An improved artificial potential field with RRT star algorithm for autonomous vehicle path planning

Abstract

To address the issues of high sampling randomness, slow convergence speed, and insufficient path smoothness in traditional RRT* algorithm, this paper proposes a bidirectional APF-RRT* algorithm called BIAP-RRT*. First, a dynamic goal bias strategy is introduced to guide random sampling points towards the target direction, reducing ineffective sampling. Second, an improved artificial potential field method is incorporated to enhance the random tree's exploration capability, enabling it to quickly escape from local optima. Third, a dual-tree growth strategy is adopted with an improved tree connection mechanism to accelerate algorithm convergence. Fourth, the path is pruned according to the triangle inequality to shorten path length, while B-spline curves combined with linear interpolation are used to smooth the pruned path, improving path quality. Finally, through comparative analysis in different environments, the BIAP-RRT* algorithm shows significant advantages over traditional RRT algorithm, RRT* algorithm, and an existing improved algorithm in terms of convergence speed, number of iterations, and path smoothness.