@article{PENG2026, 
author = {Jinghui PENG and Li DONG and Lintao QU and Qingfeng LIU and Yuantao NI and Shizheng ZHAO},
title = {Improved RRT*−DWA with greedy pruning for USV path planning in dynamic obstacle environments},
year = {2026},
journal = {Chinese Journal of Ship Research},
volume = {21},
number = {5},
pages = {71-80},
keywords = {unmanned vehicles, RRT*−DWA, greedypruning, dynamic obstacle avoidance, path post-processing},
url = {https://www.sciopen.com/article/10.19693/j.issn.1673-3185.04813},
doi = {10.19693/j.issn.1673-3185.04813},
abstract = {ObjectiveUnmanned surface vehicles (USVs) face static-dynamic obstacle threats in complex seas. Widely-used RRT*-DWA navigation frameworks suffer severe path-node redundancy, leading to frequent replanning, poor smoothness and deteriorated real-time & energy performance. This paper presents an improved path-planning method to strengthen USV dynamic obstacle-avoidance and navigation economy.Method A three-degree-of-freedom USV kinematic model considering ocean-current disturbances is constructed, and intelligent-agent-based dynamic obstacles are introduced to reproduce real-world maritime conditions. On the basis of standard RRT*-DWA, an adaptive greedy-pruning post-processing strategy is developed, with a maximum-connection-distance constraint added to guarantee safety inside narrow channels while eliminating redundant nodes. Comparative simulations are carried out against baseline algorithms including vanilla RRT*-DWA, pure DWA and Informed RRT*-DWA. Computational efficiency, path quality, replanning frequency and obstacle-avoidance performance are adopted as evaluation metrics. ResultsQuantitative simulations show that the improved RRT* yields 82.21 % lower average runtime, 5.66 % shorter average path length, 81.67 % fewer path nodes and 39.76 % reduced average path curvature. Within the integrated navigation system, mission navigation distance drops by 1.78 %. Compared with original RRT*-DWA, the proposed method cuts local replanning frequency by 65 % and per-control-cycle computation time by 24.67 %, while keeping safe clearance from dynamic obstacles. Further comparisons with APF-DWA and Informed RRT*-DWA verify its better trade-off between global optimality and local reactivity. Conclusion The adaptive-greedy-pruning-based improved RRT*-DWA effectively addresses real-time and smoothness defects originating from path redundancy. It greatly improves USV obstacle-avoidance, planning efficiency and navigation economy for dynamic marine scenarios. The embedded maximum-connection-distance constraint guarantees operational safety, offering a feasible solution for USV autonomous navigation and laying a foundation for follow-up multi-objective planning and practical sea deployment.}
}