A Restricted Resampling Algorithm of Particle Filter for Gmapping

Gmapping is a method that estimates the movement trajectory of a robot through particle filter and builds maps. Resampling is important process for Gmapping which can make effective particles replace ineffective particles. Selective resampling algorithm determines whether to perform resampling in gmapping by comparing the dispersion of normalized particle weights with a set threshold. But this method results in some ineffective particles being computed for a long time before resampling, which affects the mapping accuracy and stability. To address this issue, this paper proposes a restricted resampling algorithm to strengthen the weight of effective particles and weaken the weight of ineffective particles, and then compute the dispersion of normalized particle weights. Experimental results indicate that restricted resampling produces maps more accuracy and stability than the selective resampling, with clearer textures and outlines.