An Adaptive Unscented Particle Filter for Tracking Ground Maneuvering Target

Abstract

Ground maneuvering target tracking is a linear/ nonlinear and Gaussian/non-Gaussian filtering problem. The particle filter (PF), which is not restricted by assumptions of linearity and Gaussian noise, is an optimal estimator to address such problems. Based on the particle filter, a filtering method which uses an Unscented Kalman Filter (UKF) to generate the mean and covariance of the importance proposal distribution is developed. To reduce the computational burden, a resampling controller is designed to adjust the number of particles according to the filtering performance in the different maneuvering stages. Simulation results demonstrate that the new adaptive filtering method can obtain almost the same tracking performance with that of the UPF using fewer particles in the non-maneuvering phase and achieves more accuracy with more particles in the maneuvering phase.