A Global Localization Algorithm for Mobile Robots Based on Grid Submaps

Abstract

In this paper, we present a global localization approach for mobile robots equiped with 2D laser range finder(LRF), which combines depth-first search and grid submaps for computing scan-to-submap matches. For global localization, the maximum likelihood submap of the given scan obtained from the multi-level lookup tables, using the depth-first search method adopt the coarse-to-fine upper bound constraints. Then, a set of candidate poses and their matching scores are obtained, and the proposed algorithm could lead to the true pose convergence gradually through the mobile robot motion guided by topological information. Different from classical scan matching methods used in pose estimation and loop-closure, the proposed approach fuses the pose tracking and the global localization by using prior submaps. The mobile robot localization in current submap costs low memory, this makes the workspace easy to be expanded. Finally, the performance of the global localization algorithm was verified in experiments about the unique and similar scenes.