Multiple targets traversing for unmanned surface vehicles by bundled genetic optimization and fast-marching Q-Learning.

This paper discusses the use of unmanned surface vehicles (USVs) in the ocean for traversing multiple targets and proposes a novel multiple targets traversing algorithm that takes into account the multi-level priority of target assignment and path optimization. The algorithm includes a bundled genetic optimization target assignment method and a fast marching-Q-Learning path planning method. Comparative studies with existing methods show that the proposed algorithm have efficient performance, where the calculation time for target assignment has been reduced by 45.64%, and the path length has been further optimized by 15.3%. Considering the velocities of USVs navigation, path turning motion and multi-level priorities constraints are satisfied, and collision risk is avoided in dynamic ocean environment. Hence, the proposed multiple targets traversing algorithm plays an important role in the ocean engineering field of USVs. • Bundled-genetic target assignment method: The core of this method is to use bundling mechanism to solve target assignment efficiency problem, including improving the speed of target assignment, satisfying USV velocity and multi-level priority constraints. • FM-Q-Learning path planning method: This method utilizes FM's reward function to solve path optimization problem, including improving path convergence speed, decreasing path length, and smoothing the paths through navigation restriction. • Simulation of remote sensing satellite background: This strategy can reflect the experimental results in a real environment, which helps to verify the applicability of theoretical algorithms in practical situations. [ABSTRACT FROM AUTHOR]