Model-Driven Scenario-Based Assurance for the COLREGS Compliance of Autonomous Surface Vehicles

A doctoral symposium paper is a compact description of a thesis while it is still in progress. Mine frames the PhD around three objectives:

• O1: Specification. A language for system-level ASV test scenarios that models vessels, obstacles, environmental conditions, geometric relations and COLREGS situations through qualitative abstractions.
• O2: Generation. Automatic derivation of relevant, functionally complete, diverse and robust multi-vessel assurance suites.
• O3: Execution and evaluation. Integrating the scenarios into naval simulators and producing assurance evidence for safety and COLREGS compliance.

It also places the published work against these objectives: the ISSTA 2025 study, the MODELS 2026 refinement approach and the SceneGEMS framework. The open problem it puts on the table is robustness: how to cover a single equivalence class with a sufficiently diverse set of tests, especially near boundaries where two concrete scenarios look alike but are interpreted differently under the COLREGS.

The talk itself was a five-minute pitch followed by feedback from the symposium panel.