Optimal Control and Planning for Autonomous Driving
Abstract
Safety is an emerging task in autonomous driving that encompasses Perception, Planning, and Decision Making to improve autonomy in all driving conditions, especially in urban environments where vehicles share the road with other cars and pedestrians. Several research activities have been conducted, and promising results have been achieved.
In this master thesis, we have focused on Trajectory planning and Execution task that enables our Amesim Car to overtake safely around predefined environment attempting to reduce the error between the planning and execution. A kinodynamic motion planner like-driver was developed to mimic the human driver actions and to provide us with an executable path.
In addition, an optimal trajectory controller was designed to stabilize the vehicle and track the reference under system constraints. The project combines an RRT-based kinodynamic planner with NMPC-style trajectory control and uses a Siemens Amesim vehicle model for evaluation.
Finally, co-simulations were performed for various scenarios: double lane change maneuvers and racing track tests with different control schemes.