Research on Personalized Trajectory Decision-Making and Planning Methods for Intelligent Vehicles Based on Spatiotemporal Risk Indicators

In order to generate curved trajectories that satisfy drivers' personalized needs, this paper proposes a trajectory decision-making and planning method based on spatiotemporal risk indicators. Initially, a diverse trajectory planning approach is designed based on parametric curve fitting, enabling the generation of trajectories with significant variations. Subsequently, multiple spatiotemporal risk indicators are utilized to quantify drivers' risk perception characteristics and conduct risk assessments. Moreover, a spatiotemporal risk correlation model is developed based on neural network, establishing a nonlinear mapping between spatiotemporal risk indicators and trajectory planning parameters. This mapping facilitates the generation of trajectories that adhere to the specified spatiotemporal risk indicators. Finally, experiments are conducted with uniformly distributed spatiotemporal risk indicator combinations and an error analysis method is conducted to validate the proposed method. The testing criteria are established based on the variations in spatiotemporal risk indicators corresponding to different driving styles. Results demonstrate that the proposed trajectory decision-making and planning method satisfies the testing criteria and is capable of generating trajectories that meet various expected spatiotemporal risk indicators, thus achieving drivers' personalized needs.