Unscented Particle Filter for Bearings-only Tracking with Out-of-Sequence Measurements in Sensor Networks

An out-of-sequence measurements (OOSMs) processing algorithm for improving the passive tracking performance in wireless sensor networks is proposed. Firstly, decentralized tracking structure is organized by the dynamic clustering of sensor nodes, and cluster heads collect the measurements from child nodes to form the local estimate. Then, particle filter scheme is presented to solve OOSM problem in this decentralized structure. Due to the limited exploration capability of proposal density, unscented particle filter (UPF) is used to incorporate the most current measurement and to generate the proposal distribution of the particle filter. The detailed implementation steps of OOSM based on UPF (OOSM-UPF) are deduced. Finally, the bearings-only tracking state space is modeled by turn rate model, and 3D simulation scenario is constructed to test several filters for OOSM. Simulation results show that performance of OOSM-UPF is much improved than other schemes