Sense-Assess-Explain
For the SAX project, ORI led a diverse team of software and hardware engineers, researchers, and specialist drivers. The aim of the project was to build robots, or autonomous vehicles, that can:
- sense and fully understand their environment
- assess their own capabilities
- provide causal explanations for their own decisions
The SAX project extended methods for interpreting and representing an autonomous vehicle's observations of the environment in human-understandable terms. To this end, they have improved the way that traditional sensors (e.g., cameras, lasers) are used in complex and rare traffic situations.
The expert driver provided extensive commentary that the team were able to analyse and combine with data being gathered by the vehicle’s systems. This helped Lars and the team to look into explainability while driving on road. They used a technique called ‘commentary driving’ where you describe what you see, what you anticipate happening, then how you reacted to these situations.
The resulting ten hours of commentary (part of a wider dataset of 140 hours covering over 3,700 miles of on and off-road driving) was a valuable output from the project. It has the potential to be used to validate design and assurance capabilities.
Benefits of collaboration
As a result of the project, Lars was contacted by the European Commission and is now part of the expert group focusing on explainability for automated and autonomous driving for the Commission’s Joint Centre for Research.
Working together has also led to further collaboration. AAIP's Richard Hawkins and Lars are collaborating on a project for UKRI’s Trustworthy Autonomous Systems Hub. This follow-on project, led by Lars, explores ‘Responsible AI for Long-term Trustworthy Autonomous Systems’ (RAILS) and extends their investigations to maritime and the use of drones.
