Direction-dependent optimal path planning for autonomous vehicles

Abstract

The optimal path planning problem is considered for rovers. Tip-over risk is accurately modelled using direction dependence. In the previous direction-independent model, the value function was approximated using the Fast Marching Method (FMM). The risk was not accurately modelled. Solar energy is considered here for the first time. Minimizing path length, obstacle avoidance and soil risk are also considered. For a direction-dependent model, the value function in the optimal path planning problem can be approximated accurately using the Ordered Upwind Method (OUM) but not FMM. The value function is used to synthesize the optimal control, which is shown to have no local minima. A novel algorithmic improvement, OUM-BD over the OUM to include a bi-directional search is presented. The OUM-BD is slightly slower than the FMM, but can accurately solve a larger class of problems. The OUM-BD is faster than the existing OUM, an optimal bi-directional RRT path planner (Bi-RRT*), and a genetic algorithm (GA) path planner in terms of time, and outperforms both the GA and Bi-RRT* planner in cost in tested examples.