Trajectory Optimization of Autonomous Driving by Differential Dynamic Programming

We present a real-time optimization method for autonomous vehicle motion planning. Trajectory optimization is essential for autonomous vehicle motion planning, due to its non-holonomic motion. Conventional way of optimization is by trajectory smoothing. In this paper, we use differential dynamic programming (DDP) to optimize the trajectory. The key advantage of DDP is that it not only smoothes the trajectory but also considers the non-holonomic constrain of the vehicle. The optimized trajectory is pre-proof to be able to executed by the vehicle. Results are demonstrated with an autonomous vehicle model in Gazebo simulation environment.