ofdma cross layer resource controlofdma cross layer resource control 3/20 system model (๐‘ด users,...

20
OFDMA Cross Layer Resource Control Gwanmo Ku Adaptive Signal Processing and Information Theory Research Group Jan. 25, 2013

Upload: others

Post on 15-Mar-2020

8 views

Category:

Documents


0 download

TRANSCRIPT

OFDMA Cross Layer

Resource Control

Gwanmo Ku

Adaptive Signal Processing and Information Theory Research Group

Jan. 25, 2013

Outline

OFDMA Cross Layer Resource Control

Objective Functions

- System Throughput (L1), Total Transmit Power (L1)

Constraints

- Transmit Power Constraint (L1)

- Quality of Service (User Demand, Fairness), Buffer Status (L2-3)

- Stability (L2-3)

Generalized Cross Layer Control (GCLC)

Stochastic Network Optimization (SNO)

Network Utility Maximization (NUM)

2/20

OFDMA Cross Layer Resource Control

3/20

System Model (๐‘ด users, ๐‘ฒ subcarriers)

Base Station (eNB)

Mobile (UE)

๐’–๐Ÿ

๐’–๐‘ด

โ€ฆ

Higher Layer

Buffer

PHY

Higher Layer

Buffer

PHY

OFDMA

Ian Wong & Brian Evans

GCLC

๐’“๐’Ž

OFDMA Resource Control

4/20

Objective Functions

System Throughput Maximization

Transmit Power Minimization

Constraints

Transmit Power Constraint

Quality of Service

User Demands : Each User Required Data Rate

Fairness : Minimum User Data Rate

Stability based on Buffer Status

OFDMA Resource Allocation

5/20

Notations

๐’Ž โˆˆ {๐Ÿ,โ€ฆ ,๐‘ด} User Index

๐’Œ โˆˆ {๐Ÿ,โ€ฆ ,๐‘ฒ} Subcarrier Index

๐’‘๐’Ž,๐’Œ : Power Control Coefficient

๐œธ๐’Ž,๐’Œ : SINR for user index ๐’Ž and subcarrier index ๐’Œ

๐‘ท๐‘ป Total Transmit Power Constraint

๐’“๐’Ž Required Each User Data Rate

๐’“๐ŸŽ Required Minimum User Data Rate

๐’ƒ๐’Ž Buffer Service Rate

๐‘น๐’Ž Overall Coding Rate for User ๐’Ž

OFDMA Resource Allocation

6/20

System Throughput Maximization

Power Control

๐’‘๐’Ž,๐’Œ = argmaxE ๐‘ค๐‘š

๐‘€

๐‘š=1

log(1 + ๐‘๐‘š,๐‘˜๐›พ๐‘š,๐‘˜)

๐พ

๐‘˜=1

๐’๐ฒ๐ฌ๐ญ๐ž๐ฆ ๐“๐ก๐ซ๐จ๐ฎ๐ ๐ก๐ฉ๐ฎ๐ญ

๐ธ ๐‘๐‘š,๐‘˜

๐พ

๐‘˜=1

๐‘€

๐‘š=1

โ‰ค ๐‘ƒ๐‘‡

s.t

max (๐›ฝ๐‘š๐‘…๐‘š๐‘Ÿ0, ๐›ฝ๐‘š๐‘…๐‘š๐‘Ÿ๐‘š) โ‰ค ๐‘ค๐‘š log(1 + ๐‘๐‘š,๐‘˜๐›พ๐‘š,๐‘˜)

๐พ

๐‘˜=1

Stability

OFDMA Resource Allocation

7/20

Work by Ian Wong and Brian Evans

System Throughput Maximization with Tx. Power Constraint

๐‘๐‘š,๐‘˜ = argmax ๐„ ๐‘ค๐‘š

๐‘€

๐‘š=1

log(1 + ๐‘๐‘š,๐‘˜๐›พ๐‘š,๐‘˜)

๐พ

๐‘˜=1

๐„ ๐‘๐‘š,๐‘˜

๐พ

๐‘˜=1

๐‘€

๐‘š=1

โ‰ค ๐‘ƒ๐‘‡

๐‘ค๐‘š

๐‘€

๐‘š=1

= 1

s.t

OFDMA Resource Allocation

8/20

Optimization Framework

Dual Optimization

๐ฟ ๐‘ โ‹… , ๐œ† = ๐„ ๐‘ค๐‘š

๐‘€

๐‘š=1

log(1 + ๐‘๐‘š,๐‘˜๐›พ๐‘š,๐‘˜)

๐พ

๐‘˜=1

+๐œ† ๐‘ƒ๐‘‡ โˆ’ ๐„( ๐‘๐‘š,๐‘˜

๐พ

๐‘˜=1

)

๐‘€

๐‘š=1

๐‘”โˆ— = min๐œ†โ‰ฅ0

ฮ˜(๐œ†)

ฮ˜ ๐œ† = max๐‘ โ‹… โˆˆ๐‘ƒ๐‘‡

๐ฟ(๐‘ โ‹… , ๐œ†)

OFDMA Resource Allocation

9/20

Dual

ฮ˜ ๐œ† = max๐‘ โ‹… โˆˆ๐‘ƒ๐‘‡

๐ฟ(๐‘ โ‹… , ๐œ†)

= ๐œ†๐‘ƒ๐‘‡ + max๐‘ โ‹… โˆˆ๐‘ƒ๐‘‡

๐„ ๐‘ค๐‘š log 1 + ๐‘๐‘š,๐‘˜๐›พ๐‘š,๐‘˜ โˆ’ ๐œ†๐‘๐‘š,๐‘˜

๐‘€

๐‘š=1

๐พ

๐‘˜=1

= ๐œ†๐‘ƒ๐‘‡ + max๐‘๐‘˜ โ‹… โˆˆ๐‘ƒ๐‘˜

๐„ ๐‘ค๐‘š log 1 + ๐‘๐‘š,๐‘˜๐›พ๐‘š,๐‘˜ โˆ’ ๐œ†๐‘๐‘š,๐‘˜

๐‘€

๐‘š=1

๐พ

๐‘˜=1

= ๐œ†๐‘ƒ๐‘‡ + ๐ธ max๐‘๐‘˜ โ‹… โˆˆ๐‘ƒ๐‘˜

๐‘ค๐‘š log 1 + ๐‘๐‘š,๐‘˜๐›พ๐‘š,๐‘˜ โˆ’ ๐œ†๐‘๐‘š,๐‘˜

๐‘€

๐‘š=1

๐พ

๐‘˜=1

= ๐œ†๐‘ƒ๐‘‡ + ๐พ๐ธ๐›พ๐‘˜max

๐‘šโˆˆ{1,โ€ฆ,๐‘€}max

๐‘๐‘š,๐‘˜โ‰ฅ0(๐‘ค๐‘š log 1 + ๐‘๐‘š,๐‘˜๐›พ๐‘š,๐‘˜ โˆ’ ๐œ†๐‘๐‘š,๐‘˜)

multilevel water filling

๐‘š๐‘Ž๐‘ฅ ๐‘‘๐‘ข๐‘Ž๐‘™ ๐‘ข๐‘ ๐‘’๐‘Ÿ ๐‘ ๐‘’๐‘™๐‘’๐‘๐‘ก๐‘–๐‘œ๐‘›

OFDMA Resource Allocation

10/20

A Simple Closed Form

๐‘ ๐‘š,๐‘˜(๐œ†) =1

๐›พ0,๐‘š ๐œ†โˆ’

1

๐›พ๐‘š,๐‘˜

+

๐‘ฅ + = max(0, ๐‘ฅ)

๐›พ0,๐‘š ๐œ† =๐œ† ln 2

๐‘ค๐‘š Cutoff value

๐‘”โˆ— = min๐œ†โ‰ฅ0

[๐œ†๐‘ƒ๐‘‡ + ๐พ ๐„๐›พ๐‘˜๐‘”๐‘˜ ๐›พ๐‘˜ , ๐œ† ]

๐‘”๐‘˜ ๐›พ๐‘˜ , ๐œ† = max๐‘šโˆˆ{1,โ€ฆ,๐‘€}

{๐‘”๐‘š,๐‘˜ ๐›พ๐‘š,๐‘˜ , ๐œ† }

๐‘”๐‘š,๐‘˜ ๐›พ๐‘š,๐‘˜ , ๐œ† = ๐‘ค๐‘š log 1 + ๐‘ ๐‘š,๐‘˜ ๐œ† โˆ’ ๐œ† ๐‘ ๐‘š,๐‘˜ ๐œ†

OFDMA Resource Allocation

11/20

Optimal Solution

๐‘”๐‘š,๐‘˜ ๐›พ๐‘š,๐‘˜ , ๐œ† = ๐‘ค๐‘š log 1 + ๐‘ ๐‘š,๐‘˜ ๐œ† โˆ’ ๐œ† ๐‘ ๐‘š,๐‘˜ ๐œ†

=๐‘ค๐‘š

ln 2ln

๐›พ๐‘š,๐‘˜

๐›พ0,๐‘š ๐œ†โˆ’

๐‘ค๐‘š

ln 2+

๐œ†

๐›พ๐‘š,๐‘˜๐‘ข(๐›พ๐‘š,๐‘˜ โˆ’ ๐›พ0,๐‘š ๐œ† )

๐‘ข ๐‘ฅ = 0 ๐‘ฅ < 01 ๐‘ฅ โ‰ฅ 0

๐œ†โˆ— = ๐‘Ž๐‘Ÿ๐‘”min๐œ†โ‰ฅ0

[๐œ†๐‘ƒ + ๐พ ๐‘”๐‘˜๐‘“๐‘”๐‘˜๐‘”๐‘˜ ๐‘‘๐‘”๐‘˜

โˆž

0

]

๐‘ ๐‘š,๐‘˜(๐œ†โˆ—) =

1

๐›พ0,๐‘š ๐œ†โˆ—โˆ’

1

๐›พ๐‘š,๐‘˜

+

๐‘๐‘š,๐‘˜โˆ— = ๐‘ ๐‘š,๐‘˜ ๐œ†โˆ— 1(๐‘š = ๐‘š๐‘˜

โˆ— )

๐‘š๐‘˜โˆ— = argmax

๐‘šโˆˆ{1,โ€ฆ,๐‘€}๐‘ค๐‘š log 1 + ๐‘ ๐‘š,๐‘˜ ๐œ†โˆ— ๐›พ๐‘š,๐‘˜ โˆ’ ๐œ†โˆ—๐‘ ๐‘š,๐‘˜ (๐œ†โˆ—)

Cross Layer Control

12/20

Generalized Cross Layer Control (GCLC)

Proposed by Georgiadis, Neely, and Tassiulas

Focus on Stability based on Queuing Statistics

โ€ข Stochastic Network Optimization

โ€ข Network Utility Maximization

Network Stability

โ€ข Differential Equation of Queuing Statistics

โ€ข Lyapunov Stability

Cross Layer Control

13/20

Stochastic Network Optimization

Buffer for user ๐‘š

Arrival Rate ๐€๐’Ž Service Rate ๐๐’Ž

Backlog Queue ๐‘ธ๐’Ž (๐’•)

Network State Variable ๐‘บ(๐’•)

Control Action ๐‘ฐ ๐’• โˆˆ ๐‘ฐ๐‘บ(๐’•) feasible control region under ๐‘บ(๐’•)

๐ ๐‘ก๐‘† ๐‘ก ,๐ผ(๐‘ก)

๐(๐‘ก + 1)

Cross Layer Control

14/20

Stochastic Network Optimization

Stability Issue

๐‘„๐‘š ๐‘ก + 1 โ‰ค max ๐‘„๐‘š ๐‘ก โˆ’ ๐‘…๐‘š๐‘œ๐‘ข๐‘ก ๐ผ ๐‘ก , ๐‘† ๐‘ก , 0 + ๐‘…๐‘š

๐‘–๐‘›(๐ผ ๐‘ก , ๐‘† ๐‘ก )

๐‘ถ๐’–๐’•๐’ˆ๐’๐’Š๐’๐’ˆ ๐‘ธ๐’–๐’†๐’–๐’† ๐‘ฌ๐’๐’•๐’†๐’“๐’Š๐’๐’ˆ ๐‘ธ๐’–๐’†๐’–๐’†

lim๐‘กโ†’ โˆž

sup1

๐‘ก ๐„{๐‘„๐‘š ๐œ }

๐‘กโˆ’1

๐œ=0

< โˆž

Lyapunov Stability

If there exist ๐‘ฉ > ๐ŸŽ and ๐ > ๐ŸŽ, such that for all

times slot ๐’• we have :

Then network is strongly stable, and

๐„ โˆ† ๐‘ก Q ๐‘ก โ‰ค ๐ต โˆ’ ๐œ– ๐‘„๐‘š(๐‘ก)

๐‘€

๐‘š=1

lim๐‘กโ†’ โˆž

sup1

๐‘ก ๐„{Q ๐œ }

๐‘กโˆ’1

๐œ=0

<๐ต

๐œ–

15/20

Cross Layer Control

16/20

Find ๐‘ฐ(๐’•)

Find ๐šฒ by Lyapunov Drift

Drift Definition

๐ผโˆ— ๐‘ก = argmax๐ผ ๐‘ก โˆˆ๐ˆ๐‘†(๐‘ก)

๐‘Š๐‘Ž๐‘โˆ— ๐‘ก ๐œ‡๐‘Ž๐‘

โˆ— (๐‘ก)

๐‘Ž๐‘

๐‘Š๐‘Ž๐‘โˆ— ๐‘ก = max

๐œ‡๐‘Ž๐‘

๐‘„๐‘Ž ๐‘ก โˆ’ ๐‘„๐‘ ๐‘ก +

maximum queue backlog differential

โˆ† ๐‘ก = ๐ฟ ๐‘ก + 1 โˆ’ ๐ฟ(๐‘ก)

๐ฟ ๐‘ก =1

2 ๐‘„๐‘š

2 (๐‘ก)

๐‘€

๐‘š=1

Lyapunov Drift

17/20

โˆ† ๐‘ก = ๐ฟ ๐‘ก + 1 โˆ’ ๐ฟ(๐‘ก)

=1

2 [๐‘„๐‘š

2 ๐‘ก + 1 โˆ’

๐‘€

๐‘š=1

๐‘„๐‘š2 (๐‘ก)]

After applying ๐‘„๐‘š2 (๐‘ก + 1)

Find Lyapunov Bound with Conditional Expectation

โˆ† ๐‘ก ๐ ๐‘ก โ‰ค โ€ฆ

lim๐‘กโ†’ โˆž

sup1

๐‘ก ๐„{๐‘„๐‘š ๐œ }

๐‘กโˆ’1

๐œ=0

< โˆž

Network Utility Maximization (NUM)

18/20

Rate ๐ซ โˆˆ ๐šฒ with Maximum Utility

๐ซโˆ— = argmax๐ซโ‰ค๐›Œ

๐‘” ๐ซ |๐ซ โˆˆ ๐šฒ

๐‘”(๐ซ) : Utility Function

Minimize Cost

Generalized Cross Layer Control

19/20

General Form of GCLC

Cost Variable Vector ๐ฑ : Maximum cost constraints ๐

Utility Variable Vector ๐ฒ :Minimum utility constraints ๐‡

Stable Region ๐ซ โˆˆ ๐šฒ

โ€ข Arrival Rate Vector ๐›Œ

Minimize net cost

โ€ข Natural cost function ๐’‡(๐’™) and Concave Utility function ๐’ˆ(๐’š)

min๐ซโ‰ค๐›Œ

๐‘“ ๐ฑ โˆ’ ๐‘”(๐ฒ)|๐ช ๐ฑ โ‰ค ๐, ๐ก ๐ฒ โ‰ฅ ๐‡, ๐ซ โˆˆ ๐šฒ

OFDMA Resource Control via GCLC

20/20

OFDMA via GCLC

Cost Variable Vector ๐ฑ : power coefficients

โ€ข ๐ = ๐๐“ : Total Power Constraint

Utility Variable Vector ๐ฒ : user data rate ๐ฒ = ๐ซ

โ€ข ๐‡ : Quality of Service (User demands, Fairness)

Stable Region based on Queuing Statistics

Minimize net cost : Maximize System Throughput

min๐ซโ‰ค๐›Œ

๐‘“ ๐ฑ โˆ’ ๐‘”(๐ฒ)|๐ช ๐ฑ โ‰ค ๐, ๐ก ๐ฒ โ‰ฅ ๐‡, ๐ซ โˆˆ ๐šฒ