SOLITUDE: Safety Argument for Learning-enabled Autonomous Underwater Vehicles

Description

This project will develop a safety argument for maritime autonomous systems. Methodologically, we will build on, and extend, the existing safety argument method for traditional hardware and software systems, by focusing on the distinct features of autonomous systems – deep learning components. We will demonstrate our method over two real-world autonomous underwater vehicles (BlueRov and Saab Seaeye Falcon).

Funding Agency: Defence Science and Technology Laboratory (Dstl)

Project Time: 2020 - 2022

Personnel

  • Dr Xiaowei Huang (PI)
  • Dr Xingyu Zhao (Co-I)
  • Prof Sven Schewe (Co-I)
  • Prof Simon Maskell (Co-I)

External Collaborators

  • Dr Sen Wang (Lead of Heriot-Watt University)
  • Dr Alexander Phillips and Dr Catherine Harris (National Oceanography Centre, NOC)
  • Prof Robin Bloomfield (Adelard and City University of London)

Publications

  • Zhao, X., Huang, W., Banks, A., Cox, V., Flynn, D., Schewe, S., and Huang, X. (2021a).Assessing the reliability of deep learning classifiers through robustness evaluation andoperational profiles. InAISafety’21 Workshop at IJCAI’21.(link)
  • Zhao, X., Huang, W., Huang, X., Robu, V. and Flynn, D., 2021. BayLIME: Bayesian Local Interpretable Model-Agnostic Explanations, UAI2021. (link)
  • Salako, K., Strigini, L. and Zhao, X., 2021. Conservative Confidence Bounds in Safety, from Generalised Claims of Improvement & Statistical Evidence, DSN’21. (link)
  • Zhao, X., Huang, W., Schewe, S., Dong, Y. and Huang, X., 2021. Detecting Operational Adversarial Examples for Reliable Deep Learning, DSN’21 (fast abstract) (link)
  • Zhao, X., Banks, A., Sharp, J., Robu, V., Flynn, D., Fisher, M. and Huang, X., 2020, September. A safety framework for critical systems utilising deep neural networks. In International Conference on Computer Safety, Reliability, and Security (pp. 244-259). Springer, Cham. (link)