Skip to Main Content
Sensing reliability is an important issue in wireless sensor networks because of the generally existing failures of sensor nodes. In this paper, we design a reputation-based method to assure sensing reliability for acoustic target location estimation. Sensor reputation is defined here as the probability that a node's measurement is trustable, i.e., reputation actually represents the sensing reliability of a node. Our method has two phases, namely local estimation (LE) and global estimation (GE). In LE phase, sink node randomly selects subsets of sensor nodes and roughly localizes the target for each drawn subset by a simple reputation-weighted approach. And then in GE phase, LE results are examined by the fast minimum covariance determinant estimator (fast-MCD) to filter unreliable results and accurately evaluate the target location. The measurement error of each sensor then can be estimated and used to update node reputation by the Dirichlet process. We evaluate the performance of the proposed method by stimulation. The results verify that our method can effectively reduce the impact of unreliable measurements in the presence of failures and improve the estimation accuracy of target location.