Angle-Only, Range-Only and Multistatic Tracking Based on GM-PHD Filter

Abstract

Multi-object detection and tracking with spatially distributed sensor networks are used in many applications across the domains of autonomous and surveillance systems. The sensors typically used in these systems often provide incomplete observations such as bistatic and angle- or range-only measurements, thus posing a challenge to the task of retrieving the targets and estimating their state. In this paper, we first present a new variant of a multi-sensor tracking algorithm based on the Gaussian-mixture probability hypothesis density (GM-PHD) filter. Next, we show how it can be applied on fusing incomplete observations. For tracking asynchronous range- and angle-only measurements, we leverage the well-known concepts of angle and range parametrization, respectively, to describe the adaptive target birth density based on the parameters of received observations. In the case of multistatic tracking, we propose parametrizing the birth density from target hypotheses, generated by statically fusing bistatic range measurements, using the M-best S-D assignment algorithm. We investigate the performance using challenging simulation scenarios and evaluate it with established tracking metrics. Our preliminary results demonstrate the effectiveness of the proposed algorithms. Furthermore, for range- and angle-only fusion, the more common use case of unsynchronized sensor measurements is supported. While many algorithms in the literature are tailored for a specific problem, we show that the proposed GM-PHD tracker is generic and can be potentially leveraged in a wide range of sensor fusion and tracking applications.