Multi-Sensor Control for Distributed Multi-Target Tracking with Limited Sensor Fields-of-view

This paper addresses the multi-sensor control problem for distributed multi-target tracking (DMTT) with limited sensor fields-of-view (FoVs). Existing control methods cannot be directly used when the FoVs of sensors are very limited. To solve this problem, we propose a new multi-sensor control strategy that contains two key points. One is to generate the FoV-based pseudo measurement set, which is also used to sift through control commands. The other one is to propose a new reward function composed of the Cauchy-Schwarz (CS) divergence between the fused predicted density and the fused posterior density and a penalty function related to the reduction of cardinality estimate. Based on it, the optimal control scheme is obtained by maximizing the reward function. Moreover, we employ the clustering algorithm generalized covariance intersection (CA-GCI) method to fuse the local densities for dealing with the different FoVs cases. Simulation results in a challenging multi-object tracking scenario with three radar-type sensors demonstrate effectiveness of the proposed solution.