Multi-Vehicle Collaborative Trajectory Planning in Unstructured Conflict Areas Based on V-Hybrid A*

This work deals with the Multi-Vehicle Trajectory Planning (MVTP) problem in unstructured conflict areas. Compared with MVTP on structured roads, the more complex intersection and the more conflicting motions of vehicles, set higher requirements for the real-time of the planner. To address the issue, this work proposes a centralized decision-making distributed planning framework, to generate trajectories for multiple vehicles navigating conflict areas on unstructured roads. In the multi-vehicle collaboration process, vehicles are assigned a priority upon entering the control area, with higher-priority vehicles treated as dynamic obstacles to be avoided. The distributed planning is divided into three stages: 1) trajectory search utilizing Velocity-Hybrid A* (V-Hybrid A*), 2) path optimization with respect to both dynamic and static obstacles, and 3) speed optimization leveraging the convex space created by the initial solution. In the trajectory search of individual vehicles, each expansion node of V-Hybrid A* is given an acceleration to get a coarse trajectory with discrete velocity. Moreover, a box constraint is established to optimize the path of the coarse trajectory. Then, according to the homotopy class provided by the coarse trajectory, speed optimization is carried out to obtain the final single vehicle trajectory. In our simulation experiments, we randomized the generation of numerous vehicles assigned with varied tasks to evaluate the algorithm's effectiveness. Ultimately, real-vehicle experiments substantiated the algorithm's practical feasibility. The results indicate that the proposed method can significantly improve the traffic efficiency in conflict areas and has practical applicability.