Rui Mao, Yingchao Shen, Yumei Yang, and Haijun Lou
Improved RRT algorithm, path planning, adaptive step-length,artificial potential fields
This paper proposes an enhanced Rapidly-exploring random tree (RRT) algorithm integrating potential field guidance and adaptive strategies to mitigate limitations of traditional RRT-based path planning, including excessive sampling randomness, low exploration efficiency, and inadequate path smoothness. First, a goal-biased probability threshold function prioritises sampling in target regions, while an adaptive step-length mechanism dynamically adjusts based on search states to reduce randomness and accelerate convergence. Then, a composite potential field–integrating target- attractive and obstacle-repulsive fields–guides node expansion via directional selection to minimise redundant exploration. In addition, a post-processing optimisation method employs Dijkstra-based redundant node pruning and n-th-order B´ezier curve smoothing to ensure path optimality and curvature continuity. Finally, validated through simulations and real-world experiments, the proposed algorithm significantly outperforms the comparative algorithms in path planning performance: achieving a 33.4% reduction in overall path length, a 61.5% decrease in explored nodes, and a substantial 97% reduction in average search time. In complex environments, it further reduced search time by 58.87%, path length by 31.2%, and explored nodes by 2.21%. Real-world experiments demonstrated a high success rate, with a trajectory dimensional standard deviation of merely 0.13, conclusively demonstrating its robustness, feasibility, and effectiveness in practical applications.
Important Links:
Go Back
