A localization based on a space observation model in a crowded environment


This paper describes a localization technique for a mobile robot in an environment with many unknown obstacles, such as pedestrians. To realize robust localization against unknown obstacles by using a particle filter, a free-space observation model and an area-observation model have previously been proposed. Although these localization methods are very effective, both methods evaluate the likelihood by using only the free-space model. Thus, in the environment where only the left or the right side is open, the likelihood of a particle separated from a fixed obstacle cannot be lowered. For this reason, the particle continues to spread and cannot converge when the fixed obstacle is observed on only one side. In the present research, we solve these problems by two likelihood evaluations based on the free space and the occupied space. We evaluate the robustness and verify the validity of the proposed method by a simulation and an experiment in a real environment.

Proceedings of the 2012 IEEE/ASME International Conference on Advanced Intelligent Mechatronics (AIM)