Environment-Aware Task Allocation and Path Planning for Multi-USV in Ocean Environment

Multi-Unmanned Surface Vehicle (multi-USV) multi-task planning, coupled with discrete task allocation and continuous path planning, is important for efficient and safe cooperative maritime operations in complex ocean environments. Incorporating environmental awareness into this planning process remains challenging due to inherent nonlinearity and uncertainty. To investigate this problem, we develop an environment-aware task allocation scheme based on an unsupervised deep ensemble randomized network (U-DERN) that incorporates ocean current information during task allocation. Subsequently, we introduce an enhanced Crayfish Optimization Algorithm with Local Search Strategies (ECOA-LSS), which integrates Halton-sequence initialization to improve solution diversity, differential-evolution-inspired mutation and crossover to enhance global exploration, and a current-aware local search strategy to refine intra-route visiting orders based on current-induced navigation time. Experiments on multiple traveling salesman problem instances and simulated ocean-current-aware multi-task instances show that ECOA-LSS generally achieves better solution quality and shorter navigation time than several representative heuristic baselines, especially on larger and more complex cases. Ablation studies confirm the complementary contributions of the proposed modules. Compared with a representative multi-task allocation method, the proposed framework also achieves competitive navigation time, suggesting its potential utility for multi-USV multi-task planning under the considered ocean-environment settings.