Driving risk-aversive motion planning in off-road environment

Incremental sampling-based motion-planning algorithms, such as the rapidly-exploring random tree (RRT), are popular in motion planning because of their ability to proffer rapid solutions. Driving safety is the ability of a vehicle to prevent overturning and collision. However, the RRT algorithms neither involve any safety mechanism, nor consider any terrain measurement information of the environment, there is no assurance on the driving safety of the solutions. Therefore, obtaining an efficient trajectory while considering driving safety is a dilemma for intelligent vehicles in unstructured off-road environments. This study presents a potential-field based RRT* motion-planning algorithm for vehicle risk aversion. The algorithm constructs a configuration space using the potential field, which identifies risk range around obstacles and off-road terrains. A random-exploring tree grows through sampling inside the potential-field space with the consideration of nonholonomic constraints of the vehicles. A cost function that trades off driving safety and efficiency is used as a priority-sequence mechanism to get an optimized planning solution. A heuristic sampling method is employed to obtain a fast solution to the initial trajectory, and the trajectory can be optimized by decreasing the cost in the following sampling process thereafter. We tested the algorithm on scenarios to verify its effectiveness. Simulation results, under different scenarios, demonstrate that the proposed algorithm improves vehicles’ motion planning performance.