Upload
others
View
5
Download
0
Embed Size (px)
Citation preview
New Technologies of B3G/4G
Xiaohui LiState Key Laboratory of ISN
Xidian Univeristy, Xi’an, Shaanxi, [email protected]
31 July, 2009
Introduction of Our GroupResearch WorkProjectsAchievementsFamily AlbumFuture work
Introduction of Our GroupResearch WorkProjectsAchievementsFamily AlbumFuture work
Introduction of Our GroupCommunity of associated professor, Ph.D. and M.S. students.Research Interests mainly
Theory: Some key technologies of MIMO systemsApplication: Development of 3G Long Term Evolution(LTE)
Emphasize the integration of theory with practice, focus on science exchanges home and abroad to broad our scope
Introduction of Our GroupResearch WorkProjectsAchievementsFamily AlbumFuture work
Research works
Key technologies of MIMO systemsMulti-user schedulingMIMO detection
Development of 3G Long Term Evolution (LTE) Link level simulation and system level simulation with/without CoMP (Coordinated Multiple Point) Inter-cell Interference Co-ordination (ICIC) methodInter-cell power control technologyCell selection strategy for users
PSO Multi-user schedulingParticle Swarm Optimization
Kennedy and Eberhart, in 1995 , Proc. Int'l Conf. Neural Networks, vol. 5, Motivated by the behaviors of bird flock or bees finding food
Advantage Simple, easy realizationlow computation complexity
Evolutionary mechanism:
( ) ( ) ( ) ( )( )( ) ( )( )
1 1
2 2
1 1 1
1 1 i i i i
i
v t wv t c pbest t x t
c gbest t x t
η
η
= − + − − −
+ − − −
( ) ( ) ( )1i i ix t x t v t= − +
PSO Multi-user schedulingOur contribution:
Particle definition: candidate of the scheduled user subset
Fitness Function definition: capacity value corresponding to such particle
Key PSO scheduling method steps① Initialization. Generate certain number of particles and velocities② Evaluation. Calculate the fitness value of each particle③ Evolution. Apply PSO mechanism to these particles④ Termination. Repeat the above steps until the maximal iteration number
ParticleΔ
1Δ KΔ1, MS selected
0,otherwisek
kth⎧Δ = ⎨
⎩
( ) 2 det( ( ( )))Hi NtF log diag iτ γ= +I H H Δ
PSO Multi-user scheduling Simulation Results
0 2 4 6 8 10 12 14 16 18 200
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Capacity(bit/s/Hz)
Out
age
Pro
babi
lity
Capacity CDF curves of different algorithms
Optimal ESPSONBS[3]Random
0 2 4 6 8 10 12 14 16 18 202
4
6
8
10
12
14
16
SNR[dB]
Cap
acity
[bit/
s/H
z]
The effect of iterations on capacity performance
12 13 14
10
10.5
11
11.5
12
iteration=1iteration=5iteration=10iteration=15optimal
.
Parameters: one BS, 16 MSs with single antenna, L=4MSs scheduled
PSO: Inertia weight factor w=0.9, acceleration constant c1=c2=2.0, particle number = 20 , maximal iteration number= 15
MIMO detecting
1 2[ , , ]s s
1STBC1s
2s
Ks
2STBC
KSTBC
1X 1N
2X 2N
KX KN Nr
1
= +Y HX Z2
System model
The received signal over T time slots:
OSTBC codeword matrix of user k :
real constant coefficient matrices of the given OSTBC.
* *11 1 11 1 1 1 1 1
* *1 1
=T T
K K K K KT K KT K
= +
⎛ ⎞+ +⎜ ⎟
+⎜ ⎟⎜ ⎟+ +⎝ ⎠
Y HX Z
A s B s A s B sH Z
A s B s A s B s
KM O M
L
[ ] * *
1 1 1, , , ,k k kT k k k k kT k kT k= = + +⎡ ⎤⎣ ⎦X x x A s B s A s B sL L
, , , ,t t t = 1 TA B L
MIMO detecting
Two Existing Algorithms, Lin Dai, (IEEE Trans. Comm, vol. 53,no.9Type1 detecting:
Type 2 detecting
GroupSeperation
With ZF
Projection
1G Y
2G Y
KG Y
= +Y HX Z
1 1 1 1= +y H X Z
2 2 2 2= +y H X Z
K K K K= +y H X Z
Recombination1 1 1 1
ˆ ˆˆ = +r H s Z
2 2 2 2ˆ ˆˆ = +r H s Z
ˆ ˆˆK K K K= +r H s Z
1 1 1 1λ= +r I s Z 1 1HH r
2 2 2 2λ= +r I s Z
K K K Kλ= +r I s Z
2 2ˆHH r
ˆHK KH r
Linear decoding
1s
2s
ˆKs
Linear decoding
Linear decoding
Recombination
Recombination
= +Y HX Z
1W r
2W r
KW r
1 1 1 1= +r Ω s N
2 2 2 2= +r Ω s N
STBC ML decoding 1s
2s
Ks
1
K
k kk=
= +∑r Ω s N
K K K K= +r Ω s N
STBC ML decoding
STBC ML decoding
MIMO detectingProposed approach
Advantages :Unitary projection matrix without suffering from noise enhancement Low Linear decoding complexity compared with type 2 ML decoding
= +Y HX Z
1W r
2W r
KW r
1 1 1 1= +r Ω s N
2 2 2 2= +r Ω s N
1 1 1 1ξ= +r I s N 1 1HΩ r
2 2 2 2ξ= +r I s N
K K K Kξ= +r I s N
1 1HΩ r
HK KΩ r
Linear decoding
Linear decoding
Linear decoding
1s
2s
Ks
1
K
k kk=
= +∑r Ω s N
K K K K= +r Ω s N
Recombination
MIMO detectingSystem design under the case of transmitter employing different OSTBC
Assume K different STBCk ,with code length TkOutput of STBCk:The common code length : minimum common multipleDefine
Transmit signal matrix design:
represents the output code matrix where imag coefficient matrices
represents the output code matrix where real coefficient matrices
* *,1 ,1 , ,[ , , ]k k k k k k Tk k k Tk k= + +X A s B s A s B sL
/ , 1kGk T T k K= = L
[ ]
1,1 1, 1 1,1 1, 11 1
2,1 2, 2 2,1 2, 22 2
,1 , ,1 ,
( ) ( ) ( ) ( )( ) ( )( ) ( ) ( ) ( )( ) ( )
( ) ( )
( ) ( ) ( ) ( )( ) ( )
G G
G G
K K GK K K GKK K
⎡ ⎤⎡ ⎤⎢ ⎥⎢ ⎥⎢ ⎥⎢ ⎥= = =⎢ ⎥⎢ ⎥⎢ ⎥⎢ ⎥⎢ ⎥⎣ ⎦ ⎣ ⎦
X A X A X B X BX A X BX A X A X B X BX A X B
X X A X B
X A X A X B X BX A X B
L LL L
M M M M M MM ML L
k =B 0, ( )k jX A
, ( )k jX B
k =A 0
1 2, , KT MCM T T T= L
MIMO detectingThe received signal over T time slots:
With the proposed detecting approachFor the kth group, receive vector:
For each subgroup:The desired signal of each subgroup can be decoded by:
Note that the equivalent channel of any group has the orthogonal property
1 1 2 2 K K= + = + + + +Y HX Z H X H X H X ZL
k k k k k= = + r W r Ω s n)) )
,1 , ,1 , ,1 ,[ ]k k Gk k k k Gk k k Gk⎡ ⎤ ⎡ ⎤= +⎣ ⎦ ⎣ ⎦ r r Ω s s n n)) ) ) )L L L
, 1ki k ki ki i Gk= + = r Ω s n)) ) L
Hki k ki k ki kiξ= = +r Ω r s n
) )% %g
Hk k kξ= Ω Ω I
) )g
MIMO detecting Simulation Results
Parameters(Fig.1):Two groups, Alamouti coding, QPSK, Parameters(Fig.2): Two groups, Alamouti, 3*8 OSTBC,QPSK
0 2 4 6 8 10 12 14 16 18 2010-5
10-4
10-3
10-2
10-1
100
SNR[dB]
BE
R
BER Plots
type1proposedtype2
0 2 4 6 8 10 12 14 16 18 2010-5
10-4
10-3
10-2
10-1
100
101
SNR[dB]
BE
R
BER Plots
type1 avetype1 user1type1 user2proposed aveproposed user1proposed user2
Platform of LTE/LTE-A
Link level simulationSystem level simulation
System level simulation with CoMP
For UL-CoMP, a specific UE is processed not only by its serving cell, but also by its neighbor cell ,as its coordinate cell.
Inter-cell Interference Co-ordination (ICIC)
The whole band is equally divided into eight sub-bands B=f1+f2+...f8
f1 f2 f8f6f5f3 f4 f7
Sector A: edge user :f1,f4,f7 centre user:f2,f3,f5,f6,f8
A
B
C
3 sectors per cell-site
Sector B: edge user :f2, f8 centre user:f1,f3,f4,f5,f6,f7
Sector C: edge user :f3, f5,f6 centre user:f1,f2,f4,f7,f8
Inter-cell Interference Co-ordination (ICIC)
0 0.2 0.4 0.6 0.8 1 1.2 1.40
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
bps/Hz
CD
F
ICICNo ICIC
Inter-cell Power control
Power control in CoMP: to achieve maximum throughput The proposed power reallocation scheme
(1) Power requisition: UEs get power value according to LTE Rel.8 uplink power control formula
max 10 0_min ,10 log ( )MPUSCH MCS iP P P PL fα= + + + Δ + Δg g
Inter-cell Power control(2) Power reallocation: each UE transmit power is calculated by
This scheme does not increase the total power of system
11 22 33' ' '11 22 33 11 22 33
p p p pp : p : p (g ) : (g ) : (g )
+ + =⎧⎨ =⎩
' 1111
11 22 33
' 2222
11 22 33
' 3333
11 22 33
gp pg g g
gp pg g g
gp pg g g
⎧=⎪ + +⎪
⎪=⎨ + +⎪
⎪=⎪
+ +⎩
Inter-cell Power control
1 2 3 4 5 6 7 8 9 10
x 105
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Throughout [bit/s]
CD
F
Power control schemes compaire with different threshld
Threshold = 12 fractional power controlThreshold = 12 power reallocateThreshold = 15 fractional power controlThreshold = 15 power reallocateThreshold = 18 fractional power controlThreshold = 18 power reallocate
Cell Selection algorithmCell-selection with CoMP in Uplink
Exploit inter-cell diversity,Increase cellular capacity, Balance the traffic densities.
ARSRPCRSRP
BRSRP
Cell selection
Cell Selection algorithm
The proposed schemeDetermine the serving eNB
Determine the specified UE type: cell-edge UE or center UEDecide the serving eNB for the specified UE according to the RSRP of each eNB
Determine the CoMP setWhen serving eNB is the originally accessed eNB, determine the number of coordinated cell in active CoMP set.When serving eNB is not the originally accessed eNB, decide whether the CoMP is needed or not and the number of coordinated cells in active CoMP set.
Cell Selection algorithm
0 0.2 0.4 0.6 0.8 1 1.2 1.40
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
bps /Hz
CD
F
Non-Cell-Selection&&Non-CoMPCell-Selection&&Non-CoMPCell Selection&&CoMP
Introduction of Our GroupResearch WorkProjectsAchievementsFamily AlbumFuture work
Projects within 3 yearsKey technologies of adaptive multi-user MIMO systems,National Science Foundation of China,2008-2010Solutions on improving the throughput of edge cells for IMT-Advanced systems,Specialized project for state key laboratory,2008-2009Key technologies on MIMO for the uplink LTE systems,Project of ZTE corporation,2007-2008Enhanced technologies on MIMO for uplink LTE systems,Project of ZTE corporation,2008-2009
Introduction of Our GroupResearch WorkProjectsAchievementsFamily AlbumFuture work
Achievements of our group
Achievements:About 30 papers2 books1 translation4 patents4 proposal of B3G/4G
Introduction of Our GroupResearch WorkProjectsAchievementsFamily AlbumFuture work
Family Album
Introduction of Our GroupResearch WorkProjectsAchievementsFamily AlbumFuture work
Future workMulti-user MIMO systems
Multi-user PrecodingMulti-user scheduling combing with precoding
Cross-layer designCombing MIMO-OFDM With MAC protocol
Cognitive radioCognitive coexistences between WLAN and LTE systems