An Improved Particle Filter Based on Bird Swarm Algorithm

Particle Filter is well suited for strong nonlinear and non-Gaussian noise problem with respect to traditional Kalman filter, extended Kalman filter and unscented Kalman filter. However, problems such as particle depletion and particle degradation affect the performance of the particle filters. Optimizing the particle set to the high likelihood regions results in a more reasonable distribution of the sampling particles and more accurate state estimation. In this paper, a novel particle filter algorithm is presented based on the newly proposed bird swarm algorithm. Different behavior models are established by emulating the predation, flight, vigilance and follower behavior of the birds. The observation information is introduced into the generation process of the proposal distribution with the design of fitness function. In addition, to prevent particles from getting premature, being stuck into the local optimum and to increase the diversity of particles, levy flight is designed to increases the randomness of particle's movement. The proposed algorithm is applied to estimate the speed of the train under the condition that the noise of the axle speed is non-Gaussian. Simulation study and experimental results shows that the proposed algorithm is more accurate and has more effective particles compared to standard particle filter.