Efficient and Robust Localization using Hybrid Environment Model

Localization, including pose tracking and global localization, is a fundamental technique for autonomous mobile robotics. It is the prerequisite and one of the most challenging problems for mobile robot autonomous navigation. Single type of sensor always contains insufficient information and the fusion of heterogeneous sensors information has been the main focus of environment perception. This paper first introduces a new type of hybrid environment model which integrates the camera and 2D-Lidar sensor information to build hybrid environment model, which includes the visual features map and occupancy grid map. Then an efficient global localization method is presented to obtain the initial pose of mobile robot and a pose tracking method is proposed to perform robust pose tracking. Besides, we also propose a re-localization triggering strategy to ensure the robustness of localization. The performance of proposed method is validated in real-world experiment in different scenarios.