Dynamic interference adaptation for wireless mesh networks

Most wireless meshes will have to operate within the crowded unlicensed spectrum that is also shared by numerous uncoordinated 802.11 hotspots. This creates an unpredictable and variable spectrum space that mesh networks need to co-exist within. We propose a novel method for adapting to such external interference by dynamically changing the assignment of channels to the backbone links, yet retaining the same "logical" network-wide channel assignment. Called Connected- component based Adaption, this method ensures two important properties, (i) Allows distributed changes to the channels used by backbone links depending on local interference, and (ii) Allows a centralized algorithm to dictate the high-level channel and route assignments used by the network as a whole. We propose MeshChop as a randomized algorithm that uses channel hopping to achieve connected-component based adaptation. We show that MeshChop has minimal overheads and provides good link quality and throughput through dynamic adaptation. We present preliminary experimental results which show that MeshChop can achieve almost 80% improvement in throughput over non-adaptive schemes. We believe that component based adaptation using channel hopping is the right method to adapt to local interference conditions without causing network-wide changes.