An Improved Particle Filter SLAM Algorithm for AGVs

Abstract

Aiming at the problems of the traditional Particle Filter (PF) algorithm used in robot localization technology, for example, large computational expense, poor real-time performance, and limited positioning accuracy, an IPF-SLAM (Improved Particle Filtering SLAM Simultaneous Localization and Mapping) algorithm is proposed to tackle these difficulties. First, an interactive multi-model extended Kalman filter is used to provide a proposed distribution for particle filtering. The degree of fitting of Kalman filtering to a nonlinear system is improved by the multi-models, so that the filtering result is closer to the true value. Then, the “number of effective particles” is employed to determine the resampling timing and reduce the number of resampling. A Gaussian distribution function is introduced to randomly generate replicated particles to alleviate particle degradation. The simulation results show that the location error of IPF-SLAM algorithm is 17.26% lower than that of RBPF-SLAM (Rao-Blackwellise Particle Filter-Simultaneous Localization and Mapping) algorithm, and the calculation time is 5.7% lower. The experimental results show that the traditional algorithm is significantly improved in reducing computational complexity, improving positioning accuracy and robustness, etc. Therefore, the IPF-SLAM has a more significant positioning and mapping effects, compared with the traditional RBPF-SLAM algorithm.