Informationally Efficient Multi-user Communication

Yi Su
Advisor: Mihaela van der Schaar

Wednesday, May 19, 2010 at 9:30am
Engr. IV Maxwell Room 57-124

Abstract:
The rapid increase in the demand for data rate over wired and wireless communication networks has led to a rethinking of the traditional network architecture and design principles. In fact, communication systems are inherently informationally decentralized competitive environments, where multiple devices executing a variety of applications and services need to locally adapt their transmission strategies based on their available information and compete for scarce networking resources. The concepts and techniques that have dominated multi-user communication research in recent years are not well suited for these informationally decentralized environments. Specifically, most existing research has focused on two extreme multi-user interaction scenario, the complete information scenario with a common system-wide objective (e.g. Pareto optimality) and the private information scenario with conflicting objectives (e.g. Nash equilibrium (NE)).

The objective of this dissertation is to characterize users' optimal strategies to improve their performance subject to varying degrees of informational constraints in several classes of multi-user communication environments. We will mainly focus on fully distributed solutions without any real-time information exchange between different users, which perfectly satisfy the informationally efficient requirement in communication systems. In particular, we investigate three key problems in information-constrained multi-user communication systems. First, when will a distributed algorithm (e.g. best response dynamics) converge to a NE? And how fast? Second, if information is constrained and no information exchange between users is allowed, how to improve an inefficient NE without message passing? Last, assuming no real-time information exchange between users, can we still achieve Pareto optimality? We propose and analyze two new classes of games named additively coupled sum constrained games and linearly coupled games in which we individually address the three questions. In particular, we provide sufficient conditions that guarantee the uniqueness of NE and best response dynamics linearly converges to the NE. We also provide conjectural equilibrium based solutions that can substantially improve the performance of inefficient NE and fully recover Pareto optimality without any real-time information exchange between users. The investigated game models apply to a variety of realistic applications in multi-user communication systems, including multi-channel power control, flow control, and wireless random access.

Biography:
Yi Su received the bachelor's and master's degrees from Tsinghua University, Beijing, China, in 2004 and 2006, respectively, both in Electrical Engineering. He is currently pursuing the Ph.D. degree in the Department of Electrical Engineering, University of California, Los Angeles. During the summer of 2009, he was an intern with Qualcomm, Santa Clara, CA. He was selected by IBM Research as one of the 10 top Ph.D. students to participate in the 2009 Watson Emerging Leaders in Multimedia Workshop. He was awarded UCLA Dissertation Year Fellowship (2009-2010), Department Fellowship (Fall 2007), and University Fellowship (2006-2007). He ranked the first in signals and systems area in PhD preliminary examination, April 2007. His research interests include game theoretic solutions for multi-user communications, distributed optimization and learning in multi-agent systems, cross-layer optimization in wireless multimedia communications, and physical layer issues on wireless communications.