For full cooperative diversity gains to be achieved while still maintaining spectral and energy efficiency, relay assignment schemes for cooperative communications have been extensively studied in recent research. These schemes select only the best relay from multiple relaying candidates to cooperate with a communication link. However, it is challenging to find the optimal relay in distributed wireless networks because of the dynamic nature of such networks. In this paper, we first formulate

more »

... problem of relay assignment as a noncooperative, mixed-strategy, repeated game, where relaying candidates are modeled as rational players. We then propose a game-theory-based relay assignment scheme GTRA, in which each player plays against all the other players and determines whether to cooperate with a communication link on a packet-bypacket basis in a distributed manner. To adapt to dynamic environments, players utilized an adaptive learning algorithm, that is, modified-regret-matching algorithm, to learn optimal strategies of relay assignment, as well as to orient the game to converge to a set of correlated equilibriums, which is often more system efficient than a Nash equilibrium. To evaluate the performance of GTRA, we compare it with BR, a fictitious two-player game-based approach. Simulation results have shown that GTRA outperforms BR in terms of network throughput, especially in environments where the channel fading becomes severe. It is also shown that GTRA can converge to a correlated equilibrium in a short period that enables the GTRA to work well in dynamic environments.