Simultaneous Target and Multipath Positioning via multi-hypothesis single-cluster PHD filtering

This work considers the problem of tracking a RF source in dense multipath environments using a uniform linear receiver array (ULA) where multipath propagation is modeled as specular reflections from planar reflectors. A single cluster process is formulated using a Bayesian estimation framework for Simultaneous Target and Multipath Positioning (STAMP) where target state is defined as a parent process and the reflector state is defined as the daughter process. A multi-hypothesis data association method is used with Probability Hypothesis Density (PHD) filtering to improve the accuracy of the target state estimate. The Gaussian target state is updated based on classic multiple-scan maximum likelihood data association while the update of multipath parameters is based conventional Gaussian Mixture PHD filtering. Experimental results using real data for an indoor target positioning problem demonstrates substantial improvements in localization accuracy with this method.