A Full‐Coverage Underwater Path Planning Algorithm for Biomimetic Robotic Fish Based on a Three‐Point Crossover and Local Optimization Genetic Algorithm

This study addresses path redundancy, excessive turning, and high hydrodynamic cost in full‐coverage path planning for biomimetic robotic fish operating in cluttered underwater environments. A three‐point‐crossover genetic algorithm with local refinement is proposed for fixed‐depth two‐dimensional coverage tasks. By using discrete grid modeling, coverage, path length, smoothness, and flow‐related propulsion energy are jointly optimized in a unified fitness function. The three‐point crossover is designed to retain useful contiguous subsequences from parent paths during recombination, while a diversity‐guided adaptive mutation strategy maintains exploration when the population begins to stagnate. A gradient‐guided multi‐point refinement (GGMPR) step is then used to smooth high‐curvature and counter‐current segments. Thirty‐run benchmark results on regular quadrilateral and irregular pentagonal maps show that the proposed method reduces path length, sharp turns, and estimated energy relative to a standard genetic algorithm and gives slightly better path‐quality indicators than simulated annealing. On the two enlarged maps, TPGA‐tuned reduces mean path length, sharp‐turn count, and estimated energy relative to simulated annealing under a larger evolutionary budget. The approach therefore provides a simulation‐level planning framework for underwater inspection and ecological monitoring tasks requiring executable and energy‐aware coverage paths. © 2026 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.