Page 1
Interactive Communication for ResourceAllocation
Jie [email protected]
John MacLaren [email protected]
Adaptive Signal Processing and Information Theory GroupDepartment of Electrical and Computer Engineering
Drexel University, Philadelphia, PA 19104
This research has been supported by the Air Force Research Laboratoryunder agreement number FA9550-12-1-0086.
March 19th, 2014
Jie Ren (Drexel ASPITRG) ICFRA March 19th, 2014 1 / 20
Page 2
Introduction
Outline
1 Introduction
2 Problem Model
3 Analysis
4 Results
5 Conclusions
Jie Ren (Drexel ASPITRG) ICFRA March 19th, 2014 2 / 20
Page 3
Introduction
Motivation
M. I. Salman etc. ”IETE Technical Review”
• Which user to assign thesubcarrier to
• Which modulation and codingscheme to employ
X1
X2
X3
Jie Ren (Drexel ASPITRG) ICFRA March 19th, 2014 3 / 20
Page 4
Introduction
Adaptive Modulation and Coding
• Overheads• Reference Signals• Channel Quality Indicators• Control Decisions
• Occupy the OFDMA resourceblocks
• Approximately 1/4 to 1/3 of alldownlink transmission in LTE
Jie Ren (Drexel ASPITRG) ICFRA March 19th, 2014 4 / 20
Page 5
Introduction
BackgroundRateless Codes
http://www.telematica.polito.it/oldsite/sas-ipl/
• Almost achieve channel capacity
• Without requiring of channelinformation at the transmitterside
• Allow variable block length
3 dB
2 dB
2 dB
Ut(�t = 3dB)
X1
X2
X3
V 1t = 1
V 2t = 0
V 3t = 0
• BS: wishes to maximize thesystem throughput
• Only needs to learn the arg-max
Jie Ren (Drexel ASPITRG) ICFRA March 19th, 2014 5 / 20
Page 6
Introduction
BackgroundInteractive Communication
• Interaction for Lossy SourceReproduction (Kaspi 1985)
• Interaction for functioncomputation (Ishwar & Ma2011)
• Benefit can be arbitrarily large• Infinite rounds interaction
may help
Rt ={R|∃Ut , s.t.∀i = 1, · · · , tRi ≥ I (X ;Ui |Y ,U i−1), Ui − (X ,U i−1)− Y , i odd
Ri ≥ I (Y ;Ui |Y ,U i−1), Ui − (Y ,U i−1)− X , i even
H(f (X ,Y )|Y ,Ut) = 0}
(1)
Jie Ren (Drexel ASPITRG) ICFRA March 19th, 2014 6 / 20
Page 7
Introduction
Main Contribution
Achievable Interactive Communication Scheme for ResourceAllocation
• Determine the arg-max (use rateless codes for data transmission)
• Solve by dynamic programming
• Show huge savings
Jie Ren (Drexel ASPITRG) ICFRA March 19th, 2014 7 / 20
Page 8
Problem Model
Outline
1 Introduction
2 Problem Model
3 Analysis
4 Results
5 Conclusions
Jie Ren (Drexel ASPITRG) ICFRA March 19th, 2014 8 / 20
Page 9
Problem Model
Problem Model
3 dB
2 dB
2 dB
Ut(�t = 3dB)
X1
X2
X3
V 1t = 1
V 2t = 0
V 3t = 0
Notations
• Xi ∈ Xt = {at , . . . , bt}• Ut Broadcasting message at
round t
• V it Replied message from MS i
at round t
Achievable Interaction Scheme
1: BS broadcasts a threshold λt atround t
2: MS i replies a 1 if Xi ≥ λt and 0otherwise
3: Stops when BS knows arg-maxreliably
Jie Ren (Drexel ASPITRG) ICFRA March 19th, 2014 9 / 20
Page 10
Problem Model
Problem Model
3 dB
2 dB
2 dB
Ut(�t = 3dB)
X1
X2
X3
V 1t = 1
V 2t = 0
V 3t = 0
Some Assumptions
• BS knows the initial distributionof X
• BS knows the initial number ofMSs
• MSs are not allowed tocommunicate with each other
Jie Ren (Drexel ASPITRG) ICFRA March 19th, 2014 10 / 20
Page 11
Analysis
Outline
1 Introduction
2 Problem Model
3 Analysis
4 Results
5 Conclusions
Jie Ren (Drexel ASPITRG) ICFRA March 19th, 2014 11 / 20
Page 12
Analysis
Analysis
Non-increasing Support set of X
If some users reply 1
at+1 = λt
bt+1 = bt
Ft+1(x) =Ft(x)− Ft(λt)
Ft(bt)− Ft(λt)
(2)
If no user replies 1
at+1 = at
bt+1 = λt
Ft+1(x) =Ft(x)− Ft(at)
Ft(λt)− Ft(at)
(3)
Jie Ren (Drexel ASPITRG) ICFRA March 19th, 2014 12 / 20
Page 13
Analysis
Analysis
Aggregate rate
Rt(λ) = H(λ|λ1, · · · , λt−1) + Nt + (Ft(λ))NtR∗(Nt , at , λ)
+Nt∑i=1
(1− Ft(λ))iFt(λ)Nt−i Nt !
i !(Nt − i)!R∗(i , λ, bt) (4)
Policy Iteration
λ∗t = arg minλ
Rt(λ) (5)
Jie Ren (Drexel ASPITRG) ICFRA March 19th, 2014 13 / 20
Page 14
Analysis
Analysis
• Efficiently Encode the Threshold
H(λt |λ1, . . . , λt−1) (6)
• Why H(Nt |Nt−1) works?• Xt and Nt determines λ∗t• Xt−1,Nt−1 and Nt determinesXt
• Two other strategies• Non-conditioning Encode the
Threshold: H(λt)• Encode the Number of Users:
H(Nt |Nt−1)
Xt =
{λ∗t−1, bt−1} if Nt < Nt−1
{at−1, λ∗t−1} if Nt = Nt−1 and λ∗t−1 > xi
{λ∗t−1, bt−1} if Nt = Nt−1 and λ∗t−1 ≤ xi
(7)
Jie Ren (Drexel ASPITRG) ICFRA March 19th, 2014 14 / 20
Page 15
Results
Outline
1 Introduction
2 Problem Model
3 Analysis
4 Results
5 Conclusions
Jie Ren (Drexel ASPITRG) ICFRA March 19th, 2014 15 / 20
Page 16
Results
Results
• X = {1, . . . , 16}
2 3 4 5 6 7 8 9 10 11 125
10
15
20
25
30
35
40
45
50
number of users
overh
eads
Non−interaction
Interaction
Sending Threshold
Sending Number of Uers
One−way Limit
Jie Ren (Drexel ASPITRG) ICFRA March 19th, 2014 16 / 20
Page 17
Results
Some Extensions
Interaction with Distortion
E[max{X1, . . . ,XNt} − Xi ] ≤ D (8)
Bits Cost Vs. Time Cost
C = µR + (1− µ)T (9)
Jie Ren (Drexel ASPITRG) ICFRA March 19th, 2014 17 / 20
Page 18
Conclusions
Outline
1 Introduction
2 Problem Model
3 Analysis
4 Results
5 Conclusions
Jie Ren (Drexel ASPITRG) ICFRA March 19th, 2014 18 / 20
Page 19
Conclusions
Contribution & Future Work
Review of Contribution
• Achievable Interactive Communication Scheme for ResourceAllocation
• Solve by Dynamic Programming
Future Work
• Consider Scalar Quantization than the 1-bit Message
• Fundamental Limits (Rate-distortion Curve)
• Resource Allocation in MIMO system
Jie Ren (Drexel ASPITRG) ICFRA March 19th, 2014 19 / 20
Page 20
Conclusions
Acknowledgments
Supported by the AFOSR under agreement number FA9550-12-1-0086
Jie Ren (Drexel ASPITRG) ICFRA March 19th, 2014 20 / 20