An Improved RBPF-SLAM Algorithm for Indoor Robot

Aiming at the problems of inaccurate proposal distribution and particle degradation of traditional RPF-SLAM algorithm in robot indoor mapping, an improved RPF-SLAM algorithm was proposed. Firstly, the motion model and observation model were combined as the mixed suggestion distribution. Annealing parameters were used to optimize the mixed suggestion distribution and adjust the proportion of the two in the suggestion distribution to make the mixed suggestion distribution more accurate. Secondly, a Markov chain Monte Carlo (MCMC) resampling method is used to generate reasonable samples from the target distribution by constructing Markov chains, so that particles can move to different positions in each iteration to avoid particle degradation. Experimental results show that compared with the traditional RBPF-SLAM algorithm, the improved algorithm has lower root mean square error and calculation time, higher accuracy and better effect of map construction.