Call Us ≽ 8377877756 ≼ Call Girls In Shastri Nagar (Delhi)
Adaptive Channel Prediction, Beamforming and Scheduling Design for 5G V2I Network
1. Adaptive Channel Prediction, Beamforming and
Scheduling Design for 5G V2I Network
T. E. Bogale+
, X. Wang++
and L. B. Le+
Institute National de la Recherche Scientifique (INRS), Canada+
Western University, Canada++
Sep. 25, 2017
2. Presentation outline
Presentation outline
1 System scenario and objective
2 Multipath channel model
3 Proposed CIR prediction, beamforming and scheduling design
Adaptive RLS prediction algorithm
Joint beamforming and VE scheduling
4 Simulation results
5 Conclusions
(VTC-Fall 2017, Toronto Canada) Adaptive Design for 5G V2I Sep. 25, 2017 2 / 10
3. System scenario and objective
System scenario and objective
System scenario
1
Objective
CIR Prediction
Scheduling
Beamforming
(VTC-Fall 2017, Toronto Canada) Adaptive Design for 5G V2I Sep. 25, 2017 3 / 10
4. Multipath channel model
Multipath channel model
Multipath CIR model
Multiuser: Single antenna VEs and N antenna BS
Maximum delay spread Td and sampling period Ts
No of multipath CIR taps L = Td
Ts
¯hkn =[¯hkn1, ¯hkn2, · · · , ¯hknL]T
⇒ ¯Hk =
√
Pk
˜Hk
√
Rk
Assumption:
√
Pk [j],
√
Rk [j] are constant for Cb blocks,
¯hknl: correlated over time, Pk = diag(g2
k1, g2
k1, · · · , g2
kL)
˜˜Hk = P−1
k
¯Hk Uk D−1
k (Low dimension)
svd(Rk ) Uk Dk UH
k
Goal: Predict ˜˜Hk [i + 1] from ˜˜Hk [i], · · · , ˜˜Hk [i − Sb + 1]
(VTC-Fall 2017, Toronto Canada) Adaptive Design for 5G V2I Sep. 25, 2017 4 / 10
6. Proposed CIR prediction, beamforming and scheduling design Joint beamforming and VE scheduling
Joint beamforming and VE scheduling
Scheduling and beamforming
Schedule VEs to optimize
max
Ks,Ks,wks
Ks
k=1
log(1 + γp
ks), Pks ≤ Pmax , k ∈ Ks
γks =
Pks|wH
kshks|2
K
i=k Pis|wH
kshis|2 + σ2|wks|2
hkns =fH
s
¯hkn
with γp
ks(wks) is kth VE sth sub-carrier SINR (receive
beamforming vector), fH
s = [1, e−j 2π
M
s
, e−j 2π
M
2s
, · · · , e−j 2π
M
(L−1)s
],
and M is FFT size.
Mixed integer and non-linear problem (Non convex)
Apply greedy VE selection approach
(VTC-Fall 2017, Toronto Canada) Adaptive Design for 5G V2I Sep. 25, 2017 6 / 10
7. Proposed CIR prediction, beamforming and scheduling design Joint beamforming and VE scheduling
Proposed design summary
Summary of proposed design
1
i-Sb-1 i-1 i i+1
CIR Blocks
Predict
i-Sb-1 i-1 i i+1
i-Sb-1 i-1 i i+1
i-Sb-1 i-1 i i+1
i-Sb-1 i-1 i i+1
VE 1
VE 2
VE 3
VE 4
VE K
Schedule &
beamform
VE 1
VE 3
VE Ks
(VTC-Fall 2017, Toronto Canada) Adaptive Design for 5G V2I Sep. 25, 2017 7 / 10
8. Simulation results
Simulation results
Simulation parameters
BS antenna: N = 8
Number of multipath L=4
Gain: [0, −6.3, −25.1 − 22.7]dB
(VTV-Expressway)
Fc = 5.6 GHz, FFT size M = 64, Cb = Sb = 8
Spatial correlation
Rk =
Ck
i=1
a(θki )a(θki )H
Temporal correlation
E{( ˜Hk [i])m,n( ˜Hk [j])H
m,n} = αkij = J0(2πfk |i − j|)
J0(.): 0th-order Bessel, fk : Normalized Doppler
Each VE speed: 90 km/hr
MSEs with different CIR predictors
0 50 100 150 200 250 300 350 400
Block index (i)
3
3.5
4
4.5
5
5.5
Meansquareerror(MSE)
MMSE
Proposed RLS
Outdated CIR
(VTC-Fall 2017, Toronto Canada) Adaptive Design for 5G V2I Sep. 25, 2017 8 / 10
9. Simulation results
Simulation results cont’d
Single user average SE
0 2.5 5 7.5 10 12.5 15 17.5 20
SNR (dB)
2
4
6
8
10
12
14
AverageSE(bps/Hz)
Perfect CIR
MMSE
Proposed RLS
Outdated CIR
Random scheduling
Effect of scheduled VEs Kmax
1 2 3 4 5 6 7 8
Maximum number of scheduled VEs (K max
)
0
5
10
15
20
25
30
35
40
AveragesumSE(bps/Hz)
Perfect CIR
MMSE
Proposed RLS
Outdated CIR
Random scheduling
Main observations
Gap between MMSE and proposed is small
(proposed RLS predictor does not assume CIR model)
⇒ Advantageous
Random scheduling with predicted CIR performs bad
⇒ VE selection is crucial (for single user case)
(VTC-Fall 2017, Toronto Canada) Adaptive Design for 5G V2I Sep. 25, 2017 9 / 10
10. Conclusions
Conclusions
We propose combined CIR prediction, scheduling and
beamforming design
Proposed design applies a RLS algorithm for CIR predictor
The VE scheduling is performed to maximize the average sum
rate with beamforming
We apply greedy based VE scheduling design with ZF
beamforming technique
Proposed technique achieves better performance than other
approaches (especially for single user scheduling case)
For multiuser scheduling case, proposed approach experiences
performance degradation as Kmax increases
Potential reason: CIR prediction error accumulates with Kmax
Possible remedy: Increase SNR as it reduces prediction error
(VTC-Fall 2017, Toronto Canada) Adaptive Design for 5G V2I Sep. 25, 2017 10 / 10