Underwater Bearings-Only Multitarget Tracking Based on Modified PMHT in Dense-Cluttered Environment

Abstract

Underwater bearings-only multitarget tracking has many advantages, and the probabilistic multiple hypothesis tracking (PMHT) is an elegant algorithm for multitarget tracking problem. However, the basic PMHT has following weakness: 1) the targets' posterior probability may convergent to the local maximum, which would degrade the tracking accuracy and 2) the tracking performance is sensitive to the targets' initialization. In order to overcome these weaknesses, this paper proposes a modified PMHT algorithm. The key idea of the modified PMHT algorithm is that it allows the bearings measurements at one scan come from any Gaussian density, which has the same mean and different covariance with the same target. In addition, the modified PMHT algorithm uses the deterministic annealing to reduce the dependence on the targets' initialization. To deal with the nonlinear bearings measurements, the paper uses the unscented Kalman smoother to update the target states. The simulation treats the cross moving targets and closely spaced targets for both multiple stationary sensors and single maneuvering sensor scenarios in a dense clutter environment. The simulation results show the superiority of the modified PMHT algorithm over the basic PMHT algorithm respect to accuracy and robustness for underwater bearings-only multitarget tracking problem when setting a relatively bad initialization value.