Stochastic Local Search Based Channel Assignment in Wireless Mesh Networks

Abstract

In this paper, we consider the problem of channel assignment in multi-radio, multi-channel wireless mesh networks. We assume a binary interference model and represent the set of interfering links in a network topology as a conflict graph. We then develop a new centralised stochastic local search algorithm to find a channel assignment that minimises the network interference. Our algorithm assigns channels to communication links rather than radio interfaces. By doing so, our algorithm not only does preserve the network topology, but is also independent of the network routing layer. We compare the performance of our algorithm with that of a well-known Tabu-based approach (by Subramanian et al.) on randomly generated sparse and dense network topologies. Using graph-theoretic evaluation and ns2 simulations (a widely used discrete event network simulator), we show that our algorithm consistently outperforms the Tabu-based approach in terms of both the network interference and the throughput obtained under various offered loads. In particular, for a practical setting of 3 radio interfaces per mesh node in a dense network topology with 12 channels available, our approach achieves 70% lower network interference and thus 15 times higher average throughput than those achieved by the Tabu-based approach.