Structured Natural Language for expressing Rules of the Road for Automated Driving Systems

Abstract

Automated Driving Systems (ADSs), like human drivers, must be compliant with the rules of the road. However, current rules of the road are not well defined. They use inconsistent and ambiguous language. As a result, they are not sufficiently formal for machine interpretability, a necessity for applications of verification and validation (V&V) of ADSs. Rules must be defined in a way that make them usable to a variety of stakeholders. While first-order and temporal logic forms of rules of the road are needed for monitoring and verification during simulation and testing, a structured natural language for these rules is necessary for consistent definition. They must also adhering to standard vocabulary taxonomies of Operational Design Domain (ODD) and behaviour. This paper contributes a structured natural language based on formal logic, that allows rules of the road to be defined in a natural, yet precise manner, using concepts of ODD and behaviour, making them usable in the V&V of ADSs. We evaluate the effectiveness of the language on a selection of rules from the Vienna Convention on Road Traffic and the UK Highway Code.