Title: Cooperation in Ad Hoc Networks: a Game-Theoretic Approach
Speaker: Luiz A. Dasilva (joint work with Allen B. MacKenzie) 12:10-12:30

ABSTRACT

Ad hoc networks are built on the premise of node cooperation, due to the absence of centralized control. This cooperation can take many forms: distributed selection of channels and power levels, decisions of whether to forward packets for one's neighbors, backoff procedures in case of collision, etc. We apply cooperative and non-cooperative game theory to the study of cooperation in ad hoc networks. This presentation highlights two applications: spectrum sharing and topology control. We design a bargaining-based distributed algorithm for opportunistic spectrum sharing: nodes autonomously decide what channels to operate in and how to set power levels so as to achieve fair and efficient use of resources. We then compare our algorithm, which is based on the Nash Bargaining Solution, to a distributed greedy waterfilling algorithm; ours achieves greater fairness without sacrificing efficiency in the use of the spectrum. In the area of topology control, we use a repeated game model to design packet forwarding policies that will lead to desirable network connectivity properties while considering individual nodes' energy conservation objectives. More broadly, this work considers the analysis and design of incentive mechanisms so that self-interest by each node will lead to globally-desirable outcomes. Finally, we will outline ongoing work in the development of a multi-institution competition, the Mobile Ad Hoc Networking Interoperability And Cooperation (MANIAC) challenge, which will allow us to study the incentive-compatibility of resource sharing in the absence of monetary rewards and explicit reputation schemes.