Objective Unmanned surface vehicles (USVs) operating in complex marine environments face significant challenges due to the presence of both static and dynamic obstacles. The integration of global path planning algorithms such as the rapidly-exploring random tree star (RRT) with local reactive planners like the dynamic window approach (DWA) has been widely adopted to achieve reliable navigation. However, existing RRT*−DWA fusion frameworks often suffer from high path node redundancy, which leads to frequent local replanning, poor path smoothness, and compromised real-time performance and energy efficiency. This study aims to address these limitations by proposing an improved path planning method that enhances the dynamic obstacle avoidance capability and navigation economy of USVs.

Method To realistically simulate USV motion, a three-degree-of-freedom kinematic model incorporating ocean current disturbances is first established. The simulation environment is enriched with dynamic obstacles modeled as intelligent agents capable of trajectory prediction and collision avoidance behaviors, reflecting realistic maritime scenarios. Building upon the conventional RRT*−DWA fusion framework, we introduce an adaptive improvement to the classic greedy pruning post-processing technique tailored for USV dynamic environments. Specifically, a maximum connection distance constraint is incorporated to ensure safe navigation through narrow channels while preserving path connectivity. This constraint prevents the pruning process from generating unsafe direct connections between distant nodes, thereby reducing path redundancy without compromising safety. The effectiveness of the proposed method in mitigating the core issue of path redundancy is systematically verified through comparative simulations. Furthermore, extensive experiments are designed to benchmark the improved algorithm against the original RRT*−DWA, pure local DWA, and other mainstream dynamic planning algorithms such as A* with DWA and artificial potential field methods. Evaluation metrics include computational efficiency, path quality, local replanning frequency, and obstacle avoidance performance.

Results Quantitative results demonstrate that the improved RRT* algorithm significantly outperforms the original version in dynamic scenario path planning. Specifically, it reduces the average runtime by 82.21%, the average path length by 5.66%, the number of path nodes by 81.67%, and the average path curvature by 40.2%. When integrated into the navigation system, these improvements lead to a 1.78% reduction in the actual USV navigation distance for typical missions, indicating enhanced fuel economy. Compared with the original RRT*−DWA framework, the proposed method reduces the frequency of local replanning by approximately 65% and the average computation time per control cycle by 70% under identical scenarios. Importantly, the USV consistently maintains a safe distance from dynamic obstacles, demonstrating robust collision avoidance. Additional comparisons with APF−DWA and Informed RRT*−DWA reveal that the improved algorithm achieves a superior balance between global optimality and local reactivity.

Conclusion The proposed improved RRT*−DWA algorithm based on adaptive greedy pruning effectively resolves the real-time and smoothness issues caused by path redundancy in the original framework. It significantly enhances the obstacle avoidance capability, planning efficiency, and navigation economy of USVs in dynamic and uncertain marine environments. The incorporation of the maximum connection distance constraint ensures safety in complex scenarios, making the algorithm a practical and effective solution for autonomous USV navigation. This work provides a solid foundation for future research on multi-objective path planning and real-world deployment of USVs in challenging ocean conditions.