FRRT*-Connect: A Bidirectional Sampling-Based Path Planner with Potential Field Guidance for Complex Obstacle Environments

This paper addresses the path planning problem in high-dimensional complex environments and proposes an improved FRRT*-Connect algorithm to enhance the efficiency, precision, and robustness of path generation. The algorithm first introduces a goal-directed attractive force control mechanism, integrating artificial potential field methods to guide the tree expansion more effectively toward the goal, thereby reducing redundant sampling and significantly improving convergence speed. Secondly, an adaptive step-size strategy is proposed, dynamically adjusting the tree expansion step size based on the complexity of the environment, which enhances the algorithm's adaptability in narrow passages and complex topological structures, effectively avoiding local minima. The results show that, compared to the RRT*-Connect algorithm, the proposed method exhibits significant advantages in path quality, convergence efficiency, and success rate: the average path length is reduced by 19.7%, convergence speed is improved by 58.4%, and the success rate reaches 98% in narrow passage scenarios. These improvements effectively overcome the issues of path redundancy, slow convergence, and local minima inherent in traditional RRT-based algorithms, demonstrating superior performance in challenging scenarios with complex obstacles and narrow passages.