Optimal distributed interference avoidance: potential game and learning

Abstract

ABSTRACTThis article studies the problem of distributed interference avoidance (IA) through channel selection for distributed wireless networks, where mutual interference only occurs among nearby users. First, an interference graph is used to characterise the limited range of interference, and then the distributed IA problem is formulated as a graph colouring problem. Because solving the graph colouring problem is non‐deterministic polynomial hard even in a centralised manner, the task of obtaining the optimal channel selection profile distributively is challenging. We formulate this problem as a channel selection game, which is proved to be an exact potential game with the weighted aggregate interference serving as the potential function. On the basis of this, a distributed learning algorithm is proposed to achieve the optimal channel selection profile that constitutes an optimal Nash equilibrium point of the channel selection game. The proposed learning algorithm is fully distributed because it needs information about neither the network topology nor the actions and the experienced interference of others. Simulation results show that the proposed potential game theoretic IA algorithm outperforms the existing algorithm because it minimises the aggregate weighted interference and achieves higher network rate. Copyright © 2012 John Wiley & Sons, Ltd.