Low-overhead and High-accuracy Failure Detection Method for Wireless Multi-hop ad hoc Networks

One serious challenge of wireless multi-hop ad hoc networks is frequent link and node failures due to lossy wireless channels and low-cost hardware. Failure detection plays an important role in network maintenance. In order to reduce communication overhead, local-based failure detection methods conduct diagnosis processes in a local area. However, the overhead of them is still very large because several redundant detection processes are triggered for one failure by several nodes. In this paper, we present a new failure detection approach named Root Node Reduction (RNR), which can reduce the number of detection processes by a backoff scheme. Furthermore, we propose two additional strategies to improve diagnosis accuracy, including multi fusion which guarantees complete evidence collection, and multi report which improves the reception ratio of diagnosis conclusion report. The comparison results show that the performance of communication overhead, diagnosis accuracy and data packet reception is improved by these schemes.