Automated Generation of Functionally Complete Assurance Suites for COLREGS-Compliance of Autonomous Surface Vehicles

Abstract

Autonomous surface vehicles (ASVs) operating in maritime traffic must comply with the International Regulations for Preventing Collisions at Sea (COLREGS). Demonstrating COLREGS compliance requires systematic assurance that all relevant encounter scenarios are represented and analyzed. Recent work has proposed a research agenda for scenario-based assurance of COLREGS compliance based on qualitative abstractions and multi-level scenario modeling. In this paper, we provide the first realization of a model-based approach that systematically derives initial scenes as assurance scenarios for multi-vessel encounters. First, a complete set of functional equivalence classes is derived automatically. Then, we provide a mapping from functional equivalence classes to logical scenarios that capture geometrical constraints between potentially hazardous ship encounters. Next, initial scenes with precise vessel placements are generated from logical scenarios by (1) a search-based algorithm and (2) a novel rejection sampling-based algorithm. Finally, we derive (static) trajectories amenable to simulation for ASV assurance purposes. Our evaluation demonstrates that the approach achieves full coverage of relevant COLREGS encounters at the functional level and scales to complex scenarios up to six vessels, which represent the vast majority of open sea encounters.