Large Language Models in Intelligent Robotic Systems for Environment Clean UpIf you are interested in robotics, and have strong programming and software engineering skills, this could be a project for you. Successful completion of this project could lead to career prospects inside the UKAEA. Supervision will be provided jointly by the Department of Computer Science at the University of York and the UKAEA/RACE. The project is concerned with testing of robots that can autonomously clean up areas where we can find either hazardous materials or humans, or both. A challenge is the design and verification of systems that have a vision component to identify particular materials or humans. Use of a Large Language Model (LLM) for that is a promising prospect, but raises concerns about the reliability of a system that depends on such a component. To allow a robot to operate hazardous materials or interact with humans, we need to trust that robot. Potential consequences if they fail in their operation can range from wasted time and materials, to more serious incidents. In this project, we will apply state-of-the-art technology for modelling, simulation, and testing for analysis and verification of robots that can be trusted to work autonomously. We will consider, first, existing technology to evaluate the limits of what can be achieved with existing robots and existing Software Engineering techniques. Second, we will consider advanced testing techniques that support evaluation of designs of systems that use an LLM. Contributions of the work will span from improvements to automation to novel techniques for testing. Our challenge is to enable and promote development approaches that provide evidence that the robot can be trusted to work with sensitive materials and experiments, and potentially around humans. The successful candidate will benefit from the general training provided by the Department of Computer Science. This will cover topics such as security, research management and leadership, collaborations, employability, public engagement, and communication. Moreover, the candidate will benefit from inclusion in a cohort for our Centre for Doctoral Training on Autonomous Robotic Systems for Laboratory Experiments. This will involve regular meetings with students working on similar topics. This project will be part-funded by the UK Atomic Energy Authority’s RACE (Remote Applications in Challenging Environments, https://www.race.ukaea.uk) robotics and remote handling centre. There will be an opportunity to work from RACE about two weeks per year. RACE was founded in 2014 as part of the UKAEA’s Fusion Research and Development Programme to design and test robots for operating in some of the most challenging environments imaginable. UKAEA’s wider mission is to lead the commercial development of fusion power and related technology and position the UK as a leader in sustainable nuclear energy. Based at its Culham Campus HQ, near Oxford, and a new technology facility in South Yorkshire, UKAEA has, until recently, operated a Joint European Torus (JET) fusion experiment on behalf of scientists from 28 European countries; now it is leading the decommissioning of the JET machine. UKAEA is keeping the UK at the forefront of fusion as the world comes together to build the first powerplant-scale experiment, ITER—one step away from the realisation of fusion as a low-carbon energy source. UKAEA is involved in future fusion demonstration powerplant design activities such as DEMO and the UK’s future STEP powerplant. References
A. L. C. Cavalcanti, W. Barnett, J.
Baxter, G. Carvalho, M. C. Filho, A. Miyazawa, P. Ribeiro, and A. C.
A. Sampaio. RoboStar Technology: A Roboticist's Toolbox for Combined
Proof, Simulation, and Testing, pages 249--293. Springer
International Publishing, 2021. (http://dx.doi.org/10.1007/ W. Barnett, A. L. C. Cavalcanti, and A.
Miyazawa. Architectural Modelling for Robotics: RoboArch and the
CorteX example. Frontiers of Robotics and AI, 2022. (https://doi.org/10.3389/
Z. Attala, A. L.C.
Cavalcanti, and J. C. P. Woodcock. Modelling and verifying robotic
software that uses neural networks. In E. Ábrahám, C. Dubslaff, and
S. L. T. Tarifa, editors, Theoretical Aspects of Computing,
pages 15--35. Springer, 2023. (http://dx.doi.org/10.1007/
A. L. C. Cavalcanti,
J. Baxter, R. M. Hierons, and R. Lefticaru. Testing Robots using
CSP. In D. Beyer and C. Keller, editors, Tests and Proofs,
pages 21--38. Springer, 2019. (http://dx.doi.org/doi.org/10.
A comprehensive list of publications is
provided on the
RoboStar site
(https://robostar.cs.york.ac. To apply
|