Belief-free equilibrium of packet forwarding game in ad hoc networks under imperfect monitoring

Future applications will require autonomous devices to be interconnected and form ad hoc networks. In such networks, cooperation will be the first problem to solve at all layers of the protocol stack. This paper deals with one of the basic functions of a network, namely packet forwarding. We model packet forwarding as a stochastic game in which each node monitors the behavior of its neighbors. We consider the realistic scenario when the monitoring technology used by the nodes is imperfect. In reality, there can be inconsistencies between the true action of a node and the observations of its neighbors. Therefore, in an ad hoc network, each node receives only noisy private information about the past play of its neighbors. We develop a simple one period memory strategy that constrains self-interested nodes to cooperate under noise. We use a belief-free approach. A belief-free approach delivers a tremendous computational advantage because nodes' belief about the neighbors' private history does not need to be computed. We support our results by mathematical proofs and simulations.