Hybrid Mobile Robot Indoor Localization Using Large-Scale Metric-Topological Map

Abstract

Localization is an essential task for an autonomous mobile robot's indoor navigation. Based on the metric-topological environment modeling, a hybrid localization method, pose tracking in local metric map and global localization in topological map, are used in this paper to carry out accurate and robust localization in semi-structured indoor environment. In order to guarantee the effectiveness of global localization and also solve the kidnapped robot problem in dynamic environment with some unexpected collisions, a novel Normalized Topological Node Weights Filter (NTNWF) algorithm is presented, which makes mobile robot's localization in large-scale topological map with closing loops and symmetric configurations available and efficient in least motion steps. Experiment results implemented in both SmartROB-2 mobile robot and the simulation software designed specially for NTNWF algorithm show the hybrid localization method's validity and practicability.