A Model based Path Planning Algorithm for Self-driving Cars in Dynamic Environment

Self-driving cars require robust and fast path planning algorithms to operate in dynamic environment. In the last years, model based path planning has emerged as an effective solution to real-time path planning taking into account the kinematics of the vehicle. However, this approach only validates in simple environment such as static or single moving obstacle. In this paper, we propose a model based path planning algorithm for dynamic environment. The approach generates candidate trajectories online, then evaluates and selects the most appropriate one according to real-time environment information. The selected trajectory can be directly tracked until emergency occurs and replanning starts. Thus, the planning algorithm is anytime and dynamic, and also effective in complex multi-obstacle environment. Simulation results of different scenarios show the validity of the proposed algorithm.