Developing a Novel Path Planning Algorithm for an Autonomous Experimental Rover Platform

Sammanfattning

This thesis introduces a novel and computationally efficient, random-search based path planner as a local path planner for the experimental planetary platform Sawppy. The Sawppy platform utilizes an on-board 3D Light Detection And Ranging (LIDAR) system and relies on a locally computed 3D Voxel Map to 2D Occupancy Map Converter (3DV-2DO) to provide a binary occupancy grid for path planning. A quadtree-based cell-decomposition method is employed by the planner to identify convex obstacle-free regions which are utilized as sampling candidates by the underlying Rapidly Random Tree (RRT)* algorithm, successfully guiding it towards safer paths with an increased exploration speed with results comparable to regular RRT* at a significantly faster rate. In addition to the planner, the developed planner-system bridge enables seamless integration of the Python-based planner and the live operating system of the rover platform and handles Robot Operating System (ROS) topic communication to from the traversability mapper, to the hardware controller responsible to steer the platform. Both the planner and the planner-system bridge successfully completed hardware in loop test using the Sawppy platform. While improvements on both the software and hardware remain, the rover’s capability to perform autonomous navigation and exploration, and the proposed path planner’s capability of handling real navigational tasks, is successfully demonstrated.