Simultaneous localization and mapping and autonomous navigation on SPOT robot from Boston Dynamics

Achieving autonomous navigation of robots can open doors to many applications. Boston Dynamics has developed the Spot robot to enable robust, safe, and agile movement over many types of terrain. However, Spot’s navigation is only semi-autonomous – it needs an operator to create the path that the robot can then repeat on its own. This navigation method is not adaptive as it requires the operator to repeat the initialization step should Spot need to take another path. In a very dynamic and changing environment such as a construction site, this can become a problem. Indeed, it is necessary to change very regularly the predefined paths to maintain efficiency in the navigation, if not to avoid the robot to be lost. In addition, Spot doesn’t have a solution for mapping environments in a precise way and the methods available today for that purpose are slow and cumbersome. To improve the navigation and the possible applications of Spot (e.g., mapping, monitoring), we have developed a Simultaneous Localization and Mapping (SLAM) algorithm based on a graph optimization technique. It effectively uses the point clouds features received from the Spot’s LiDAR and detects loop-closures in the robot’s path to allow precise localization. This enables the mapping of the robot’s environment in a precise and real-time way. In addition, we have also developed a path-planning algorithm based on an improvement of A*: Jump Point Search. A filtering method of the computed path has also been built to transform the output of the navigation into an efficient set of instructions for the robot. Tests carried out on Spot in buildings demonstrate the effectiveness of the proposed solutions. Environments can be precisely reconstructed and the robot can navigate autonomously inside them. The need for an operator to construct predefined paths that Spot can take is removed.