Collision mitigation for cognitive radio networks using local congestion game

This paper investigates the problem of distributed channel selection for cognitive radio networks, where mutual inference only occurs among nearby users, using a game based solution. Firstly, we propose a local congestion game, in which the utility function is determined by its own channel selection and the channel selection profile of its neighboring users. The game is proved to be a potential game with the aggregate collision level serving as the potential function. Then, a stochastic learning automata based distributed channel selection algorithm is proposed, with which the users learn the desirable channel selections from their action-payoff history. It is analytically shown that the proposed learning algorithm converges to pure strategy Nash equilibrium (NE) points without information exchange. Moreover, it maximizes the aggregate collision level globally or locally, and hence achieves higher network throughput.