重庆理工大学学报2025，Vol.39Issue(17)：13-22,10.DOI:10.3969/j.issn.1674-8425(z).2025.09.002
改进RRT*算法在复杂环境下的路径规划研究
Improved RRT* algorithm for path planning in complex environments
摘要
Abstract
With the rapid development of autonomous driving technologies,path planning is key to intelligent vehicles.Path planning algorithms are the very foundations for safe,efficient,and reliable vehicle operation in highly complex environments.Among the numerous approaches,the Rapidly-exploring Random Tree Star(RRT*)algorithm stands out as one of the most representative sampling-based planners.Thanks to its probabilistic completeness and asymptotic optimality,RRT*plays a critical role in vehicle path search,trajectory generation,and motion planning tasks,paving the way for higher-level autonomous driving functions. Although RRT*offers several advantages,it still exhibits some limitations in real-world scenarios.First,reliance on purely random sampling often generates a large proportion of nodes within obstacle regions,which leads to ineffective exploration,excessive computational overhead,and prolonged planning time.Second,the raw paths that RRT*generates frequently redundant nodes and abrupt curvature changes,producing trajectories not suitable for direct execution by vehicles.These compromise path smoothness,stability,and controllability—essential requirements for safe and comfortable autonomous navigation.Therefore,improving both computational efficiency and trajectory quality remains a critical challenge for advanced path planning of intelligent vehicles. To address these problems,this paper proposes an enhanced algorithm,ADBI-RRT*(Artificial Potential Field and Double-tree Bidirectional Improved RRT*),which introduces multiple complementary strategies that jointly improve sampling efficiency,accelerate convergence,and generate high-quality trajectories.The first strategy employs a goal-biased sampling mechanism,which increases the likelihood of generating nodes near the target region,reduces unnecessary randomness,guides the tree toward promising areas,and improves overall convergence speed.The second one integrates an improved artificial potential field that applies directed perturbations aligned with the goal orientation.This design strengthens goal-directed search behavior and enables the algorithm to escape local optima efficiently,thereby enhancing robustness in cluttered and complex environments.The third one adopts a bidirectional growth framework in which two trees expand simultaneously at the start and goal regions.When their distance falls below a predefined threshold with no obstacles between them,the trees connect directly.This approach improves connection efficiency,reduces unnecessary iterations,and markedly cut the overall search time.After the algorithm obtains the initial feasible path,which is further refined to improve the quality of the trajectory and meet the needs of driving. The algorithm applies the triangulation principle to remove redundant nodes,reducing curvature fluctuation and path length.Then it further smooths the simplified trajectory by combining linear interpolation and B-spline curves,which improves the smoothness of the path and ensures the controllability of the path.By integrating these improvements,ADBI-RRT*improves computational efficiency and trajectory quality,reduces the cost of vehicle traveling,enhances the smoothness and the ride comfort of the vehicle and its traveling transmission.Finally,extensive simulation experiments in MATLAB evaluate the performance of ADBI-RRT*under environments with varied obstacle densities,distributions,and complexities.The comparative analysis considers the traditional RRT,RRT*,and another representative improved variant.Evaluation metrics include path generation time,number of iterations,total path length,and path smoothness.Compared with other methods,ADBI-RRT*cuts path generation time,reduces the number of iterations,and produces shorter trajectories.Moreover,the refined paths maintain smoother curvature transitions,fully meeting driving requirements in different scenarios.关键词
路径规划/ADBI-RRT*算法/目标偏置/改进人工势场/距离阈值Key words
path planning/ADBI-RRT* algorithm/target bias/artificial potential field/distance threshold分类
信息技术与安全科学引用本文复制引用
韩毅,孔米闯,李建庆,秦瑞泽,姚静彤..改进RRT*算法在复杂环境下的路径规划研究[J].重庆理工大学学报,2025,39(17):13-22,10.基金项目
陕西省西安市西咸新区科技计划-秦创原总窗口揭榜挂帅项目(2022-ZDJS-002) （2022-ZDJS-002）
