Synthesizing deterministic maritime traffic situations for testing autonomy

Abstract

This work presents a hybrid scenario synthesis approach combining static and adaptive techniques to evaluate autonomous maritime systems under multi-factor conditions. Static scenarios offer efficient, expert-driven testing but can falter in complex environments where assumptions no longer hold. Adaptive scenarios capture system adaptability through evolving interactions but demand greater design and computational resources. The hybrid method deterministically generates maritime traffic interactions, supporting black-box evaluations resilient to unpredictable system motion. Key innovations include precise interaction design, closed-loop scenario selection based on prior system behavior, and efficient filtering of test cases. The method is implemented within the U.S. Navy’s Autonomous Systems Test Capability, specifically leveraging the Virtual Maritime Testing Environment which supports high-fidelity, multi-agent simulations with runtime triggers and intelligent traffic behaviors that enable repeatable, goal-driven testing scenarios. Verification through MATLAB-based implementation and Monte Carlo simulations showed consistent scenario execution. These results, when compared to static scenario analysis, demonstrate the approach’s robustness.