Modified particle filter algorithm for mobile robot simultaneous localization and mapping

Simultaneous localization and mapping (SLAM) is an important topic in the autonomous mobile robot research. A modified Rao-Blackwellised particle filter (MRBPF) algorithm is proposed for the mobile robot to SLAM, which can simultaneously localize the robot and build up the map in the structured indoor environment. Firstly, MRBPF respectively uses particle filters (PF) to estimate the posterior probability distributions of robot postures and landmarks in the environment map. Secondly, it adapts the re-sampling process based on the effective sample size (ESS), and improves the computation methods of sample weights so as to guarantee MRBPF to have enough re-sampling numbers. Furthermore, a robust motion model and an observation model with only ranging sensor and odometer are constructed. Experimental results show that MRBPF-SLAM performs well on both weight variance and the number of effective samples. More over, the estimation accuracy of path and map is improved to some extent, and the simulation results also indicate that the methods are valid.