EnergyEnergy-Efficient Transmission Scheduling with Strict Underflow Constraints

Graduate Symposium 2009

David I Shuman and Mingyan Liu, Electrical Engineering and Computer Science Department, University of Michigan Contact: dishuman@umich.edu

Motivating Application Wireless Media Streaming

Single source transmitting data streams to one or more receivers/users over a shared wireless channel Available data rate of the channel varies with time and from receiver to receiver
Sender
Buffer 1 User 1 Buffer 2

Summary of Contributions
1. Relation to inventory theory In inventory language, our problem is a multi-period, multiitem, discrete time inventory model with random ordering prices, deterministic demand, and budget constraints
Items

Mobile Receivers

Data streams for each of the mobile receivers Receiver buffers Random ordering prices Random channel conditions Deterministic demand Users packet requirements for playout Budget constraint Transmitters power constraint
Inventories

Scheduler

Buffer 3

Wireless Channel

User 2 User 3

2. Structure of optimal policy for the single receiver case

Low SNR Regime. Linear Power-Rate Curve with Peak Power Constraint Results in Modified Base-Stock Policy
y * ( x, c) x n
Transmission Cost (Power Consumed)

Buffer M User M
Optimal Number of Packets to Transmit

y * ( x, c ) n bn (c ) P c
0

Timing in Each Slot

Transmitter learns each channels condition Transmitter allocates some amount of power (possibly zero) for transmission to each user Transmission and reception Packets are removed / purged from each receivers buffer for playing

P c
0

P
P c
Packets Transmitted (a)

Optimal Buffer Level After Transmission

bn (c)

P c
(b)

bn (c )

bn (c)

P c
(c)

bn (c)

Buffer Level Before Transmission

Buffer Level Before Transmission

High SNR Regime. Piecewise Linear Convex Power-Rate Curve Results in Finite Generalized Base-Stock Policy
z * ( x , s) n x + z * ( x , s) n

Transmission Cost (Power Consumed)

bn,k 1 ( s)
Optimal Number of Packets to Transmit
~ ( s) z0

~ (s ) zk ~ (s ) z k1

Optimal Buffer Level After Transmission

bn,k ( s)

0
~ ( s ) ~ ( s) z1 z2

x 0 bn ,k 1 ( s) ~k 1 ( s ) z bn,k ( s ) ~k 1 ( s ) z

x
0

bn , k ( s) ~k 1 ( s) z bn, k ( s ) ~k ( s ) z

Packets Transmitted

Buffer Level Before Transmission (b)

Buffer Level Before Transmission (c)

Transmission Scheduling Objectives

Avoid underflow so as to ensure playout quality Minimize system-wide power consumption in order to:
Prolong the systems lifetime Maximize the number of receivers the sender can support Limit interference

(a)

We can also determine the critical numbers when certain technical conditions are satisfied 3. Structure of optimal policy for the two receiver case
x2

Main idea Opportunistic Scheduling

Exploit the temporal and spatial variation of the channel by transmitting more data when the channel condition is good and less data when the channel condition is bad Challenge is to determine what is a good condition and how much data to send accordingly Strict underflow constraints represent a competing quality-ofservice interest

Buf f er Level of User 2 Bef ore Transmission

2 bn (c1, c 2 )

x1

b1 (c1, c 2 ) n

Buf f er Level of User 1 Bef ore Transmission

Ongoing Work
Develop near-optimal policies and better intuition for general M-receiver case through numerical approximation techniques

Aerospace Engineering Applied Physics Atmospheric, Oceanic and Space Sciences Biomedical Engineering Chemical Engineering Civil and Environmental Engineering Electrical Engineering and Computer Science Industrial and Operations Engineering Interdisciplinary Programs Macromolecular Science and Engineering Materials Science and Engineering Mechanical Engineering Naval Architecture and Marine Engineering Nuclear Engineering and Radiological Sciences