15 limits equate to –41.25 dBm/MHz

For comparison, PSD limits for 2.4 GHz ISM and 5 GHz U-NII bands are +40 dB higher per MHz

Total emissions over several gigahertz of bandwidth are a small fraction of a milliwatt


56 Comparison between FCC and EU masks

−40

−50z]

−60

y [d

Bm

/MH

z]

−70

IRP

den

sity

[

−80

EIR

FCC maskEU mask

0 1 2 3 4 5 6 7 8 9 10 11Frequency [GHz]

−90


Frequency [GHz]

57 Why do we need UWB systems?

M ti tiMotivations Main issues

•High temporal resolution (l li i l i h) •Keep complexity low (localization, multipath)

•High material penetration (according to the band used)

•Keep complexity low

•Large bandwidth antennas

I t f t to the band used)

•Underlay technology (efficient spectrum usage)

•Interference management (coexistence)

p g )

•Multiple access

•Low probability of detection (LPD)Low probability of detection (LPD)

Narrowband (30kHz)

Wideband CDMA (5 MHz)

UWB (several GHz)

Part 15 Limit


UWB (several GHz)

Frequency

58 How to generate a UWB modulated signal

Conventional techniques over sinusoidal carrier

such as OFDM, DS-CDMA, FH-CDMA

Higher complexity

Impulse radiopu se ad o

Base-band transmission of short pulses (monocycles)

Lower complexity


59

Impulse Radio - UWB


60 Impulse Radio UWB (IR-UWB)

L l ( RF d IF )Base-band generation

of the signal

•Low complexity (no RF and IF stages)

•Low consumption

Example of Gaussian

6th derivative monocycle Spectrum6th derivative monocycle Spectrum

2

3

4

0

10

−1

0

1

Am

plitu

de

−20

−10

rmal

ized

PS

D [d

B]

normalized FCC maskpulse spectrum

−4

−3

−2

−1

−50

−40

−30norm


−0.5 −0.4 −0.3 −0.2 −0.1 0.0 0.1 0.2 0.3 0.4 0.5time (ns)

−40 1 2 3 4 5 6 7 8 9 10 11 12

Frequency (GHz)

−50

61 Example of simple impulse generator

From J.S. Lee, C. Nguyen e T. Scullion, IEEE Microwave Theory and Tech. 2001

Bit 0

Bipolar version realized at WiLAB – University of Bologna

Bit 0


62 IR-UWB: signaling


63

UWB signals are typically modulated pulse trains (low duty cycle)

IR-UWB: signaling

UWB signals are typically modulated pulse trains (low duty cycle)

At each bit several pulses (100-1000) are transmitted according to a certain “randomization” technique such as Time hopping (TH) or certain randomization technique such as Time-hopping (TH) or Direct Sequence (DS) to allow multiple access (each user adopts a different “code”)

Modulation techniques include pulse-position modulation (PPM),pulse amplitude modulation (PAM) and others

δ Tfδ f

TbPPM signaling


Randomized time hopping

64

UWB propagation

and channel models


65 The UWB channel

Multipath propagation( )

Rx Direct path

g(t): channel gainh(t): channel impulse responses(t): transmitted signalr(t): received signal

Tx UWB

UWBr(t): received signal

Multi-pathNarrowband channel:

Fading large link budget marginr(t) = g(t) · s(t)

Wideband channel

g g g g

Echoes signal distortion(t) h(t)⊗ (t) Echoes signal distortion

Single paths are typically not resolvable, i.e., echoes amplitudes exhibit fading

r(t) = h(t)⊗ s(t)

Ultra wideband channel

Echoes signal distortion r(t) = h(t)⊗ s(t)


Single paths may be resolved potential diversity gain

r(t) = h(t)⊗ s(t)

66 UWB Propagation Experiment

• Modern building with offices• Modern building with offices and labs

• Multipath profilesT = 300ns

• 14 rooms 49 idi t t i t49 equidistant points

• 7 x 7 measurement grid with 90 cm sidewith 90 cm side

Moe Z Win and Robert A Scholtz “Characterization of Ultra-Wide Bandwidth Wireless Indoor Channels: A Communication-


Moe Z. Win, and Robert A. Scholtz, Characterization of Ultra-Wide Bandwidth Wireless Indoor Channels: A Communication-Theoretic View”, IEEE Journal On Selected Areas In Communications, Vol. 20, No. 9, December 2002

67 Example of channel impulse responses

Room F1

LOS high SNR

Room P

NLOS high SNR

Room H

NLOS medium SNRNLOS medium SNR

Room B

NLOS low SNR

In general, dense multipath is present composed of tens or hundreds paths


68 Channel model for dense multipath: IEEE 802.15.4a

The IEEE 802.15.4a UWB channel model is based on an extended version of the The IEEE 802.15.4a UWB channel model is based on an extended version of the classical Saleh-Valenzuela (SV) indoor channel model, where multipath components arrive at the receiver in groups (clusters) following the Poisson distribution. According to the SV model, the complex baseband channel impulse g p presponse is given as

h0(t) =PK

k=0

PLl=0 ak,le

jφk,lδ (t− Tk − τk,l) ,0( )P

k=0

Pl=0 k,l ( k k,l) ,

ak,l: amplitude of the lth path in the kth clusterTk: delay of the kth clusterTk: delay of the kth clusterτk,l: delay of the lth path relative to the kth cluster arrival time Tkφk,l are uniformly distributed in the range [0, 2π)The number of clusters K is assumed to be Poisson distributed.The ray arrival times τk,l are modeled with mixtures of two Poisson processes

The Power Delay Profile (PDP) is assumed exponential negative within each cluster

The small-scale fading, which characterizes the path amplitudes ak,l, follows aNakagami-m distribution

The Power Delay Profile (PDP) is assumed exponential negative within each cluster

Λk,l = En|ak,l|

2o∝ exp(−τk,l/²k) ²k is the intra-cluster decay time constant.


69 Simplified channel model for lower frequencies

Another widely adopted model is the dense multipath model with a single cluster composed of L independent equally spaced paths and exponential PDP. In this case the real impulse response is given by

h(t) =PL

l=0 al pl δ (t− τl)

τl = τ1 + (l − 1)∆∆: resolvable time intervalal: path amplitude with Nakagami-m statistics

{ }

Exponential negative PDP

pl is a r.v. which takes, with equal probability, the values {−1,+1}

Λl = En|al|

2o= (e∆/²−1)e−∆(l−1)/²

e∆/²(1−eL∆/²) ²: decay time constant.

A. F. Molisch, D. Cassioli, C.-C. Chong, S. Emami, A. Fort, B. Kannan, J. Karedal, J. Kunisch, H. Schantz, K. Siwiak,d M Z Wi “A h i t d di d d l f lt id b d ti h l ” IEEE T A t


and M. Z. Win, “A comprehensive standardized model for ultrawideband propagation channels,” IEEE Trans. AntennasPropag., vol. 54, no. 11, pp. 3151–3166, Nov. 2006, Special Issue on Wireless Communications.

70

Ranging using UWB signals


71 Impulse Radio UWB (IR-UWB)

L l ( RF d IF )Base-band generation

of the signal

•Low complexity (no RF and IF stages)

•Low consumption

Example of Gaussian

6th derivative monocycle Spectrum6th derivative monocycle Spectrum

2

3

4

0

10

−1

0

1

Am

plitu

de

−20

−10

rmal

ized

PS

D [d

B]

normalized FCC maskpulse spectrum

−4

−3

−2

−1

−50

−40

−30norm


−0.5 −0.4 −0.3 −0.2 −0.1 0.0 0.1 0.2 0.3 0.4 0.5time (ns)

−40 1 2 3 4 5 6 7 8 9 10 11 12

Frequency (GHz)

−50

72 Theoretical Performance Limits

Theoretical performance limits (bounds) serve as useful benchmarks in Theoretical performance limits (bounds) serve as useful benchmarks in assessing the performance of newly developed TOA estimation techniques

The performance of the TOA estimator, like all non-linear estimators, is characterized by the presence of distinct SNR regions corresponding to different modes of operation:

- Low SNR region (a priori region), signal observations provide little new information- High SNR region (asymptotic region), the MSE is accurately described by the CRB- Medium SNR (transition region or ambiguity region), observations are subject to

bi i i h d f b h CRBambiguities that are not accounted for by the CRB

This behaviour is referred to as the threshold effect


73 Improved bounds

Unfortunately, the CRB is not accurate at low and moderate SNR and/or when the observation time is short improved bounds

The Ziv-Zakai Lower Bound

The Ziv-Zakai bound (ZZB) can be applied to a wider range of SNRs, ( ) pp gbut more complex than CRB for analytical evaluation


74 The Ziv-Zakai Lower Bound

When τ is uniformly distributed in [0, Ta), the ZZB is

ZZB =1

T

Z Ta

z (Ta − z)Pmin (z) dz

where Pmin (τ ) is the error probability of the classical binary detection scheme

Ta

Z0

( a ) min ( )

with equally probable hypothesis:

H1 : r(t) ∼ p {r(t)|τ}

H2 : r(t) ∼ p {r(t)|τ + z}

In AWGN the minimum attainable probability of error becomes

Pmin (z) = Q

µqSNR (1− ρp(z))

¶where Q (·) is the Gaussian Q-function and ρp(z) is the autocorrelation functionof p(t).


75 TOA estimation theoretical limits using UWB pulses

CRB and Ziv-Zakai Bound in AWGN (single path)

10-7

CRB, RRC, =3.2 ns

CRB and Ziv Zakai Bound in AWGN (single path)

10-8

CRB, RRC, τp=3.2 ns

ZZB, RRC, τp=3.2 ns

CRB, RRC, τp=1 ns

ZZB, RRC, τp=1 ns

CRB, Gauss. deriv., n=2

3 mA priori RMSE Ta/

√12 A priori region

10-10

10-9

SE

(s)


ZZB, Gauss. deriv., n=2


ZZB, Gauss. deriv., n=6

30 cm

3 cm

10-11

10

RM

S 3 cm

Ambiguity region

-13

10-12 Asymptotic region

0 5 10 15 20 25 30 35 40SNR (dB)

10-13

From D Dardari C -C Chong and M Z Win “Improved lower bounds on time-of-arrival estimation error in realistic


From D. Dardari, C.-C. Chong, and M. Z. Win, Improved lower bounds on time-of-arrival estimation error in realistic UWB channels,” in IEEE International Conference on Ultra-Wideband, ICUWB 2006, (Waltham, MA, USA), pp. 531–537, Sept. 2006.

76

Main sources of error

in TOA Ranging


77 Main Issues in UWB TOA Estimation

Rich multipathRich multipath(ambiguities in first path detection, paths overlapping)

Non Line-of-Sight (NLOS) g ( )

conditions•direct path blockage•extra propagation delay due •extra propagation delay due to different e.m. propagation speeds

Interference(narrowband and wideband)

In addition:• clock drift (time synch algorithms)


( y g )• design of low-complexity TOA estimators

78 TOA Estimation in Multipath Environments

L

r(t) =pEp

Xl=1

αl p(t− τl) + n(t)p(t)

In general, dense multipath is composed of tens or hundreds paths.It may be difficult to recognize the first path, especially at low and medium SNRs

detection of first path may be challengingPaths could be partially overlapped (not resolvable channel)


Moe Z. Win, and Robert A. Scholtz, “Characterization of Ultra-Wide Bandwidth Wireless Indoor Channels: A Communication-Theoretic View”, IEEE Journal On Selected Areas In Communications, Vol. 20, No. 9, December 2002

79 TOA Estimation in Multipath Environments

Received signalReceived signal

r(t) =pE

LXαl p(t τl) + n(t)r(t) =

pEp

Xl=1

αl p(t− τl) + n(t)

Problem formulation

• we are interested in the estimation of the ToA, τ = τ1, of the direct path byobserving the received signal r(t) within the observation interval [0, Tob);

• we consider τ to be uniformly distributed in the interval [0, Ta), withTa < Tob;

t f i t U { }• set of nuisance parameters U = {τ2, τ3, . . . , τL,α1,α2, . . . ,αL}


80 CRB and ZZB for TOA Estimation with Multipath

Using the IEEE802.15.4a channel model

10-8

CRB ZZB, CM1

g

10-9

ZZB, CM1ZZB, CM4ZZB, CM5

10-10

RM

SE

(s)

10-11

R

5 10 15 20 2510-12

5 10 15 20 25SNR (dB)

D. Dardari, A. Conti, U. Ferner, A. Giorgetti, and M. Z. Win, “Ranging with Ultrawide Bandwidth Signals in Multipath Environments”, Proc. of IEEE (Special Issue on UWB Technology & Emerging Applications), Feb. 2009.


Environments , Proc. of IEEE (Special Issue on UWB Technology & Emerging Applications), Feb. 2009.

81 ZZB for TOA Estimation using Measured Data

hi h i C. –C. Chong and S. K. Yong, “A generic statistical based UWB channel model for high-rise apartments,” IEEE Trans. Antennas and Propagation, vol. 53, no. 8, pp. 2389-2399, Aug. 2005.

high-rise apartment

10-7

10-8

10LOS (Rx2)LOS (Rx3)NLOS (Rx4)NLOS (Rx5)NLOS (Rx6)NLOS (Rx7)

10-10

10-9

RM

SE

(s)

NLOS (Rx7)NLOS (Rx8)

10-11

10-10

0 5 10 15 20 25 30SNR (dB)

10-12


D. Dardari, C.-C. Chong, and M. Z. Win, “Improved lower bounds on time-of-arrival estimation error in realistic UWB channels,” in IEEE International Conference on Ultra-Wideband, ICUWB 2006, (Waltham, MA, USA), pp. 531–537, Sept. 2006.

82 ML TOA Estimator in Multipath Environments

When channel parameters are unknown TOA estimation in multipath environ-When channel parameters are unknown, TOA estimation in multipath environments is closely related to channel estimation, where path amplitudes and TOAτ = [τ1, τ2, . . . , τL]

T are jointly estimated using, for example, a ML approach

The ML estimate of τ is τ = argmaxτ

{χH(τ )R−1(τ )χ(τ )}, where R(τ ) is the

t l ti t i f (t) dautocorrelation matrix of p(t) and ⎡⎢ p(t− τ1)p(t− τ2)

⎤⎥χ(τ ) ,

Z Tob

0

r(t)

⎢⎢⎢⎢⎢⎢⎣p(t τ2)

.

.

.

⎥⎥⎥⎥⎥⎥⎦ dt

Problems:

⎣p(t− τL)

⎦

• Implementation at Nyquist sampling rate or higher difficult with UWB signals

• High complexity


g p ylooking for sub-optimal schemes

83 Reducing the Estimator Complexity

Examples:

Sub optimal ML -Sub-optimal ML

-Generalized ML

Si l th h ldi-Simple thresholding

-Multi-resolution

( )-Delay & average (noisy template)

-Two-stages approaches

-Energy detection based

-……………………


84 Sub-optimal ML Algorithms

• They can be derived from the ML channel estimation criterion and based on a simple peak detection process

– Single Search– Search and Subtract– Search, Subtract and Readjust

These algorithms essentially involve the detection of N largest positive and negative values of the MF output and the determination of the corresponding time locationsof the MF output and the determination of the corresponding time locations .

While these algorithms are equivalent when the multipaths are separable, the last two algorithmsg q p p , g

take into account the effects of a non-separable channel (overlapped echoes)


85 Single Search

• The delay and amplitude vectors are estimated with a single look.

( )r t ( )y t ( )y t ˆ ˆk kcτ( ) ( )y

MF

( )Peak Detector

,i ik kcτ

1,2,...,ik N=

ABS ()

1k 2k

, , ,i

( )y t

3k

3k

( )y t


86 Search and Subtract

This algorithm provides a way to detect multipath components in a non-This algorithm provides a way to detect multipath components in a nonseparable channel.

( ) ( ) ( )1ir t r t r t−= −

( )r tMF Peak DetectABS ()

( )y t ( )y t( )r t

( )r t

1kτ 1,2,...,ik N=ˆ ˆ,i ik kcτ

received signal

Path estimatorpath 2 path 1

( ) ( )ˆ ˆ ˆi ii k kr t c p t τ= −

( )1ir t

received signalafter subtract


+( )1i t−

87 Search Subtract and Readjust

MF

( ) ( ) ( )1ir t r t r t−= −

( )y t ( )y tPeak Detect

τ

( ) ( )ph t p T t= −

MFABS ()

1 2k N

( )y t ( )y t

ikτ

Peak estimator

1,2,...,ik N=

So far (Search and Subtract) the delay and

{ }1

ˆ ˆj

i

k jc c=

Peak estimatoramplitude of each path are estimated separately at each step; in this algorithm a joint estimation of the amplitudes of different paths is introduced. { }

1jk j=

Path estimator

( ) ( )1ˆ ˆ ˆ

j j

ii k kj

r t c p t τ=

= −∑+ ( )1ir t−

( )r t


++

88 Simple Thresholding

Another classical approach is the Thresholding Estimator

thresholdthreshold

Threshold η

The optimum choice of η depends on channel statistics and SNR:

• Small η → high probability of early detection prior to the first path (earlyTOA estimation) due to noise and interferenceTOA estimation) due to noise and interference

• Large η→ low probability of detecting the first path and a high probabilityof detecting an erroneous path (late TOA estimation) due to fading


89 Performance in Realistic EnvironmentsResults from: C. Falsi, D. Dardari, L. Mucchi, and M. Z. Win, “Time of arrival estimation for UWB localizers in realistic environments,” EURASIP J. Appl. Signal Processing (Special Issue on Wireless Location Technologies and Applications), 2006.

Al i h

Lower complexity

Simple thresholding peak detectionAlgorithmstested

Simple thresholding, peak detection

Iterative cancellation techniques (to deal with paths )

Higher complexityoverlapping)

Decreasing SNR

Main observations:

-There is no a large performance difference between simple and complex algorithms between simple and complex algorithms considered (especially at medium and large SNRs)

-Ranging resolutions on the order of 10cm are Ranging resolutions on the order of 10cm are achievable!


90 Sub-Nyquist Sampling Rate TOA Estimators

TOA ti t b d d t ti (ED) t t bTOA estimators based on energy detection (ED) can operate at sub-Nyquist sampling rate low complexity

• T[.]: pre-processing filter– to improve the first path detection– to mitigate the effect of the interferenceto mitigate the effect of the interference

• First path detector: several algorithms can be adopted

The TOA resolution is bounded by the ED integration time


Asymptotic MSE = T 2int/12

91 Example of ranging preamble structure

Each user has a different time-hopping sequencepp g qto allow multi-user communication


92 ED-based TOA estimation algorithms

Collected energy vectorCollected energy vector

Si l th h ldiPossible algorithms

• Simple thresholding• P-MAX• Backward search• Jump Back and Search Forward• ………


93 Simple thresholding

Detect the time slot within the observation interval that contains the first pathDetect the time slot within the observation interval that contains the first pathby comparing each element of {zk} to a fixed threshold η.The first threshold crossing event is taken as the estimate of the TOA

Decision vector

The optimum choice of η depends on channel statistics and SNR:

• Small η → high probability of early detection prior to the first path (earlyTOA estimation) due to noise and interferenceTOA estimation) due to noise and interference

• Large η→ low probability of detecting the first path and a high probabilityof detecting an erroneous path (late TOA estimation) due to fading


94 P-Max

The P Max criteria is based on the selection of the earliest sample among theThe P-Max criteria is based on the selection of the earliest sample among theP largest in {zk}.

Decision vector

TOA estimation performance depends on the parameter P , where P can beoptimized according to received signal characteristicsoptimized according to received signal characteristics

- D. Dardari, C.-C. Chong, and M. Z. Win, “Analysis of threshold-based ToA estimators in UWB channels,” in European Signal Processing Conference,


EUSIPCO 2006, Florence, ITALY, Sep. 2006.

95 Backward Search

The search begins from the largest sample in {zk}, with index kmax, and theThe search begins from the largest sample in {zk}, with index kmax, and thesearch proceeds element by element backward in a window of length Wsb untilthe sample-under-test goes below the threshold η

Decision vector

- I. Guvenc and Z. Sahinoglu, “Threshold-based TOA estimation for impulse radio UWB systems,” in Proc. IEEE Int. Conf. on Utra-Wideband (ICU), Zurich Switzerland Sep 2005 pp 420–425


Zurich, Switzerland, Sep 2005, pp. 420–425.

96 Jump Back and Search Forward

The leading edge of the signal is searched element-by-element in a window oflength Wsb samples preceding the strongest one.The search proceeds forward until the sample-under-test crosses the threshold

Decision vector

The optimal selection ofWsb and η depend on the received signal characteristics


97 Performance comparison between TOA est. techniques

10-7

Signal bandwidth 1.6 GHz

10-8

Center frequency 4 GHzED integration time 2 nsPreamble length 400 pulsesF d ti 120

RM

SE

(s)

MAX

Frame duration 120 nsIEEE802.15.4a (CM4)channel model

10-9

MAX

P-Max, P=3

JBSF, Wsb=20

Simple Thresholding

SBS, W =20

Error floor Tint/√12

10 15 20 25 30 35 40SNR (dB)

10-10

SBS, Wsb=20

SBSMC, Wsb=30, D=5

From: D. Dardari, A. Conti, U. Ferner, A. Giorgetti, and M. Z. Win, “Ranging with Ultrawide Bandwidth Signals in Multipath Environments”, Proc. of IEEE (Special Issue on UWB Technology & Emerging Applications), Feb, 2009.


Multipath Environments , Proc. of IEEE (Special Issue on UWB Technology & Emerging Applications), Feb, 2009.

98 Some advanced issues

• Robust ranging (interference mitigation)• Cognitive ranging• NLOS condition• Secure ranging• Secure ranging• Passive ranging and localization• New generation RFID systemsg y• ……….


99

Interference effects

on ranging


100 UWB: the quiet neighbor

• Wide bandwidth and very low power spectral densityy p p y• “Underlay” technology coexists with other services


101 Interference effects

UWB systems are expected to work in coexistence with other UWBand narrowband systems as an underlying technology

Both wideband interference (WBI) and narrowband interference (NBI)can be present and degrade the ToA estimation

Either non-linear and linear 2D filtering techniques can be applied to mitigate the effect of the interference

Another possibility is the exploitation of the cognitive radio paradigm


102 Interference Mitigation Techniques (1/2)

True ToA

WBI+NBI


103 Interference Mitigation Techniques (2/2)

Min Filter

E h l f { } b d f llEach column of {vn,k} can be processed as follows

zk =

Nt−HXmin{vn,k, vn+1,k, . . . , vn+H−1,k},

Xn=0

{ , , + , , , + , },

where k = 0, . . . , K − 1, and H is the length of the filter.

Min Filter+Differential Filter

To improve the estimation performance in the presence of both NBI and WBI,the min filter is first applied to each matrix column to produce the intermediatevector {zk}.Th {˜ } i f h d (diff i l fil ) f llThe vector {zk} is further processed (differential filter) as follows

zk = zk − zk+1, k = 0, . . . ,K − 1 ,


with zK = 0.

104 Energy matrix filtering examples

First step: after the min filterBefore filtering

True ToATrue ToA

WBI+NBI

Second step:

WBI+NBI

after the differential filter


True ToA

105 Performance comparison between mitigation techniques

10-7

NBI & MUI

10-8

s)

10-9

RM

SE (s

)

10-9

Averaging filterDifferential filterMin filterMin & differential filters

w/o interf.

10 15 20 25 30 35 40SNR (dB)

10-10

Min & differential filters

Nsym = 400: preamble length.

Performance of the threshold-based estimator with different 2D filtering tech-niques in the presence of both NBI, INR = 35 dB, and WBI, SIR = −15 dB.

sym p g


From D. Dardari, A. Giorgetti, and M. Z. Win, “Time-of-arrival estimation in the presence of narrow and wide bandwidth interference in UWB channels,” in IEEE International Conference on Ultra-Wideband, ICUWB 2007, Singapore, Sep. 2007.

106

Cognitive ranging


107 UWB towards the Cognitive Radio paradigm

CR based on UWB represents a more complete solution as it:CR based on UWB represents a more complete solution as it:

actively looks for unused spectrum (Spectrum sensing)

begins to transmit inside those bands (Agile spectrum generation)

eventually being get out if needed when the primary users show up

r Spe

ctru

m

y g g p y p

r Spe

ctru

m

holes

frequency

Pow

e

frequency

Pow

er

Spec

trum

frequency

Pow

er

A Giorgetti M Chiani and D Dardari “Coexistence issues in cognitive radios based on ultra wide bandwidth systems ” in 1st


A. Giorgetti, M. Chiani, and D. Dardari, Coexistence issues in cognitive radios based on ultra-wide bandwidth systems, in 1st International Conference on Cognitive Radio Oriented Wireless Networks and Communications, CROWNCOM 2006, Mykonos Island, GREECE, June 2006, pp. 1–5.

108 Cognitive ranging

mTransmitting only in the spectrum holes

Pow

er S

pect

rumTransmitting only in the spectrum holes

could not be the best solution

frequency

P

We expect that a higher ranging accuracy rumWe expect that a higher ranging accuracy

can be achieved if we tolerate the presence of interference under certain power mask constraints (underlay Po

wer

Spe

ctr

power mask constraints (underlay transmission) . frequency

Target:Find the optimal transmitted signal spectrum shape which maximizes the p g p ptheoretical ranging accuracy


109 The approach followed

Input:p• According to the cognitive radio cycle, we assume the interference

spectrum is available (measured) • Constraint: Transmitted spectrum masks (e.g., FCC)

First step Determine the analytical expression of the Cramer-Rao bound (CRB) onDetermine the analytical expression of the Cramer Rao bound (CRB) on the time-of-arrival estimation MSE in the presence of interference

Determine the optimal power allocation scheme (i.e., transmitted signal spectrum shape) which minimizes the CRB subject to:

Second step

p p ) j-Transmitted signal spectrum mask constraints (e.g., FCC) - Interference spectrum


110 Signal and channel models

Transmitted signal(OFDM-like)

Received signal noise+interference

Interference


known at the receiver (sensing)

111 Cramer-Rao bound in the presence of interference

Assume

wherewhere


112 Optimal signal design

Define

Optimization problem Solution


113 Numerical example

• System parameters

Interference level

Interference avoidance and interference overlapping (all subcarriers)


Interference avoidance and interference overlapping (all subcarriers) strategies are considered and compared

114 Optimum subcarrier allocation

(all subcarriers)


(all subcarriers)

115 Numerical results


116

NLOS condition


117 NLOS condition: extra propagation delay (1/2)

The extra delay ∆τ introduced by a homogeneous material with thickness dWThe extra delay ∆τ introduced by a homogeneous material with thickness dWis given by

∆τ = (√²r − 1)

dWc

where ²r is the relative electrical permittivity of the material.

D. Dardari, A. Conti, J. Lien, and M. Z. Win, “The effect of cooperation in UWB based positioning systems using


p p g y gexperimental data,” EURASIP Journal on Advances in Signal Processing, Special Issue on Cooperative Localization inWireless Ad Hoc and Sensor Networks, 2008.

118 NLOS condition: extra propagation delay (2/2)

Extra propagation delay effect leads to biased estimations, typically some nanoseconds

Ranging error on the order of 50 cm!

Possible solutions:

• larger number of anchor nodes

• knowledge of some a priori information (e.g., scenario layout)knowledge of some a priori information (e.g., scenario layout)

• cooperation among nodes


119

The IEEE 802.15.4a standard


120 Overview of IEEE 802.15.4 (ZigBee)

Features:• Low Data Rate• Low Power Consumptionp• Low Cost• Self-Organization and Flexible Topology


121 The IEEE 802.15.4a

As an amendment to 802.15.4 for an alternative PHY to provide (2007):

• High precision ranging/location capabilityg p g g p y• Ultra low power• Lower cost• High aggregate throughputHigh aggregate throughput• Scalability to data rates• Longer range

Home automation

and.. WSNs, health care monitoring…


Asset tracking

122 Physical layers

The amendment adds two new PHYs:

• 150 - 650 MHz and 3.1 - 10.6 GHz Ultra-Wide Band (UWB)– Impulse Radio (IR) based signaling– Mandatory nominal data rate of 1 Mbps– Modulation: Combination of BPM (burst position mod.) and BPSK

• 2450 MHz Chirp Spread Spectrum (CSS)


123 UWB PHY: Channels

A compliant UWB device shall be capable of transmitting in at least f th ifi d b done of three specified bands.

• Sub-GHz Band: 150 – 650 MHz

• Low Frequency Band (LFB): 3244 - 4742 MHzLow Frequency Band (LFB): 3244 4742 MHz

• High Frequency Band (HFB): 5944 – 10234 MHz


124 Ranging in the IEEE802.15.4a standard (1/2)

• Ranging is optional (UWB PHY only)

• Main ranging protocol: Two-way TOA ranging (others possible)

• Private ranging protocol to protect against malicious devices (optional)

• Both coherent and non coherent estimation

S f f S f

IEEE 802.15.4a packet structure

Preamble

[16,64,1024,4096] symbols

Start of frame delimiter, SFD

[8,64] symbols

PHY header Data field

Preamble: acquisition channel sounding and leading edge detectionPreamble: acquisition, channel sounding and leading edge detectionSFD: frame synchronization, ranging counter management.

Designed to minimize the frame synchronization error

•The preamble and SFD lengths are specified by the application based on channel conditionsand receiver capabilities (coherent/non-coherent)

•The application bases its choice on figure of merit reports from PHY


on good range measurement is •Longer lengths are preferred for non-coherent receivers to improve SNR via processing gain

125 Ranging in the IEEE802.15.4a standard (2/2)

• Each symbol is composed of a ternary sequence taken from a set of 8 sequences

E h d i h i h id l • Each ternary sequence, and its non coherent version, has ideal periodic autocorrelation function (no side lobes) so that what is observed at the receiver is only the channel response

Autocorrelation function

Example of one the 8 possible length-31 ternary sequence

[ 000+0 0+++0+ 000+ 000+ +++00 +0 00][-000+0-0+++0+-000+-000+-+++00-+0-00]


126

Secure Ranging


127 Secure Ranging

Ranging can be subjected to hostile attacks in certain environments.

In order to make impostor and snooper attacks more difficult, the IEEE p p ,802.15.4a standard includes an optional private ranging mode

P i t iPrivate ranging

After a preliminary authentication step, nodes exchange information, via a secure communication protocol, on both the sequences to be used in the next p , qranging cycle as well as their timestamp reports

Node A Node B

The response delay τd is kept secret


The response delay τd is kept secret

128

Passive Ranging and Localization


129 Passive localization

• Besides the localization of “friendly” collaborative objects (active tags), passive geolocation, i.e., the possibility of detecting and tracking non collaborative objects (targets) is gaining an increasing attention.

• Attractive to monitoring critical environments:– Power plants– Reservoirs– Critical structures vulnerable to attacks

• Attractive for rescue in disaster scenarios (e.g., to quickly localize people trapped in collapsed buildings or in the presence of dense p p pp p g psmoke…)

targets typically human beings or vehicles


g yp y g

130 Anti-intruder radar systems

• To this aim a wireless infrastructure composed of cooperative UWB To this aim, a wireless infrastructure composed of cooperative UWB nodes could represent a cheap solution thanks to high resolution capabilities and low cost signal processing techniques

• Tactical wireless sensor networks• Radar sensor networks


• Multistatic UWB radars

131 Radar basics

• Monostatic radarTX and RX are colocated

target

• Bistatic radarBistatic radarTx and RX are separated by a distance comparable to the target

distance

target


132

• Multistatic radar

Multistatic radar basics

Txs and RXs are separated by a distance comparable to the target distance

– Multiple TXs and one RX– One TX and multiple RXs

target

• They do not suffer from coupling between TX and RX• Can detect stealth targets

RX d bl• RXs are not detectable• They do not require directional antennas to locate the target• With more RX we can increase the radar sensitivity and counteract

f di h d i d l


fading, shadowing and clutter• They require proper information exchange (cooperation is needed)

133 Some recent books and special issues

R. Verdone, D. Dardari, G. Mazzini, A. ContiWireless Sensors and Actuator Networks: Technologies, Analysis and Design, Elsevier, 2008

- Proc. of IEEE - Special Issue on UWB Technology & Emerging Applications -, Feb. 2009.Guest editors: A. Molisch, J. Zhang, M. Win, D. Dardari, W. Weisbeck

- D. Dardari, A. Conti, U. Ferner, A. Giorgetti, and M. Z. Win, “Ranging with Ultrawide Bandwidth Signals in Multipath Environments”, Proc. of IEEE (Special Issue on UWB Technology & Emerging Applications), Feb. 2009.

-EURASIP Journal on Advances in Signal Processing, Special Issue on Cooperative Localization in Wireless Ad Hoc and Sensor Networks, 2008.

Guest editors: D Dardari D Jourdan C C Chong L MucchiGuest editors: D. Dardari, D. Jourdan, C-C. Chong, L. Mucchi

- Z. Sahinoglu, S. Gezici, I. Guvenc, “Ultra-wideband positioning systems: theoretical limits, ranging algorithms, and protocols”, Cambridge University press, 2008.


g g g p g y p

134 Selected references

D Dardari A Conti U Ferner A Giorgetti and M Z Win “Ranging with Ultrawide Bandwidth Signals in Multipath D. Dardari, A. Conti, U. Ferner, A. Giorgetti, and M. Z. Win, Ranging with Ultrawide Bandwidth Signals in Multipath Environments”, Proc. of IEEE (Special Issue on UWB Technology & Emerging Applications), Feb. 2009.

R. Verdone, D. Dardari, G. Mazzini, A. Conti, Wireless Sensors and Actuator Networks: Technologies, Analysis and Design, Elsevier, 2008

D. Dardari, C.-C. Chong, and M. Z. Win, “Improved lower bounds on time-of-arrival estimation error in realistic UWB

Z. Sahinoglu, S. Gezici, I. Guvenc, “Ultra-wideband positioning systems: theoretical limits, ranging algorithms, and protocols”, Cambridge University press, 2008.

S. Gezici, Z. Tian, G. B. Giannakis, H. Kobayashi, A. F. Molisch, H. V. Poor, and Z. Sahinoglu, “Localization via ultra-wideband radios: a look at positioning aspects for future sensor networks ” IEEE Signal Processing Mag vol 22 pp 70 84

g pchannels,” in IEEE International Conference on Ultra-Wideband, ICUWB 2006, (Waltham, MA, USA), pp. 531–537, Sept. 2006.

wideband radios: a look at positioning aspects for future sensor networks, IEEE Signal Processing Mag., vol. 22, pp. 70–84, Jul. 2005.

D. Dardari, C.-C. Chong, and M. Z. Win, “Threshold-based time-of-arrival estimators in UWB dense multipath channels,”IEEE Trans. Commun., vol. 56, no. 8, Aug. 2008.

I. Guvenc and Z. Sahinoglu, “Threshold-based TOA estimation for impulse radio UWB systems,” in Proc. IEEE Int. Conf. on Utra-Wideband (ICU), Zurich, Switzerland, Sep 2005, pp. 420–425.

D. Dardari, A. Giorgetti, and M. Z. Win, “Time-of-arrival estimation in the presence of narrow and wide bandwidth interference in UWB channels,” in IEEE International Conference on Ultra-Wideband, ICUWB 2007, Singapore, Sep. 2007.

Z. Shainoglu and I. Guvenc, “Multiuser interference mitigation in noncoherent UWB ranging via nonlinear filtering,”EURASIP J Wi l C i i d N ki l 2006 1 10 2006


EURASIP J. Wireless Communications and Networking, vol. 2006, pp. 1–10, 2006.


C. Falsi, D. Dardari, L. Mucchi, and M. Z. Win, “Time of arrival estimation for UWB localizers in realistic environments,”

D. Dardari, A. Conti, J. Lien, and M. Z. Win, “The effect of cooperation in UWB based positioning systems usingexperimental data,” EURASIP Journal on Advances in Signal Processing, Special Issue on Cooperative Localization in

C. Falsi, D. Dardari, L. Mucchi, and M. Z. Win, Time of arrival estimation for UWB localizers in realistic environments,EURASIP J. Appl. Signal Processing (Special Issue on Wireless Location Technologies and Applications), 2006.

g gWireless Ad Hoc and Sensor Networks, 2008.

P. Cheong, A. Rabbachin, J. Montillet, K. Yu, and I. Oppermann, “Synchronization, TOA and position estimation forlow-complexity LDR UWB devices,” in Proc. of IEEE Int. Conf. on Ultra-Wideband (ICUWB), Zurich, SWITZERLAND,Sep 2005, pp. 480–484.

J.-Y. Lee and R. A. Scholtz, “Ranging in a dense multipath environment using an UWB radio link,” IEEE J. Sel. AreasCommun., vol. 20, no. 9, pp. 1677–1683, Dec. 2002.

Y. Shen and M. Z. Win,“Effect of path-overlap on localization accuracy in dense multipath environments,” in Proc. IEEE Int. Conf. on Commun., Beijing, CHINA, May 2008.

Y. Shen and M. Z. Win, “Fundamental limits of wideband localization accuracy via Fisher information,” in Proc. IEEE

B. Alavi and K. Pahlavan, “Modeling of the TOA-based distance measurement error using UWB indoor radiomeasurements,” IEEE Commun. Lett., vol. 10, no. 4, pp. 275–277, Apr. 2006.

Y. Shen and M. Z. Win, Fundamental limits of wideband localization accuracy via Fisher information, in Proc. IEEEWireless Commun. and Networking Conf., Kowloon, HONG KONG, Mar. 2007, pp. 3046–3051.

D. Dardari and M. Z. Win, “Ziv-Zakai bound of time-of-arrival estimation with statistical channel knowledge at the receiver,” in IEEE International Conference on Ultra-Wideband, ICUWB 2009, Vancouver, BC, Canada, Sep. 2009, pp. 624


receiver, in IEEE International Conference on Ultra Wideband, ICUWB 2009, Vancouver, BC, Canada, Sep. 2009, pp. 624 — 629


M Z Win and R A Scholtz “Impulse radio: How it works ” IEEE Commun Lett vol 2 no 2 pp 36 38 Feb 1998M. Z. Win and R. A. Scholtz, Impulse radio: How it works, IEEE Commun. Lett., vol. 2, no. 2, pp. 36–38, Feb. 1998.

Moe Z. Win, and Robert A. Scholtz, “Characterization of Ultra-Wide Bandwidth Wireless Indoor Channels: A Communication-Theoretic View”, IEEE Journal On Selected Areas In Communications, Vol. 20, No. 9, December 2002

C.-C. Chong and S. K. Yong, “A generic statistical-based UWB channel model for high-rise apartments,” IEEE Trans.Antennas Propag., vol. 53, pp. 2389–2399, 2005.

D. Cassioli, M. Z. Win, and A. F. Molisch, “The ultra -wide bandwidth indoor channel: from statistical model tosimulations,” IEEE J. Sel. Areas Commun., vol. 20, no. 6, pp. 1247–1257, Aug. 2002.

A. F. Molisch, D. Cassioli, C.-C. Chong, S. Emami, A. Fort, B. Kannan, J. Karedal, J. Kunisch, H. Schantz, K. Siwiak,and M. Z. Win, “A comprehensive standardized model for ultrawideband propagation channels,” IEEE Trans. AntennasPropag., vol. 54, no. 11, pp. 3151–3166, Nov. 2006, Special Issue on Wireless Communications.

“IEEE standard for information technology - telecommunications and information exchange between systems - local andgy g ymetropolitan area networks - specific requirement part 15.4: Wireless medium access control (MAC) and physical layer(PHY) specifications for low-rate wireless personal area networks (WPANs),” IEEE Std 802.15.4a-2007 (Amendment toIEEE Std 802.15.4-2006), pp. 1–203, 2007.

G M B Fid B D O A d “Wi l t k l li ti t h i ” C t N t k I 51 G. Mao, B. Fidan, B.D.O. Anderson, Wireless sensor network localization techniques, Computer Networks, Issue 51, 2007, Elseveir, pp. 2529-2553



M Maroti P Völgyesi S Dora B Kusy A Nadas A Ledeczi G Balogh and K Molnar “Radio interferometric M. Maroti, P. Völgyesi, S. Dora, B. Kusy, A. Nadas, A. Ledeczi, G. Balogh, and K. Molnar, Radio interferometric geolocation,” in Proc. ACM Conference on Embedded Networked Sensor Systems, San Diego, USA, Nov. 2005, pp. 1–12.

B. Zhen, H.-B. Li, and R. Kohno, “Clock management in ultra-wideband ranging,” Proc. IST Mobile and WirelessCommun. Summit, pp. 1–5, Jul. 2007.

F. Sivrikaya and B. Yener, “Time synchronization in sensor networks: a survey,” IEEE Netw., vol. 18, no. 4, pp. 45–50,2004.

Y. Jiang and V. Leung, “An asymmetric double sided two-way ranging for crystal offset,” Int. Symp. on Signals, Systemsand Electronics (ISSSE), pp. 525–528, July 30 2007-Aug. 2 2007.

I. Guvenc, C.-C. Chong, and F. Watanabe, “NLOS identification and mitigation for UWB localization systems,” in Proc.g g yIEEE Wireless Commun. and Networking Conf., Kowloon, HONG KONG, Mar. 2007, pp. 1571–1576.

C.-C. Chong, F. Watanabe, and M. Z. Win, “Effect of bandwidth on UWB ranging error,” in Proc. IEEE Wireless Commun and Networking Conf Kowloon HONG KONG Mar 2007 pp 1559–1564Commun. and Networking Conf., Kowloon, HONG KONG, Mar. 2007, pp. 1559–1564.

W. Suwansantisuk and M. Z. Win, “Multipath aided rapid acquisition: Optimal search strategies,” IEEE Trans. Inf. Theory, vol. 53, no. 1, pp. 174–193, Jan. 2007.

H. L. Van Trees, Detection, Estimation, and Modulation Theory, 1st ed. New York, NY 10158-0012: John Wiley &Sons, Inc., 1968.

D Ch M Z k i d J Zi “I d l b d i l t ti ti ” IEEE T I f Th


D. Chazan, M. Zakai, and J. Ziv, “Improved lower bounds on signal parameter estimation,” IEEE Trans. Inf. Theory,vol. 21, no. 1, pp. 90–93, Jan. 1975.


J. Zhang, R. A. Kennedy, and T. D. Abhayapala, “Cramer-rao lower bounds for the synchronization of UWB signals,” in g, y, y p , y g ,EURASIP Journal on Wireless Communications and Networking, vol. 3, 2005.

V. Lottici, A. D’Andrea, and U. Mengali, “Channel estimation for ultra-wideband communications,” IEEE J. Sel. AreasCommun., vol. 20, no. 9, pp. 1638–1645, Dec. 2002.C , 0, 9, pp 638 6 5, 00

H. Zhan, J. Ayadi, J. Farserotu, and J.-Y. Le Boudec, “High-resolution impulse radio ultra wideband ranging,” Proc. ofIEEE Int. Conf. on Ultra-Wideband (ICUWB), pp. 568–573, Sep. 2007.

S. H. Song and Q. T. Zhang, “Multi-dimensional detector for UWB ranging systems in dense multipath environments,”IEEE Trans. Wireless Commun., vol. 7, no. 1, pp. 175–183, 2008.

A R bb hi I O d B D i “ML i f i l i i b d l l i UWB A. Rabbachin, I. Oppermann, and B. Denis, “ML time-of-arrival estimation based on low complexity UWB energydetection,” in Proc. of IEEE Int. Conf. on Ultra-Wideband (ICUWB), Waltham, MA, Sep. 2006, pp. 599–604.

A. A. D’Amico, U. Mengali, and L. Taponecco, “Energy-based TOA estimation,” IEEE Trans. Wireless Commun., vol. 7,no 3 pp 838 847 March 2008no. 3, pp. 838–847, March 2008.

Z. Lei, F. Chin, and Y.-S. Kwok, “UWB ranging with energy detectors using ternary preamble sequences,” in Proc. IEEEWireless Commun. and Networking Conf., vol. 2, Las Vegas, NE, USA, Apr 2006, pp. 872–877.

Z. Tian and G. B. Giannakis, “A GLRT approach to data-aided timing acquisition in UWB radios-Part I: Algorithms,”IEEE Trans. Wireless Commun., vol. 4, no. 6, pp. 1536–1576, Nov. 2005.



D. Dardari, Y. Karisan, S. Gezici, A. A. D’Amico, and U. Mengali, “Performance limits on ranging with cognitive radio,”( )

L. Yang and G. B. Giannakis, “Timing ultra-wideband signals with dirty templates,” IEEE Trans. Commun., vol. 53, no. 11, pp. 1952–1963, Nov. 2005.

in International Workshop on Synergies in Communicaions and Localization (SyCoLo 2009), Dresden, Germany, June 2009.

11, pp. 1952 1963, Nov. 2005.

C. Gentile and A. Kik, “A comprehensive evaluation of indoor ranging using ultra-wideband technology,” EURASIP J.Wireless Commun. and Networking, vol. 2007, no. 1, pp. 12–12, 2007.

C. Mazzucco, U. Spagnolini, and G. Mulas, “A ranging technique for UWB indoor channel based on power delay profileanalysis,” in Proc. IEEE Semiannual Veh. Technol. Conf., vol. 5, Milan, ITALY, May 2004, pp. 2595–2599.

D D d i “P d d ti UWB fl t f t i ” IEEE C L tt l 8 10 608D. Dardari, “Pseudo-random active UWB reflectors for accurate ranging,” IEEE Commun. Lett., vol. 8, no. 10, pp. 608–610, Oct 2004.

D. B. Jourdan, D. Dardari, and M. Z. Win, “Position error bound for UWB localization in dense cluttered environments,”IEEE Trans. Aerosp. Electron. Syst., vol. 44, no. 2, pp. 613–628, Apr. 2008.IEEE Trans. Aerosp. Electron. Syst., vol. 44, no. 2, pp. 613 628, Apr. 2008.

E. Paolini, A. Giorgetti, M. Chiani, R. Minutolo and M. Montanari, “Localization Capability of Cooperative Anti-intruder Radar Systems,” EURASIP Journal on Advances in Signal Processing, (Special Issue on Cooperative Localization in Wireless

Ad Hoc and Sensor Networks), vol. 2008, Article ID 726854, 14 pages, 2008

S. Severi, D. Dardari, G. Destino, and G. Abreu, “Efficient and accurate localization in multihop networks,” in Proc. of the Asilomar Conference on Signals, Systems, and Computers, Pacific Grove, CA, Nov. 2009, pp. 1—6.


S. Severi, G. Abreu, and D. Dardari, “A quantitative comparison on multihop algorithms,” in Proc. Workshop on Positioning, Navigation and Communication (WPNC 10), Dresden, Germany, Mar. 2010, pp. 1—6.


G. Mao, B. Fidan, B.D.O. Anderson, “Wireless sensor network localization techniques,” Computer Networks, Issue 51, 2007,G. Mao, B. Fidan, B.D.O. Anderson, Wireless sensor network localization techniques, Computer Networks, Issue 51, 2007,Elseveir, pp. 2529-2553

H. Wymeersch, J.Lien, M.Win, “Cooperative Localization in Wireless Networks“, Proc. of the IEEE – S.I. on UWB Technology& Emerging Applications, Feb. 2009

D. B. Jourdan, D. Dardari, and M. Z. Win, “Position error bound for UWB localization in dense cluttered environments,” inIEEE Trans. on Aerospace and Electronic Systems, vol. 44, issue 2, April 2008, pp. 613-628.

M. Gavish, A.J .Weiss, “Performance analysis of bearing-only target location algorithms,” IEEE Trans. on Aerospace and El.Sys., Vol. 28, Issue 3, 1992, pp. 817-828

A. Savvides, H. Park, M. Srivastava, The bits and flops of the n-hop multilateration primitive for node localization problems,First ACM Workshop on Wireless Sensor Networks and Application (WSNA), pp112-121, Atlanta GA, September 2002.

D. Niculescu, B. Nath, Ad hoc positioning system (APS), IEEE Proc. of Globecom 2001, pp. 2926-2931.

C. Savarese, J.M. Rabaey, J. Beutel, Locationing in distributed ad-hoc wireless sensor networks, IEEE Proc. of ICASSP2001

L. Doherty, K.S.J. Pister, L. El Ghaoui, “Convex Position Estimation in Wireless Sensor Networks,” IEEE Proc. of INFOCOM2001

N. Patwari, J. N. Ash, S. Kyperountas, A. O. Hero, III, R. L. Moses, and N. S. Correal, “Locating the nodes: cooperativelocalization in wireless sensor networks,” IEEE Signal Processing Magazine, vol. 22, no. 4, 54 pages, 2005.



J J Caffery Jr “A new approach to the geometry of TOA location ” in Proceedings of the 52nd IEEE Vehicular J. J. Caffery, Jr., A new approach to the geometry of TOA location, in Proceedings of the 52nd IEEE Vehicular Technology Conference (VTC ’00), vol. 4, p. 1943, Boston, Mass, USA, September 2000.

D. Dardari and A. Conti, “A sub-optimal hierarchical maximum likelihood algorithm for collaborative localization in ad-hocnetwork,” in Proceedings of the 1st Annual IEEE Communications Society Conference on Sensor and Ad HocComm nications and Net orks (SECON ’04) p 425 Santa Clara Calif USA October 2004Communications and Networks (SECON 04), p. 425, Santa Clara, Calif, USA, October 2004.

T. Pavani, G. Costa, M. Mazzotti, D. Dardari, and A. Conti, “Experimental results on indoor localization technique throughwireless sensors network,” in Proc. IEEE Vehicular Tech. Conf. (VTC 2006-Spring), (Melbourne, AUSTRALIA), May 2006.

D.Dardari, A.Conti, J.Lien, M.Win, “The Effect of Cooperation on Localization Systems Using UWB Experimental Data“,EURASIP Journal on Advances in Signal Processing - S.I. on Cooperative Localization in Wireless Ad Hoc and SensorNetworks, 2008

I. Guvenc, C.-C. Chong, and F. Watanabe, “Analysis of a linear least-squares localization technique in LOS and NLOSI. Guvenc, C. C. Chong, and F. Watanabe, Analysis of a linear least squares localization technique in LOS and NLOSenvironments,” in Proceedings of the 65th IEEE Vehicular Technology Conference (VTC ’07), p. 1886, Dublin, Ireland, April2007

A.Conti, D.Dardari, M.Win, “Experimental Results on Cooperative UWB Based Positioning Systems”, Proc. of ICUWB 2008

A.Conti, D.Dardari, L.Zuari, “Cooperative UWB Based Positioning Systems: CDAP Algorithm and Experimental”, Proc. ofIEEE ISSSTA 2008

M. R. Gholamiy, E. G. Strom, F. Sottile, D. Dardari, S. Gezici, M. Rydstrom, M. A. Spirito, and A. Conti, “Static y, , , , , y , p , ,positioning using UWB range measurements,” in IEEE International Symposium on Wireless Pervasive Computing (ISWPC 2010), Modena, Italy, May 2010,

N. Decarli, D. Dardari, S. Gezici, and A. D’Amico, “LOS/NLOS detection for UWB signals: a comparative study using i t l d t ” i IEEE I t ti l S i Wi l P i C ti (ISWPC 2010) M d It l M


experimental data,” in IEEE International Symposium on Wireless Pervasive Computing (ISWPC 2010), Modena, Italy, May 2010,

142

For further information please contactFor further information please contact

Davide Dardari, [email protected]

Special thanksG Ab (CWC Fi l d)G. Abreu (CWC, Finland)O. Andrisano (University of Bologna, Italy)M. Chiani (University of Bologna, Italy)A. Conti (University of Ferrara, Italy)A. Conti (University of Ferrara, Italy)C-C. Chong (DoCoMo Labs, USA)N. Decarli (University of Bologna, Italy)U. Ferner (MIT, USA)S G i i (Bilk U i i T k )S. Gezici (Bilkent University, Turkey)W. Gifford (MIT, USA)A. Giorgetti (University of Bologna, Italy)J. Lien (JPL, Nasa, USA)( , , )D. Jourdan (Athera, Washington, USA)L. Mucchi (University of Florence, Italy)S. Severi (University of Bologna, Italy)R V d (U i it f B l It l )


R. Verdone (University of Bologna, Italy)M. Z. Win (MIT, USA)………………………

Documents

Short-Range Positioning Systems: Fundamentals and Advanced … Positioning... · The purpose of any localization algorithm is, given a set of measurements, to find the locations of