A Game Theoretic Solution for Relay Selection in Dynamic Cooperative Networks

This paper investigates the problem of relay selection in dynamic networks where the active source node set is randomly changing from slot to slot. Furthermore, each source node only has its individual information and there is no information exchange among source nodes. The goal of the source nodes is to determine their relay selections. This optimization problem is hard to solve, even in a centralized method, because the active source node set is changing randomly, and the probability distributions of the active source node sets are unknown. We formulate the optimization problem as a non-cooperative game which is a congestion game with player-specific payoff functions. To cope with the dynamics and incomplete information, we propose a multi-user learning algorithm with low complexity to achieve the stable solution. Simulation results validate the convergence of the algorithm. And the achieved expected throughput of the proposed approach is close to the optimum.