Game Theoretical Framework for Cooperation in Autonomous Wireless Networks

Abstract

With the explosive growth of wireless networking techniques in the last decade, connecting to the world from any place, at any time and for any body is no longer just a dream. New concepts of network infrastructures such as mobile ad hoc networks or dynamic spectrum access networks emerged in recent years to provide more flexible wireless networking, efficient spectrum usage and robust network connections. With the development of intelligent wireless devices such as cognitive radios, the network users' capability has been largely increased. It becomes important to analyze and understand the network users' intelligent behaviors, especially selfish behaviors. Therefore, we focus our study on these new types of wireless networks with selfish users, which need to be self-organizing and decentralized. They are also referred to as autonomous wireless networks.

In order to analyze the selfish behaviors of network users for efficient autonomous wireless networking, we analyze the cooperation in autonomous wireless networks under a comprehensive game theoretical framework. Game theory models strategic interactions among agents using formalized incentive structures. It not only provides game models for efficient self-enforcing distributed design but also derives well-defined equilibrium criteria to measure the optimality of game outcomes for various scenarios.

In this dissertation, we first study the cooperation enforcement in autonomous wireless networks under noise and imperfect information. We model the interactions among users as multi-stage games and propose a set of belief-assisted approaches to ensure cooperation by allowing reputation effects or retribution. For instance, a user in an ad hoc network will forward packets for the others if they have built up high belief values through their past cooperative behaviors, i.e., forwarding packets. Further, we investigate the impacts of network dynamics on game theoretical cooperation stimulation/enforcement in autonomous wireless networks. We model the dynamic interaction among users as multi-stage dynamic games and develop various dynamic pricing approaches to stimulate cooperation among users by using payments as incentives based on auction rules and dynamic programming. Finally, we exploit the collusive selfish behaviors in autonomous wireless networks in a non-cooperative game theoretical framework and devise countermeasures to combat or alleviate collusive behaviors.