A Novel Disturbance Particle Filtering Algorithm for Reduced Degeneracy and Improved Tracking Accuracy

Particle filtering is a Bayesian technique used for tracking moving targets in nonlinear dynamic systems. However, the resampling step in this filter is computationally intensive, limiting its application in real-time tracking. This paper introduces a methodology that eliminates the need for resampling while effectively parallelizing the particle filter. The approach involves iteratively sampling multiple heading disturbances for each particle until unbiasedness is achieved. This method is highly suitable for parallel processing since all particles are handled in batches, rather than sequentially as in the resampling process. The benefits of the proposed technique are demonstrated through a simulated study.