Nonparametric Multitarget Data Association and Tracking for Multistatic Radars

Abstract

Multistatic radar systems provide better detection performance for stealth airborne platforms and are resilient to single-point failures. However, when multiple targets are present over the radar surveillance region, incorrect target associations to the measurements could create ghost targets. Computationally efficient and accurate de-ghosting and tracking multiple targets are critical tasks in real-time distributed radar systems. By exploiting the geometry of the measurement model in the association process, we propose a novel and efficient data association approach followed by a tracking algorithm in this work. It utilizes the time-of-arrival and bistatic Doppler frequency measurements of the targets with respect to different transmitter–receiver pairs to accurately determine and track the 3-D positions and velocities of the targets. The proposed approach is nonparametric as it does not need any assumption on the initial states or the number of targets and their motion models, but only uses the knowledge of the geometry of the terrestrial radar sensors. This nonparametric data association and tracking (NPDAT) algorithm is tested with multiple targets in two significant scenarios. First, all the targets are simultaneously present in the region, and then, targets arrive and depart the region based on a random arrival pattern. Our approach precisely tracks targets even during crossover and also tracks fast-maneuvering targets. This NPDAT algorithm is compared with popular existing methods and is shown to exhibit superior performance in estimation accuracy and maneuvering target tracking ability, even while enjoying a significantly lower time and implementation complexity.