Abstract: Objectives In order to solve the problems such as low efficiency of path planning,node redundancy and non-shortest path of RRT* algorithm in complex navigation environment,a GBE-PS-RRT* algorithm(Goal-Bidirectional-Eliptic-Pruning-Smoothing-RRT*) is proposed,which combines elliptic restricted sampling,bidirectional search,goal bias extension and constraint sampling,redundant node pruning and path smoothing. Methods The sampling space is reduced by ellipse-restricted sampling,and the goal bias and sampling position constraints are introduced to make the node sampling more goal-oriented. When new nodes are generated,different weights are assigned to the sampling points and the goal points to guide new nodes to reach the goal points quickly. After the initial path is generated,the path length is optimized by redundant node pruning algorithm,and the path is smoothed by cubic B-spline curve. Finally,the planned path meets the actual navigation requirements. Results Through MATLAB simulation test,1000 repeated tests were done in each of the four complex navigation obstacle environments,and the comparison analysis was made with the GCSE-RRT* baseline algorithm from three dimensions,such as search efficiency,path length and node utilization. The results show that the proposed algorithm is more stable than the baseline algorithm in path planning,and its worst index value is better than the baseline algorithm,and the total average running time is shortened by 48.7%,the average path length is shortened by 9.5%,and the average node utilization is increased by 42.8%.To further verify the superiority of the algorithm, a comparative experimental analysis was conducted between the algorithm and the APF-RRT algorithm in both simple and complex Marine environments. The results show that in the simple sea area environment, the algorithm in this paper reduces the path length by 9.5% and the number of nodes by 58.3% compared with the APF-RRT algorithm. In the complex Marine environment, the algorithm in this paper reduces the path length by 5% and the number of nodes by 18.2% compared with the APF-RRT algorithm. Conclusions The GBE-PS-RRT* algorithm proposed in this study has significant advantages in improving the efficiency and quality of path planning for unmanned ships, especially demonstrating stronger adaptability and stability in complex navigation environments. The algorithm effectively reduces the consumption of computing resources while ensuring the optimality of the path, enhances the safety and smoothness of the path, and provides a reliable path planning solution for the autonomous navigation of unmanned ships in real complex Marine environments. It has good engineering application prospects and promotion value.