Joint sequential target estimation and clock synchronization in wireless sensor networks

We propose a method to jointly estimate sequential target states and network synchronization status based on observations obtained by an unsynchronized wireless sensor network. We build an unsynchronized multi-sensor state-space model to connect asynchronous observations with target state transition. Under the built model, we solve the joint estimation problem via the expectation-maximum (EM) algorithm, assuming known temporal order of sensor clocks. Based on the solution and a Bayesian inference method developed to learn temporal order from observations, we solve the joint estimation problem in a distributed manner, assuming unknown temporal order. We use Monte Carlo methods to approximate our solutions, in order to account for nonlinear models and non-Gaussian noise. Moreover, we develop a recursive and parallel algorithm to compute the EM covariance matrix under Monte Carlo approximations. Numerical examples are presented to demonstrate the performance of the proposed method, and show that sequential target estimation benefits from the concurrent clock synchronization.