A planner for All-Terrain Vehicles on unknown rough terrains based on the MPC paradigm and D*-like algorithm

A novel conceptual design of a planner for a mobile vehicle, operating on poorly traversable unknown rough terrains, is discussed. Finding a way to include a vehicle model into the planning stage, while coping with unknown or partially known terrains, is a challenging and rarely addressed optimization setup. The main advantages of a possible solution of such a problem would be twofold. First, the planner would give trajectories which are feasible to follow by the vehicle, which is not the case in many other state of the art planning algorithms especially for large vehicle speeds. Second, those trajectories would be the optimal ones in accordance to the current vehicle states and knowledge on its environment. We propose a solution based on an MPC planning paradigm, wherein the planner solves a constrained optimal control problem at each time instant using the current knowledge on the terrain, which is caught appropriately by an objective function. Solving an optimal control problem allows for the vehicle model being included into the planning stage, while the repeated optimization allows for taking continuously into account new terrain information. To deal with the information given beyond the sensor range and to guarantee reaching a given goal position, we have adopted a D∗-like algorithm for rough terrains being used as a cost-to-go term within the optimization setup.