Automated driving vehicles have problems in local trajectory planning, such as insufficient consideration of vehicle handling stability, excessive simplification of vehicle models, and lack of objective evaluation of vehicle comfort. Considering the stability of vehicle handling, a three-degree-of-freedom model of vehicle is established. The lane changing scene of the self-driving car is simulated. The lane-changing trajectory is output by inputting the wheel angle，and the parameterization equation and driving trajectory characteristics of the vehicle are calculated. The BP neural network is used to identify the trajectory characteristics, and the change relationship between the lane change duration and lateral offset distance of the autonomous driving vehicle is obtained. At different lane changing speeds, according to different lane changing durations and lateral offset distances, the input wheel angles are used to obtain the lane change optimized by the trajectory clusters and steering stability parameters. Based on the conventional trajectory optimization method that only considers the driving efficiency and safety, the trajectory optimization objective function is constructed, by using the value of the yaw, roll, and lateral acceleration rate of the vehicle lane change process. A comprehensive trajectory optimization method based on driving efficiency, safety, comfort and steering stability is proposed. The trajectory optimization objective function is solved to obtain the optimal lane change trajectory. The joint simulation results show that the method is superior to the conventional trajectory optimization method, and the comfort and steering stability are improved by more than 20%.