Skip to Main Content
Sensing reliability is an important issue in wireless sensor networks due to generally existing failures in sensor nodes. In this paper, we focus on the issue of sensing reliability in an acoustic target localization application, and a reputation-based algorithm is proposed to assure sensing reliability. Similar to the random sampling consensus (RANSAC), we select multiple subsets of nodes to roughly estimate target locations each time and choose the best subset receiving the most support from all nodes to compute the final target locations. The effectiveness and the efficiency of the algorithm are ensured by accurate estimation of the contamination ratio of node data. We define a node's reputation to indicate its sensing reliability, and the contamination ratio of the node can be represented by the reputation. As time passed, a node's reputation will gradually be evolved to the true value by exerting the Dirichlet process. In the experiments, we investigate several common failures in the sensor network. The results verify that the method can effectively detect unreliable measurements and improve the localization performance.