Intelligent Vehicle Path Planning Based on Q-Learning Algorithm with Consideration of Smoothness

Q-learning usually carries out global path planning in grid environment, which is difficult to satisfy the requirements of vehicle dynamics in practice. In this paper, a local path planning for intelligent vehicle based on Q-learning algorithm is proposed. Firstly, the vehicle-road model is established. The information of lane boundary and lane center line is obtained by interpolation method, and the position relationship between main vehicle and environment vehicle is determined. Secondly, the variables that can reflect the driving state of the vehicle and the relationship with the surrounding vehicle position are determined to describe the vehicle states. According to the actual driving situation and the mechanical structure of the vehicle, the action that the vehicle can take is determined. Then, the reward and punishment function is designed through the optimization objectives: driving in the lane, no collision with the surrounding vehicles and so on. Finally, the vehicle path discrete sequence is obtained after training, it is necessary to smooth the path to ensure that the vehicle can track the planned path well. In this paper, the cubic spline interpolation is used to process the discrete path sequence into smooth continuous path. In order to verify the effectiveness of the planning method, the path tracking simulation is carried out on MATLAB/CarSim software. The results show that the path solved by this method can satisfy the requirements of vehicle dynamics.