Trajectory Planning of Autonomous Vehicles Based on Parameterized Control Optimization for Three-Degree-of-Freedom Vehicle Dynamics Model

In contemporary trajectory planning research, it is common to rely on point-mass model for trajectory planning. However, this often leads to the generation of trajectories that do not adhere to the vehicle dynamics, thereby increasing the complexity of trajectory tracking control. This paper proposes a local trajectory planning algorithm that combines sampling and sequential quadratic optimization, considering the vehicle dynamics model. Initially, the vehicle trajectory is characterized by utilizing vehicle dynamic control variables, including the front wheel angle and the longitudinal speed. Next, a cluster of sampling points for the anticipated point corresponding to the current vehicle position is obtained through a sampling algorithm based on the vehicle's current state. Then, the trajectory planning problem between these two points is modeled as a sequential quadratic optimization problem. By employing an offline method, the optimal trajectory set between the present position and the anticipated point cluster is computed. After acquiring clusters of candidate trajectories, each candidate trajectory is evaluated to determine its feasibility and cost, considering factors such as efficiency and comfort. The best trajectory is then chosen as the local trajectory of vehicle. The trajectories generated using the proposed method and the quintic polynomial method are both tracked and controlled using a 3-DOF vehicle dynamic model. The results clearly demonstrate that the trajectories generated by the proposed method exhibit superior tracking performance.