Task assignment and path planning methods for marine unmanned aerial vehicles based on improved intelligent optimisation algorithms

Marine UAV swarm planning requires heterogeneous task allocation and route planning under range, payload, obstacle, risk-area, and environmental-cost uncertainty constraints. This study develops a two-stage framework that explicitly couples task-cluster generation with route-level feasibility verification. In the first stage, a problem-specific NL-DW-SA allocation pipeline with common ACO-based within-cluster sequencing combines nonlinear capability budgeting, distance-aware clustering, and simulated-annealing boundary repair to generate capability-constrained and spatially compact task clusters. In the second stage, LS-EL-DE-WOA optimises the visiting sequence within each cluster and constructs closed polyline routes connecting the base station and task nodes. When maximum-range violations, hard-obstacle conflicts, or excessive route-cost imbalances occur, route-level information is returned to the assignment stage for task migration or exchange. A unified AHP-based evaluation system is used to assess solution quality, stability, and computational efficiency. Under author-constructed simulation scenarios, the proposed task-assignment and route-sequence variants achieved the highest comprehensive scores of 0.9368 and 0.8427, respectively. Across 20 baseline runs, backward feedback was activated in eight cases and restored feasibility in seven cases within at most four feedback iterations. Weight and parameter sensitivity analyses, together with disturbance and extreme-scenario tests, indicate that the principal comparative conclusions are stable within the tested ranges. These results provide algorithm-level evidence for coordinated marine-UAV task assignment and route-sequence planning, but do not establish real-world flight readiness or field-deployment performance. Not applicable.