Dynamic Perceptive Bat Algorithm Used to Optimize Particle Filter for Tracking Multiple Targets

Abstract

Resampling of standard particle filters will cause particle depletion and require abundant particles in the state estimation, which can hardly meet the accuracy and velocity requirements of a modern radar tracking system. This paper proposes an improved multiple-maneuvering-target tracking algorithm based on a novel intelligent particle filter. The improved algorithm combines the bat algorithm and particle filters and takes particles as bats to simulate behavior of bats in pursuit of prey. By adjusting frequency, volume, and pulse rate, particle groups search for the optimal value and move to high likelihood areas intelligently under the guidance of the optimal particle. Meanwhile, it improves the optimization mechanism of the bat algorithm; dynamic control of searching velocity and perception range are proposed. It makes the algorithm seek optimization within a self-adaptive cognition range, and the optimizing rate can be adjusted dynamically to control the balance of global and local optimizing abilities. Furthermore, the improved algorithm combines interacting multiple model and joint probabilistic data association, which enables improved accuracy in target tracking and robustness in a complex environment by iterative optimization. Simulation results show that the improved algorithm enhances the performance of a multiple-maneuvering-target tracking system.