A visual-assisted relocalization add-on for autonomous robots

We introduce a novel visual-assisted relocalization method for autonomous robots, enhancing both indoor and outdoor navigation. Recognizing the limitations of relying solely on light detection and ranging (LiDAR) sensors for simultaneous localization and mapping (SLAM), our approach integrates visual sensors to complement existing LiDAR-based systems without necessitating a complete overhaul. The proposed relocalization strategy, which treats both point and object features in image frames as scene descriptions, can be seamlessly integrated with multiple indoor–outdoor LiDAR-based SLAM algorithms. Establishing a link between these features and the robot’s poses transforms them into natural landmarks. Via an efficient feature-matching procedure, initial pose estimates can be obtained and provided to the primary LiDAR-based relocalization scheme to compute final pose estimates. The novelty of our method lies in its flexible add-on design, its compatibility with various LiDAR-based SLAM algorithms, and its ability to significantly boost the robustness and adaptability of robotic navigation. Extensive experiments across diverse environments confirm the performance superiority of the proposed strategy, demonstrating its practical value for real-world applications.