On the robustness of wireless opportunistic routing toward inaccurate link-level measurements

Abstract

The quality of wireless links is inherently dynamic, and this often makes the measurements of link delivery probabilities inaccurate over short timescales. We present Stable Opportunistic Routing (SOR), which improves unicast throughput for wireless mesh routing in the presence of inaccurate link-level measurements. In essence, SOR integrates two key features of prior approaches: (i) nodes trigger or suppress transmissions by inferring the actual reception of packets by neighboring nodes through channel overhearing, and (ii) nodes use network coding to avoid redundant transmissions. In addition, SOR is less dependent on accurate timing estimates or measured loss rates than prior approaches. Our stance is to argue that by carefully incorporating prior approaches into the design space of SOR, we can make opportunistic routing more robust toward link-level measurement errors, a practical issue in wireless mesh routing. Using nsclick simulation, we show that SOR has higher throughput than existing shortest-path and opportunistic routing protocols in large-scale networks, and the performance gain is more prominent when link-level measurements are erroneous.