Simultaneous registration and estimation for distributed non-uniformly sampled sensor network

Abstract

In practical multi-sensor tracking systems, target states are estimated based on measurements which are often obtained non-uniformly from each local sensor, and the system bias widely exists in these measurements. However, conventional methods can only deal with the two issues respectively. Based on an expectation-maximization (EM) algorithm and improved Kalman filter, a novel algorithm is proposed to solve the two problems together for the first time. In each local sensor, sensor bias registration and target state estimation are performed with EM at each measurement sampled time. With the non-uniform filtering result at previous time, extrapolation and innovation correction are performed at the uniform system time. Finally, these local target states at the global system time are transmitted and fused in the fusion center. The simulation shows that the proposed method is effective and reliable for the practical distributed sensor network.