Dr Siddartha Khastgir, UKRI Future Leaders Fellowship
Enabling a Novel Evaluation Continuum for Connected & Autonomous Vehicles (CAV)
Each year, approximately 1.25 million people die on the roads globally. It is approximated that 90% of these accidents are due to driver error. By assisting the driver or removing the driver, Connected and Autonomous Vehicles (CAV) could help to reduce this fatality number. However, safety of CAV remains the biggest challenge to their adoption.
The global CAV industry is estimated to be worth over £50billion (by 2035), with the UK CAV industry projections being over £3billion.
However, in order to achieve the safe introduction of CAV, significant research is needed to overcome barriers associated with the public deployment of CAV.
Developing pioneering testing methodologies
WMG’s Dr Siddartha Khastgir will use his Fellowship to help position the UK as a world leader in CAV research and innovation, to help create lasting societal and economic benefits.
His research will develop pioneering testing methodologies and standards to enable robust and safe use of CAVs with a focus on creating both fundamental knowledge and applied research methods and tools.
Dr Khastgir talks about his Fellowship and how it will support positioning the UK as a world leader in CAV research and innovation for long lasting societal and economic benefits.
A major focus of the Fellowship is the concept of safe AI, and the evidence needed to argue for safety of a AI (non-deterministic) based system.
Safety concerns
While prototype CAV technologies have existed for some time now, ensuring the safety of these technologies has proven to be a hindrance to commercialisation of CAV. It is suggested that to prove that CAV are safer than human drivers, they will need to be driven for more than 11 billion miles, an unfeasible proposition. Industry trends in CAV indicate widespread use of Artificial Intelligence (AI) in the autonomous control systems.
AI systems are non-deterministic in nature, resulting in different behaviours and a lack of transparency around the CAV system. This means it’s often difficult to identify reasons for a particular failure in such AI-based systems and take the corrective measures.
Creating evidence to prove CAV and AI-based systems are safe
Dr Khastgir's Fellowship will focus on three themes:
- Test Scenarios: Building on his previous research of Hazard Based Testing, Dr Khastgir will develop AI-driven approaches to identify scenarios which expose failures in CAV. Additionally, this project will develop a methodology to extract scenarios from raw real-world vehicle data.
- Simulation: It is widely accepted that simulation will play a key role in testing and safety evaluation of CAVs. However, understanding the fidelity requirements for simulation to enable its use for safety evidence generation still evades the industry – a research question for this project.
- Safety Evidence: A major focus for Dr Khastgir’s Fellowship is the concept of safe AI, and the evidence needed to argue for safety of an AI (non-deterministic) based system.
Further Impact: Influencing International Standards
Dr Khastgir represents the UK on various national and international standardisation committees. Research outputs from the three themes of the fellowship will provide inputs to new BSI, ISO and SAE standards, providing further insight and a clear route to deliver impact from the proposed research through the development of international standards, while also ensuring that the UK becomes a global leader in this area.
Dr Khastgir served as the lead author for international standards on Low-Speed Automated shuttles (ISO 22737) and national standard on taxonomy for Operational Design Domain (BSI PAS 1883). The vision for this Fellowship is to apply the tools, techniques and methods developed for safety of CAV to other safety critical domains, such as healthcare, electric vehicles/aviation and automated manufacturing.
To learn more about Dr Khastgir’s Fellowship including updates and a monthly blog, please visit his Fellowship website.