Research on Simultaneous Location and Mapping Algorithm of Intelligent vehicle Based on Improved Particle Filter Resampling

Abstract

For the estimation problem that the Simultaneous Location and Mapping algorithm used in Intelligent vehicle suffers from sample impoverishment, a Multi-threshold differential evolutionary particle resampling method is proposed. The method adopts different particle updating Strategys according to different thresholds of effective particle numbers. particle diversity will be rapidly improved by differential evolution when particle degradation is severe. When the particle degradation is light, the system resampling method is directly used to copy the large weight particles and eliminate the small weight particles. By adopting different particle regeneration strategies for particles with different degrees of degradation, the problem of particle degradation and depletion in the process of state estimation can be effectively solved. Simulation experiments show that the root mean square error of the improved algorithm pose is reduced by 2.592m compared with the traditional algorithm. The road sign estimation root mean square error is also reduced by 2.428m compared with the traditional algorithm. Experiments show that the multi-threshold differential particle filter resampling method can effectively improve the particle diversity and ensure the state estimation accuracy of the vehicle.