Source localization in large-scale asynchronous sensor networks

Abstract

The problem of uncooperative source localization and synchronization in asynchronous sensor networks has been considered previously in a centralized manner, where all the raw measurements are delivered to a processing center. However, for large-scale sensor networks the transmission of raw measurements over the networks requires considerable communication overhead, besides being vulnerable to transmission failure and interference, and the computational load at the processing center is rather heavy. A more suitable candidate to this problem is the distributed estimation considered in this paper, where each sensor node communicates with its neighboring nodes only, and the parameters are estimated at each sensor via information cooperation. In order to decouple the unknown source positions and sensor clock offsets, we propose to update them iteratively based on the belief propagation (BP) framework. To reduce the complexity of parameter update at each sensor, the sigma point based estimation method is adopted. Theoretical analyses concerning the computational complexity and communication overhead are carried out. Simulation results demonstrate that the proposed method could estimate the source positions distributively with much lower complexity and communication overhead compared with the centralized methods at the cost of acceptable performance degradation.