Motion planning for passenger vehicles - force field trajectory optimization for automated driving

Kurzfassung

Trajectory planning is one of the key software modules to enable the development of advanced driver assistance sys tems or autonomously driving cars to further enhance traf fic safety. This paper presents a novel coupled trajectory planning approach for autonomous driving along a road. It uses a two-step algorithm to achieve integrated planning of path and velocity profile to best create feasible, collision free trajectories through dynamic traffic environments. The first step, the trajectory initialization, uses an A* based search in a combined search space that consists partly of command space and state space, to efficiently plan a tra jectory specifying position, velocity and time. The local search algorithm used for trajectory optimization is based on trajectory deforming force fields. It operates on a low dimensional search space in , , , but employs a depen dent higher dimensional space to account for the dynamic quantities of the motion model, (velocities, accelerations, jerk), as well as the extrapolated movements of obstacles. The approach is able to handle arbitrary road curvatures. The trajectory planning approach has been implemented and demonstrated in an autonomous test vehicle. First ex perimental results are shown and discussed.