Improved Particle Filter Based On The Grey Wolf Optimizer

Particle Filter (PF) is a complex model estimation technique based on Monte Calo simulation. The resampling procedure of particle filter suppresses particle degradation. However, it leads to loss of particle diversity and limits estimation accuracy. In order to alleviate these problems, Grey Wolf Optimizer (GWO) is combined with PF in this paper. The specific strategy is to treat the particle set obtained from the importance sampling as a wolf group and the particle weight as the fitness of the grey wolf. By GWO, the particle with the smaller weight is moved to the position where a larger weight can be obtained. The importance sampling process actually optimizes the particle distribution so that it can obtain more accurate state estimation. The simulations results show that the proposed method exhibits better performance than the traditional PF, Unscented Kalman filter (UKF) and Unscented PF. By adjusting the particle number and the iterations number of GWO, the proposed method can balance the accuracy and efficiency.