New Path Planner Methods for Planetary Explorations

In this thesis a new path planner method based on NURBS (Non Uniform Rational Basis-Splines) geometric curves is presented and applied in the context of planetary surface exploration. A review on robot navigation and path planner algorithms with particular focus on rover exploration techniques used by NASA and ESA is done. After that, NURBS theory and integration in a new path planner algorithm is explained in detail. Last, a comparison between classical path planner algorithms, like A*, and NURBS path planner is discussed, focusing on the advantages and results of the latter. The methodologies used are many, from NURBS mathematics, a kind of parametric curves with their properties, to computer vision and C++ algorithms. The designed path planner is composed of three different parts. The first is the obstacle detection algorithm that takes a top-view surface image (such as moon or mars surface) of the surrounded robot environment and detects the obstacles based on computer vision and perception algorithms. The second is the path planner algorithm based on NURBS mathematics and SISL (Sintef Spline Library) library, and last is a graphical viewer derived from OpenGL (Open Graphics Library) that renders obstacles, rover, and paths. This last one is also used for a dynamic simulation of the rover along the path with a continuously re-planning of the trajectory in the waypoints. Finally, results, simulation and tests are provided and explained. Comparison with A* is done and advantages of NURBS path planner are discussed. Advantages can be found in the peculiar properties of NURBS geometric curves. It is possible to define curves (i.e., rover trajectories) with less complexity and greater accuracy regardless of the number of obstacles, and few points can express and save the information of a NURBS. These properties will have benefits in terms of performance, usage memory, and smoother trajectories. This will result in having easier curve to manipulate since very low degree can be imposed independently, less memory is used to save the generated trajectory respect to mesh approximation, and smoother trajectories are achieved, that are optimal in the planetary exploration context.