Evaluation of selected localization and mapping techniques for the Boston Dynamics Spot robot

Sammanfattning

This study evaluates the accuracy of different mapping and localization techniques for the Boston Dynamics Spot robot in various environments. The motivation comes from Spot's built-in Simultaneous Localization and Mapping (SLAM) and autonomous traverse system, GraphNav, which requires a user to manually map the environment using Spot. The research partly aims to explore whether integrating an open-source SLAM technique such as SLAM Toolbox can improve Spot's utility in missions like search and rescue, high-risk construction and inspection missions. The study was done in collaboration with the Artificial Intelligence and Integrated Computer Systems (AIICS) department at Linköping University, who lately are focusing on applications between autonomous robotic systems like drones, ground vehicles and Spot for public safety. Through indoor and outdoor experiments, this study investigates the limitations of GraphNav and compares that with open-source SLAM techniques, specifically SLAM Toolbox. The results of this study reveal that GraphNav's navigation is reliable indoors using waypoint anchoring with an accurate ground truth (Vicon), but show limitations outdoors due to GPS inaccuracies. SLAM Toolbox using 2D laser scans did not prove to be better than GraphNav in terms of accuracy for this study, but further research is needed. This study highlights the need for further research to improve Spot's autonomous capabilities, with for example another GPS setup and different terrain.