Journal of Marine Science and Engineering, Год журнала: 2025, Номер 13(2), С. 338 - 338

Опубликована: Фев. 12, 2025

Collision avoidance algorithms play a crucial role in ensuring the safety and effectiveness of autonomous ships, which require comprehensive testing realistic multi-ship encounter scenarios. However, existing scenario generation methods often inadequately represent spatiotemporal complexity dynamic risk interactions real-world encounters, leading to biased evaluations. To bridge this gap, paper proposes combinatorial-testing-based framework integrated with optimisation. First, full-process representation model is developed by abstracting navigation features into discretised parameter space. Subsequently, method adopted cover space, generating high-coverage set. Finally, introduced filter out oversimplified scenarios extremely dangerous Experiments demonstrated that 13.7% generated were eliminated as unrealistic or trivial, while high-risk interaction amplified 7.96 times 5.84 times, respectively. Compared conventional methods, optimised set exhibited superior alignment complexity, including escalation coordination challenges. The proposed not only advances methodology through its integration combinatorial complexity-driven optimisation, but also provides practical tool for rigorously validating ship systems.

Язык: Английский