Page 1 |
Save page Remove page | Previous | 1 of 195 | Next |
|
small (250x250 max)
medium (500x500 max)
Large (1000x1000 max)
Extra Large
large ( > 500x500)
Full Resolution
All (PDF)
|
This page
All
|
STOCHASTIC OPTIMIZATION OVER PARALLEL QUEUES: CHANNEL-BLIND SCHEDULING, RESTLESS BANDIT, AND OPTIMAL DELAY Copyright 2011 by Chih-ping Li A Dissertation Presented to the FACULTY OF THE USC GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial FUlfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (ELECTRICAL ENGINEERING) August 2011 Chih-ping Li
Object Description
Title | Stochastic optimization over parallel queues: channel-blind scheduling, restless bandit, and optimal delay |
Author | Li, Chih-ping |
Author email | chihpinl@usc.edu;chihping.li@gmail.com |
Degree | Doctor of Philosophy |
Document type | Dissertation |
Degree program | Electrical Engineering |
School | Viterbi School of Engineering |
Date defended/completed | 2011-06-22 |
Date submitted | 2011-07-29 |
Date approved | 2011-07-29 |
Restricted until | 2011-07-29 |
Date published | 2011-07-29 |
Advisor (committee chair) | Neely, Michael J. |
Advisor (committee member) |
Caire, Giuseppe Jain, Rahul Krishnamachari, Bhaskar Ward, Amy R. |
Abstract | This dissertation addresses several optimal stochastic scheduling problems that arise in partially observable wireless networks and multi-class queueing systems. They are single-hop network control problems under different channel connectivity assumptions and different scheduling constraints. Our goals are two-fold: To identify stochastic scheduling problems of practical interest, and to develop analytical tools that lead to efficient control algorithms with provably optimal performance. ❧ In wireless networks, we study three sets of problems. First, we explore how the energy and timing overhead due to channel probing affects network performance. We develop a dynamic channel probing algorithm that is both throughput and energy optimal. The second problem is how to exploit time correlations of wireless channels to improve network throughput when instantaneous channel states are unavailable. Specifically, we study the network capacity region over a set of Markov ON/OFF channels with unknown current states. Recognizing that this is a difficult restless multi-armed bandit problem, we construct a non-trivial inner bound on the network capacity region by randomizing well-designed round robin policies. This inner bound is considered as an operational network capacity region because it is large and easily achievable. A queue-dependent round robin policy is constructed to support any throughput vector within the inner bound. In the third problem, we study throughput utility maximization over partially observable Markov ON/OFF channels (specifically, over the inner bound provided in the previous problem). It has applications in wireless networks with limited channel probing capability, cognitive radio networks, target tracking of unmanned aerial vehicles, and restless multi-armed bandit systems. An admission control and channel scheduling policy is developed to achieve near-optimal throughput utility within the inner bound. Here we use a novel ratio MaxWeight policy that generalizes the existing MaxWeight-type policies from time-slotted networks to frame-based systems that have policy-dependent random frame sizes. ❧ In multi-class queueing systems, we study how to optimize average service cost and per-class average queueing delay in a nonpreemptive multi-class M/G/1 queue that has adjustable service rates. Several convex delay penalty and service cost minimization problems with time-average constraints are investigated. We use the idea of virtual queues to transform these problems into a new set of queue stability problems, and the queue-stable policies are the desired solutions. The solutions are variants of dynamic c-mu rules, and implement weighted priority policies in every busy period, where the weights are functions of past queueing delays in each job class. ❧ Throughout these problems, our analysis and algorithm design uses and generalizes an achievable region approach driven by Lyapunov drift theory. We study the performance region (in throughput, power, or delay) of interest and identify, or design, a policy space so that every feasible performance vector is attained by a stationary randomization over the policy space. This investigation facilitates us to design queue-dependent network control policies that yield provably optimal performance. The resulting policies make greedy and dynamic decisions at every decision epoch, require limited or no statistical knowledge of the system, and can be viewed as learning algorithms over stochastic queueing networks. Their optimality is proved without the knowledge of the optimal performance. |
Keyword | stochastic network optimization; queueing theory |
Language | English |
Part of collection | University of Southern California dissertations and theses |
Publisher (of the original version) | University of Southern California |
Place of publication (of the original version) | Los Angeles, California |
Publisher (of the digital version) | University of Southern California. Libraries |
Provenance | Electronically uploaded by the author |
Type | texts |
Legacy record ID | usctheses-m |
Contributing entity | University of Southern California |
Rights | Li, Chih-ping |
Physical access | The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given. |
Repository name | University of Southern California Digital Library |
Repository address | USC Digital Library, University of Southern California, University Park Campus MC 7002, 106 University Village, Los Angeles, California 90089-7002, USA |
Repository email | cisadmin@lib.usc.edu |
Archival file | uscthesesreloadpub_Volume6/etd-LiChihping-206.pdf |
Description
Title | Page 1 |
Contributing entity | University of Southern California |
Repository email | cisadmin@lib.usc.edu |
Full text | STOCHASTIC OPTIMIZATION OVER PARALLEL QUEUES: CHANNEL-BLIND SCHEDULING, RESTLESS BANDIT, AND OPTIMAL DELAY Copyright 2011 by Chih-ping Li A Dissertation Presented to the FACULTY OF THE USC GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial FUlfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (ELECTRICAL ENGINEERING) August 2011 Chih-ping Li |