Meet EVENTS team-WMG

WMG is a research-intensive academic department of the University of Warwick (UoW) with over 800 staff and one of the world’s leading research and education groups, designing solutions and overcoming challenges through collaborative R&D.

WMG is internationally renowned for its research in the areas of Connected and Automated Vehicles (CAVs) and manufacturing. WMG is well integrated with the UK automotive industry and hosts the UK’s flagship National Automotive Innovation Centre, established with some £150m investment by JLR, Tata Motors’ European Technical Centre, and the UK Research Partnership Investment Fund, addressing the ‘smart and connected car’ agenda of the UK Automotive Council.

In EVENTS, WMG is represented by the Connected and Cooperative Autonomous Systems (CCAS) team and the Safe Autonomy team, which are both part of WMG’s Intelligent Vehicles group.

What is your role and involvement in the EVENTS project?

The involvement of WMG in EVENTS is two-fold.

Firstly, the CCAS team is focused on designing perception integrity monitoring systems using Deep Neural Networks (DNNs) for onboard object detectors, utilising 2D camera-based or 3D LiDAR-based technologies. This research is motivated by the need for resilient perception systems, which would be critical for achieving SAE L3 and higher levels of autonomy. For instance, several accidents of commercial SAE L2 autonomous vehicles have been attributed to undetected perception errors under adverse weather or lighting conditions. Throughout the project, the design of perception integrity monitoring systems benefits from the WMG prototype vehicle for real-world data collection, field trials, and experimentations. Additionally, the CCAS team is developing integrity monitoring systems for LiDAR-based distance estimation to the leading vehicle aiming at improving the reliability of Adaptive Cruise Control (ACC) systems.

Secondly, the Safe Autonomy Team at WMG performed STPA (Systems-Theoretic Process Analysis) as part of the vehicle system hazard analysis & risk assessment, to identify the risks and potential hazards associated with each use case identified as part of the project. By applying STPA (which treats safety as a dynamic control problem rather than a failure prevention problem and is suitable for complex safety-critical systems) in combination with the traditional HAZOP (Hazard and Operability Study) process, we have demonstrated an enhanced safety analysis approach, making it more tailored for Connected and Automated Vehicles (CAVs).

The Safe Autonomy team also leads the Standardisation & international liaison activities of the EVENTS project to centrally coordinate all the inputs stemming from EVENTS to different standardization bodies and organisations. The project team members are involved in various international standardization and regulatory activities. These include 1) ISO: ISO TC22 SC33 WG9 (test scenarios); ISO TC204 WG14 (functional requirements for ADAS and ADS); ISO TC22 SC32 WG13 (safety of automated driving systems) 2) ASAM: OpenODD (project lead); OpenLabel (work package lead); OpenXOntology (work package lead); OpenSCENARIO 1. x & 2.0; OpenDRIVE; Open Simulator Interface (OSI); OpenTest 3) SAE: On-Road Automated Driving (ORAD) V&V task force; ORAD metrics task force 4) UNECE FRAV and VMAD: UK representatives on the committees and 5) The British Standard Institute (BSI): BSI PAS 1883; BSI Flex 1889. EVENTS aims to provide a concrete and direct input feeding into relevant standardization processes through WMG, which is participating in or chairing several relevant Working Groups such as “ISO TC204 WG14: Functional requirements for ADAS and ADS systems”, “SAE ODD definition”, “ASAM OpenODD”, “ASAM OSI”, “BSI PAS 1883” and “BSI Flex 1889”.

What are the key challenges you face in developing perception integrity monitoring techniques for CAVs? 

The perception integrity monitoring systems developed by WMG adopt an actor-critic architecture, where a separate system (the critic) continuously monitors and assesses the performance of the onboard object detector (the actor). In our implementation, the critic employs DNNs for this task. A key challenge in the design lies in the selection of the most appropriate monitoring parameters, such as raw sensor data, the list of detected objects and/or intermediate outputs from the object detector. Another significant challenge is ensuring the critic system’s performance under domain shifts. Furthermore, for real-time demonstrations using prototype vehicles, several implementation challenges within the Autoware platform must be resolved.

In terms of functional safety, we align with the latest regulations (like DCAS and ALKS) and standards (ISO 26262, ISO 21448 for SOTIF). Leveraging our experience collaborating with OEMs, we bring forth innovative solutions that consistently push the boundaries of the automotive industry’s latest advancements, staying at the forefront of technological excellence.

Additionally, our ability to integrate hardware and software into prototype vehicles and test the full system supports the V-cycle development process expected in the EVENTS project. And lastly, our experience managing automotive projects helps ensure we contribute efficiently across the EVENTS work packages.

What are the key challenges you face regarding standardisation and/ international liaison activities? 

Some of the challenges faced are creating the applied research evidence, translating those into standards, driving consensus within a large group and working with different cultures. Working with various industrial partners as part of the EVENTS project and the wider CAV ecosystem have helped identify gaps in standardisation activities and ensure that research outputs are practically implementable.

In your view, what is still needed to develop trustworthy perception systems for CAVs? 

While perception integrity monitoring systems are certainly important to initiate a safe Minimum Risk Manoeuvre in SAE L4 and L5, or promptly alert the human driver to take over the control of the vehicle in SAE L3, simply raising an alert could be insufficient without providing context. Therefore, integrating explainability into perception integrity monitoring systems would be a stride from perception to situational awareness. Explainability would also be vital for building trust between the autonomous systems and various automotive stakeholders including the end users, policymakers and developers, ultimately supporting the broader adoption of self-driving vehicles in our society.

Anything else you would like to mention or highlight?

Industry, academics and the governments have highlighted the importance of the public being able to trust Automated Vehicles to use them. To envision a safe AV future industry, academia, policymakers, and the governments need to work together to share knowledge, exchange information, and tackle the social and technical challenges. It is important to gain public trust and acceptance of self-driving technology through accurate and inclusive awareness and education programmes. The public need to understand the AV and its functions, benefits and limitations, ethics and regulation in this space and use cases. At WMG, we have been leading and facilitating cross-sector collaborations to ensure the safe deployment of Automated Vehicles on the UK’s roads.

We would like to express our gratitude to all the EVENTS project partners and hope that we can continue to collaborate on future projects.

Connected and Cooperative Autonomous Systems (CCAS) team and

Safe Autonomy team,

WMG’s Intelligent Vehicles group