Improvement of Gaussian Resampling Algorithm in Particle Filter Algorithm

The improved resampling algorithms commonly in particle filter (PF), increase particles’ diversity by making new particles with various methods, and thus improve PF’s accuracy. However, they also increase the distance of particle probability distribution before resampling and reduce theactual estimation accuracy. To solve this problem, thispaper proposes an improved Gaussian resampling(IGR) algorithm, based on Gaussian Resampling (GR) Algorithm. Under the premise of maintaining the diversity of particles, we enable new particles to contain part of the low-weighted particles’ information by conducting proper linear combination with low-weighted particles. Simulation experiments conducted on single variable non-growth model suggest that the improved algorithm reduces the particles’ Kullback-Leibler(K-L) distance and improves the final tracking accuracy of PF.