Skip to main content

Work Package 4 – Assurance

Generating better evidence and arguments for ethical, safe, and secure design and implementation of robotics and autonomous systems.

This work package focuses on the assurance challenges that come with Autonomous Robotic Systems. What kind of guarantees do we need? We want guarantees of safety, but also security and potentially ethics. How do we design our systems so we can evidence those guarantees? And how can we curate and present that evidence?

EVIDENCE GENERATION

The evidence required by a regulator or other body, typically involves not just the system itself (the robot and its software) but evidence about the way that system was developed, how hazards were identified and mitigations devised. Therefore, we are looking both at techniques for verifying the behaviour of Robotic Autonomous Systems, but also how those techniques fit into toolchains for generating and organising evidence throughout the development and deployment lifecycles.

INFORMING FUTURE STANDARDS

Much of the evidence needed will involve demonstrating adherence to standards.  However, there is a lack of widely accepted standards for the development of AI and Robotic Autonomous Systems.  We plan to perform basic research into what it means for a robot to be safe, secure and ethical and how this can be established through the development process so that this can inform future standards.

DESIGN FOR ASSURANCE

Together with WP2 we will investigate methodologies for designing Robotic Autonomous Systems in ways that will enable us to evidence their safety.  This will include developing component-based architectures for these systems, linked with techniques for validating individual components and patterns of assurance that can be related to design patterns.

Updates

Research Spotlight – Dr Dhaminda Abeywickrama

Dhaminda’s current research is focused on developing a ‘reference assurance case’ for a ground-based autonomous inspection robot. A reference assurance case is a structured framework that serves as a standardized template (e.g., patterns) and as an example for developing assurance cases across various industries. It provides a baseline of accepted practices, arguments, and evidence, which can be adapted to specific projects or systems. The reference case will also incorporate physical/functional architecture, hazard analysis, requirements, safety architecture, and arguments, with an initial application in a case study on verge inspections for the highways sector. In the first year, Dhaminda has published three research papers, which acknowledge CRADLE, on topics including risk analysis for autonomous robotic swarms, standards for soft robotics, and a corroborative approach to the verification and validation of robotic swarms.

Research Spotlight – Dr Yasmeen Rafiq

Yasmeen joined the team in April 2025. She brought with her expertise in model-based software engineering and verification for robotic systems. She has been working with WPs 1 and 5 on the design and verification of a sewer inspection robot as well as proposing a new case study associated with Robot Assisted Dressing.

Reference Assurance Case

We have made exciting progress towards producing a reference assurance case for an autonomous inspection robot. Our vision is that this should present a case for the safe deployment of an inspection robot in a hypothetical nuclear environment. The assurance case will focus particularly on those aspects related to autonomy but with clear placeholders for whether other evidence is needed. This work is informed by the knowledge and experience shared during our year one regulator engagement workshop. The year started with scoping work to define our methodology and approach. This resulted in two publications (at ICSR and AREA 2025). We had a follow up webinar in July with various regulators reporting on progress and receiving feedback on the proposed approach. Work has now begun in earnest with the aim of producing a draft report in the summer of 2026. Dr Dhaminda Abeywickrama has been leading this research strand.

    Explore Work Packages

    WP 1 Components
    WP2 Architectures
    WP3 Interactions
    WP4 Assurance
    WP5 Demonstrators