An augmented state Gaussian mixture probability hypothesis density filter for multitarget tracking of autonomous underwater vehicles

Abstract

Multitarget tracking is an important task for autonomous underwater vehicles (AUVs). Due to the limited space on AUVs, the aperture of the sonar array mounted on an AUV is small. The finite resolution of AUV sonar systems results in weak targets being masked by strong targets when they are close in angular space, leading to the trajectory fracture problem of weak targets in the azimuth crossing scenario. To address this problem, an augmented state Gaussian mixture probability hypothesis density (GM-PHD) filter is proposed. Since the acoustic signals radiated by targets contain some narrowband discrete components (known simply as tonals or lines), we focus on leveraging the tonals to enhance tracking performance. First, we model the augmented state dynamics, which contain both direction-of-arrival (DOA) and tonal components, and DOA tracks can be generated by the augmented state GM-PHD filter. Second, track segments are clustered by the modified density-based spatial clustering of applications with noise (DBSCAN) algorithm according to the estimated tonals. Finally, track segments with the same clustering results are associated with the same target, and the track stitching method is applied to overcome the trajectory fracture problem. The effectiveness of the algorithm is verified by simulation experiments in different scenarios.