HTP Graphics

LEAD RESEARCHER: MATT LUCKCUCK I am a post-doctoral research associate working on formal verification of autonomous and automated systems. My work in the RAIN Hub began with an extensive survey paper, and then focussed on linking heterogeneous verification approaches applied across an autonomous software system, and runtime verification of an autonomous system's behaviour. I also lead a collaboration with the UK's Office for Nuclear Regulation on developing guidance for developers of autonomous systems that ensures their systems are amenable to robust verification and can provide useful assurance evidence. SUMMARY // Autonomous Systems present unique development and assurance challenges, in whichever sector the system will be used. The failure of an autonomous system in the nuclear sector could be disastrous. A variety of standards and recommendations are being, or have been, developed, but there is no de-facto or de-jure standard as yet, which leaves a gap in the guidance on how to develop and assure systems that have any level of autonomy. Along with the UK's Office for Nuclear Regulation, we held workshops and produced a white paper that explored the assurance challenges for autonomous robotic systems, and provide guidance about how they can be overcome. The workshops provided a forum for academia, the regulator, and supply chain businesses to discuss the challenges autonomous systems present. The white paper presents high-level principles for the design and assurance of autonomous systems for the nuclear sector, but with a view to wider applicability. DEVELOPMENT AND ASSURANCE GUIDELINES UNIQUENESS // The unique angle of this strand of work has been the involvement of the Office for Nuclear Regulation (ONR). Some supply chain businesses and nuclear sites are already developing robotic and autonomous systems, so this collaboration has helped to start the conversation about how the UK nuclear industry can assure the safety of these systems. The collaboration has involved a lot of knowledge exchange, with discussion about the ONR's regulatory approach, current capabilities of autonomy 22 and verification, and how to present assurance evidence. The involvement of the ONR in drafting the white paper helps to ensure that its recommendations are suitable for producing assurance evidence in a form that is acceptable to the regulator. Their insights were invaluable, both for the white paper's utility and for potentially adapting it to ensure it is compatible with regulatory approaches in other sectors where autonomous systems might be used.

• Cover