Relay assignment over distributed wireless networks: A game theoretic approach

In this paper, we focus on the relay assignment issue in an ad-hoc network environment. Our objective is to improve the total capacity of all users randomly located in an area. We designed two distributed algorithms, Distributed Game-theoretic Relay Assignment (DGRA) and Distributed Fast-converge Relay Assignment (DFRA), which can achieve correlated equilibrium (CE) and low time complexity respectively. Compared with the centralized algorithms, our mechanisms do not need global information or centralized control. The distributed algorithms can fit the ad-hoc network much better. We consider all the relay candidates as rational players and formulate the relay assignment issue as a non-cooperative, mixed-strategy and repeated game. Each player adaptively makes decisions attending to achieve a higher utility in a modified-regret-matching manner by stages. Also, we show that algorithm DGRA converges to a set of CE(s). DFRA is designed as a fast converge algorithm to gain low time complexity with comparable performance. Extensive simulations show that a huge computation reduction is achieved compared with the centralized algorithm.