Minimax Rao-Blackwellized Particle Filtering in 2D LIDAR SLAM

The localization of a robot is an important problem for accurate mapping via the mobile robot. In this paper, we propose minimax particle filtering (MPF) for the pose estimation of a mobile robot in the problem of simultaneous localization and mapping (SLAM) based on 2D LIDAR scans. Standard particle filtering has generic drawbacks such as particle degeneracy and particle impoverishment issues that cause the degraded quality of particles and result in unsatisfactory performance in practice while particle filtering is expected to show optimal performance in nonlinear problems, theoretically. Recently proposed MPF overcomes these limitations in PF implementation with increased quality of the particles in terms of particle diversity and variance of the weights. We test the proposed SLAM algorithm based on MPF with the datasets that were used for testing the standard Rao-Blackwellized PF (RBPF) SLAM and show the outperforming results, particularly in terms of the maximum translational/rotational errors that result in the overall diminished average errors.