A Distributed PHD Filter for On-line Joint Sensor Registration and Multi-target Tracking

Abstract

In this paper, we address the joint sensor registration and multi-target tracking problem using PHD filter and generalized Covariance Intersection (GCI) fusion rule. Each sensor is assumed to have no knowledge about the positions (referred to as the drift parameters) of its neighboring nodes in its local coordinate system. In our method, the drift parameters are estimated first, then the multi-target states are obtained based on the estimated drift parameters at each common global reference time step. The GCI divergence is used to measure the degree of discrepancy between multiple local posterior densities, and the drift parameters are obtained by minimizing the GCI divergence. On account of the optimization problem constructed by GCI divergence is a non-convex optimization problem when Gaussian mixture (GM) implementation is adopted, we provide a fast but effective optimization method on the basis of particle swarm optimization (PSO) algorithm. The efficiency of the proposed method is demonstrated in the numerical results.