Path planning of long-range unmanned ship based on improved ant colony algorithm
QIAO Zhen,YIN Chuan-zhong,QIU Xin
Addressing the issue of insufficient endurance for unmanned surface vehicles (USVs), an angle-priority ant colony optimization algorithm based on reinforcement learning is proposed to plan long-range paths for USVs. The Canny operator is employed to extract environmental information, and a combination of pixel grayscale weighted averaging and Gaussian filtering smoothing is applied to analyze the gradients in the image and extract edge features. The ginput function is then used to extract the edge coordinates. Based on the MAKLINK graph theory principle, the edge coordinates are defined as nodes, and the link lines and path points between nodes are used to represent the structure and connectivity of the navigation environment, thereby establishing a navigation environment model. By comparing the shortest paths obtained under the influence of different heuristic factors in the navigation environment, the optimal combination of heuristic factors for the algorithm is determined. An angle-priority mechanism is introduced to improve convergence speed, and reinforcement learning reward and penalty coefficients are utilized to adjust pheromone concen-tration, optimizing the algorithm flow and generating the optimal path. Experimental results demonstrate that the improved algorithm yields a USV navigation path with smooth corners, enhances the endurance of the USV by 4.6%, and accelerates the algorithm convergence speed by 68.9%.
