Path planning for autonomous vehicles based on an improved PRM-A* fusion algorithm

This paper proposes an unmanned vehicle path planning algorithm based on a hybrid of an improved PRM and A* algorithm, providing a practical solution for unmanned vehicle path planning. The optimized PRM algorithm effectively addresses the issue of poor search directionality by optimizing the sampling space, while the bidirectional collision detection method resolves the problem of time-consuming construction of undirected path point graphs, thereby improving both spatial and temporal efficiency. By introducing an A* algorithm with extended search neighborhoods into the undirected path point graph for path search, this approach effectively resolves the issues of large node information storage requirements and low path search efficiency inherent in traditional single A* algorithms. Finally, cubic Bézier curves are used to improve the smoothness of the planned paths generated by the hybrid algorithm, resulting in optimal paths that better align with the vehicle’s dynamic characteristics and enhance driving safety and stability. Simulation results show that, in a 20 m × 20 m global environment map, the improved fusion algorithm reduces the optimal path length by 4.84% and 2.98% compared to the traditional A* algorithm and the commonly used PRM+A* algorithm, respectively, and shortens path search time by 4.28% and 1.09%, respectively and the number of turns by 36.36% and 30%, respectively. These metrics also showed significant improvements in a 30 m × 30 m global environment map. The data indicates that the algorithm proposed in this paper has distinct advantages over both the traditional A* algorithm and the commonly used PRM+A* algorithm, primarily in terms of obstacle avoidance capabilities in complex environments, path length, and the search efficiency of the fusion algorithm. Finally, the effectiveness and practical feasibility of the algorithm proposed in this paper were verified using a ROS-based experimental platform.