TOA Based Indoor Localization and Tracking via Single-Cluster PHD Filtering

This paper addresses the issue of target tracking in indoor multipath environment using Time-of- Arrival (TOA) measured at multiple reference devices (RDs) with the assistance of inaccurate floor plane map. Due to the inaccuracy, the map features are estimated jointly with the target state, which is called map assisted simultaneous localization and mapping (MASLAM). To solve the problem, the multisensor single cluster probability hypothesis density (SC-PHD) filter is used to recursively estimate the state of mobile device (MD) and map features. In indoor environment, the received signal at RD consists of line-of-sight (LOS) path, single reflections, multiple reflections and diffraction paths. By treating the multiple reflections and diffraction paths as clutter, and modelling the LOS path and single reflections separately, a new update formula of multisensor SC-PHD filter for map features is derived. In addition, to mitigate the extremely high computation load of multisensor SC-PHD filter, a data association (DA) process is introduced. Furthermore, a particle filter-Gaussian mixture (PF-GM) implementation of the proposed multisensor SC-PHD filter is presented. Results based on TOA measurement data in the typical office area with 3 RDs show that the multisensor SC-PHD MA-SLAM scheme achieves an excellent target tracking performance with 90% of the estimation error smaller than 0.5m.