Intelligent Filter-Based Slam For Mobile Robots With Improved Localization Performance

Mingwei Lin,Canjun Yang,Dejun Li,Gengli Zhou
DOI: https://doi.org/10.1109/ACCESS.2019.2934995
IF: 3.9
2019-01-01
IEEE Access
Abstract:Fast simultaneous localization and mapping (FastSLAM) is one of the most popular methods for autonomous navigation of mobile robots. However, FastSLAM is essentially a particle filter (PF) that suffers from particle impoverishment and degeneracy problems. To improve its localization performance, this paper proposes an improved FastSLAM algorithm that contains an intelligent bat-inspired resampling whose iteration times can be adaptively tuned based on the degree of filter diverging. Additionally, the square root cubature filter is merged into the algorithm for better proposal distribution and mapping results. The advantages of the proposed method are verified by simulation and dataset-based tests. The test result demonstrates that the proposed IFastSLAM has better accuracy, computational efficiency and filter consistency compared to that of the square root unscented FastSLAM (SRUFastSLAM) and strong tracking square root central difference FastSLAM (STSRCDFastSLAM). Finally, a pool experiment is demonstrated to further verify the advantages of the proposed algorithm.
What problem does this paper attempt to address?