A Learning-based Approach for Distributed Multi-Radio Channel Allocation in Wireless Mesh Networks

We consider the distributed channel allocation problem in IEEE 802.11-based multi-radio wireless mesh networks. We develop a new scalable protocol termed LCAP for efficient and adaptive distributed multi-radio channel allocation. In LCAP, nodes autonomously learn their channel allocation based on neighborhood and channel usage information, which is obtained via a novel neighbor discovery protocol that enables neighboring nodes to efficiently discover each other even when they do not share a common channel. Extensive simulation-based evaluation of LCAP relative to the state-of-the-art Asynchronous Distributed Coloring (ADC) protocol demonstrates that LCAP is able to achieve its stated objectives of efficient channel utilization across diverse traffic patterns, protocol scalability and adaptivity to factors such as external interference. We also present a prototype implementation of the LCAP neighbor discovery module that is key to implementing the proposed approach.