A risk field-based trajectory planning method for autonomous vehicles to enhance real-time performance, safety, and traffic efficiency

Addressing the issues of poor real-time performance and inadequate safety in trajectory planning for autonomous vehicles, this paper proposed a risk field-based trajectory planning method. In the path planning layer, an improved two-dimensional Gaussian distribution function was utilized to construct a risk field that incorporates road boundaries and obstacles. The constraints of the risk field determine the adaptive sampling region for path planning. When designing the cost function, a risk field term was introduced to consider path deviation, smoothness, and safety. Dynamic programming was employed to pre-plan the path with the minimum cost, followed by quadratic programming to refine and obtain the final path. In the speed planning layer, the adaptive sampling region is determined by combining the risk field constraints, kinematic constraints, and speed constraints. A cost function incorporating a risk field term was designed to account for driving efficiency, comfort, and safety. The combination of dynamic and quadratic programming methods was used to achieve the final speed planning outcome. Simulation and driver-in-the-loop experimental results show that the proposed method significantly enhances the real-time performance, safety, and traffic efficiency of trajectory planning, enabling the generation of smooth trajectories even under emergency conditions.