Increasing autonomy on mobile robots is essential. The main reason is to provide them with the ability to perform operations on their own. This skill is desirable in fields such as planetary exploration or search and rescue operations. Mobile robots could carry out more tasks in a time window by not relying on human intervention. These systems could, for instance, drive longer distances and hence reach more places. Autonomous navigation rests on the use of path planning algorithms. These algorithms generate a path that guides the robot towards a target location. Yet, off-road and unstructured environments can pose a challenge to its locomotion capabilities. Thus, the path planner must address them at the same time as the robot mobility skills. In this way, this algorithm finds the path that minimizes a metric such as energy consumption. Not always precise information describing the environments is complete. The path planner must hence dynamically update the generated path whenever necessary. The robot may deal with terrain elements that were not addressed before in situ. This thesis tackles problems that arise in autonomous navigation on irregular terrains. As a first step, it presents an overview of the existing path planning algorithms. This overview focuses on those algorithms that are compatible with ground mobile robots. It serves to build a classification of the existing approaches. This classification rests on their functionality. In a few words, it tackles the different ways in which path planner model and process the environment and the robot mobility. This thesis puts the focus on some of them that are capable to produce the globally optimal path given a cost map. These are called PDE (Partial Derivative Equation) solving algorithms. How to adapt these algorithms to the autonomous navigation of irregular terrains is the cornerstone of this thesis. It deals first with the use of a path planner along with a reconfigurable robot.
Buscar en Dimension
