Automated Diagnosis for Fault Detection, Identification and Recovery in Autosub 6000
The objective of the project is to provide an automated fault diagnosis system for a number of subsystems of Autosub 6000, an autonomous underwater vehicle (AUV) operated by the National Oceanography Centre, Southampton (NOCS). This project, funded by the National Environment Research Council, runs from October 2008 to the end of 2011.The project is funded from the Oceans 2025 Strategic Ocean Funding Initiative.
The aim of the project is to provide a lasting legacy to the Autosub program in the form of a working diagnosis system for at least some vehicle subsystems that can be used and extended by members of the Autosub team at NOCS. Hence, the project involves not just developing algorithms and software to diagnose faults, but also deploying and testing the software on-board Autosub during real missions. While there is a significant research component to the project, these requirements mean that it also requires a software engineering effort to ensure reliable and efficient operation.
The project will use two different fault diagnosis technologies. The first, is the Livingstone FDIR engine, developed at NASA Ames Research Center, and successfully flown on the Deep Space 1 and EO-1 spacecraft. Livingstone is a discrete diagnosis system, and will be used for vehicle subsystems that can relatively easily be treated as discrete systems. The work on this part of the project will be on optimising Livingstone to run on the AUV, and to build models of AUV subsystems.
For AUV subsystems that can’t easily be modelled as discrete systems, a hybrid diagnosis system will be used. A likely approach is to use particle filters to track the continuous state of the system, and build differential equation models of the subsystems and their nominal and fault behaviours. There are significant research challenges in this approach, including issues of scaling existing approaches to work on Autosub, and in terms of how to combine these approaches with the discrete approach of Livingstone in order to produce overall estimates of the system state.
- Dr. Richard Dearden
- Dr. Jeremy Wyatt
- Dr. Juhan Ernits
We work in close collaboration with the Autosub Team.
We use a custom Eclipse IDE based model-based diagnosis development environment called Piccard to develop Livingstone 2 models.
Juhan Ernits, Richard Dearden. 2011. Fault Diagnosis Challenge in a Flight-Class Autonomous Underwater Vehicle. In Proceedings of the 22nd International Workshop on the Principles of Diagnosis (DX 11).
Juhan Ernits, Richard Dearden. 2011. Towards Diagnosis Modulo Theories. In Proceedings of the 22nd International Workshop on the Principles of Diagnosis (DX 11).
Juhan Ernits, Richard Dearden, Miles Pebody, and James Guggenheim. 2010. Diagnosis of Autosub 6000 using Automatically Generated Software Models. In Proceedings of the 21st International Workshop on the Principles of Diagnosis (DX-10).
Zeyn Saigol, Frederic Py, Kanna Rajan, Conor McGann, Jeremy Wyatt and Richard Dearden. 2010. Randomized Testing for Robotic Plan Execution for Autonomous Systems. In the IEEE Ocean Engineering Society Autonomous Underwater Vehicles 2010.
Juhan Ernits, Richard Dearden and Miles Pebody. 2010. Automatic Fault Detection and Execution Monitoring for AUV Missions. In the IEEE Ocean Engineering Society Autonomous Underwater Vehicles 2010.
Minlue Wang and Richard Dearden. 2009. Detecting and Learning Unknown Fault States in Hybrid Diagnosis. In Proceedings of Diagnosis 2009 (DX-09)
Juhan Ernits, Richard Dearden and Miles Pebody. 2009. Formal Methods for Automated Diagnosis of Autosub 6000. In Proceedings of the NASA Formal Methods Symposium, 2009.
Autosub 6000 data
In collaboration with the Autosub Team, we are in the process of making the data recorded during a number of missions available to the public. There are missions where no faults occurred and there are missions with actual faults. The data is available here.