Game Theory in Communications

Multi-user communication theory typically studies the fundamental limits of communication systems, and considers communication schemes that approach or even achieve these limits. The functioning of many such schemes assumes that users always cooperate, even when it is not in their own best interest. In practice, this assumption need not be fulfilled, as rational communication participants are often only interested in maximizing their own communication experience, and may behave in an undesirable manner from the system's point of view. Thus, communication systems may operate differently than intended if the behavior of individual participants is not taken into account. In this thesis, we study how users make decisions in wireless settings, by considering their preferences and how they interact with each other. We investigate whether the outcomes of their decisions are desirable, and, if not, what can be done to improve them. In particular, we focus on two related issues. The first is the decision-making of communication users in the absence of any central authority, which we consider in the context of the Gaussian multiple access channel. The second is the pricing of wireless resources, which we consider in the context of the competition of wireless service providers for users who are not contractually tied to any provider, but free to choose the one offering the best tradeoff of parameters. In the first part of the thesis, we model the interaction of self-interested users in a Gaussian multiple access channel using non-cooperative game theory. We demonstrate that the lack of infrastructure leads to an inefficient outcome for users who interact only once, specifically due to the lack of coordination between users. Using evolutionary game theory, we show that this inefficient outcome would also arise as a result of repeated interaction of many individuals over time. On the other hand, if the users correlate their decoding schedule with the outcome of some publicly observed (pseudo) random variable, the resulting outcome is efficient. This shows that sometimes it takes very little intervention on the part of the system planner to make sure that users choose a desirable operating point. In the second part of the thesis, we consider the competition of wireless service providers for users who are free to choose their service provider based on their channel parameters and the resource price. We model this situation as a two-stage game where the providers announce unit resource prices in the first stage and the users choose how much resource they want to purchase from each provider in the second stage. Under fairly general conditions, we show that the competitive interaction of users and providers results in socially optimal resource allocation. We also provide a decentralized primal-dual algorithm and prove its convergence to the socially optimal outcome.