Hybrid Strategy of Motion Planning with Kinematic Optimization for Autonomous Driving

Autonomous driving has revolutionized the conventional driving. In this paper, we present our motion planner for autonomous driving. Two main issues of motion planning are addressed in the paper: kinematical feasibility and consistence of motion control. The kinematical feasibility is achieved by a prediction engine, which generates a reference path for vehicle to follow based on vehicle kinematic model. The core of the engine is based on a differential dynamic programming (DDP) optimization. The consistence of motion control is solved by a hybrid motion planning strategy, which also considers the previous reference path in the motion planning. The proposed method is verified on our autonomous vehicle platform driving on a public road and performing vehicle avoidance.