In what feels decades ago, I had the opportunity to join on a mission to Singapore on the topic of connected and autonomous vehicles (CAV). It was an occasion to discuss how collaboration could help deliver the benefits we have all been looking forwards to: reduced incidents on the roads, a more accessible transport system and a better performing road network.
The mission included visiting the Nanyang Technological University modelling teams, technology providers, CETRAN testbed and discussing with the Land Transport Authority their approach to enable the sector. At the time, I was leading the first dedicated CAV team in a UK local authority (Oxfordshire County Council1, OCC) and my interests were specific: CAVs are a booming industry2 in Oxfordshire3, even before considering the close-by regions such as West Midlands4, London5, Bristol6 or Cambridge7. How can we make it easier for developers to grow locally, reduce trials‘ risks, and collaborate internationally? What steps do we need to take to prepare the road network for this revolutionary shift of how people and goods are transported?
Singapore laid out a logical plan: a three-stage approach progressing from simulation to testbed and then real-world testing for CAV suppliers to scale their applications. Combined with this, there was an ongoing upgrade of roadside infrastructure and their traffic control centre capabilities. Their level of awareness of what is happening on the network was impressive. Strict requirements for data shared back with the authorities were set out8. It was centralised in many senses: one testbed, one simulation package, one highways authority, for one flourishing city. However, both literally and metaphorically, I felt this was miles away from where I came.
In the UK, we have 153 local authorities, each with their combination of systems to control the roads installed at different times by different suppliers under various contracts from companies.
The authorities can be unitary, two-tier or even combinations, and they range from entirely rural, almost off-grid, to Europe’s only megacity. There is no centralised procurement for all traffic lights or even an up-to-date map of mobile connectivity. According to the Department for Transport’s code of practice9, CAV trials’ requirements are light touch to enable the industry, with the liability staying with the trialling organisation.
Perhaps because of this, UK CAV suppliers have developed systems with minimum external input requirements. Although work is needed in collaboration and integration, e.g. with traffic management to optimise the flow, the UK setting has given rise to a plethora of impressive solutions. Additionally, the Zenzic CAM testbeds have developed to allow for testing in those different conditions.
Representing a mostly rural county, I always felt this to be an advantage of the UK approach. While megacities host most of the world’s population10, these usually already have multiple transport options. On the other hand, the highest car dependency, isolation and fatal collisions occur on rural roads. Thus, to deliver inclusive and safer mobility, all terrains must be considered. Except in cases where large infrastructure investment is possible, CAVs will have to deal with the real world as it is, not how their software wished it could be, without all conditions being specified and ideal. That is not just for Oxfordshire’s benefit: export markets to Africa or Eastern Europe would benefit from such a minimal-requirement approach.
But how could we tell which solutions were appropriate? What are “similar conditions”? There was a gap in our roadmap to collaborate internationally: a systematic way to assess how CAVs will perform on different terrains. The solutions we saw in that trip were impressive, but I had questions to ask them about dealing with the UK roads: from faint road markings to obscured traffic signs and the unique British weather, we had things in Oxfordshire that Singapore would not have prepared them for.
The challenge of creating a taxonomy able to describe the conditions anywhere in the world at the appropriate level was daunting. It is to the credit of our industry and CCAV that since 2020 we have a BSI world-first standard the creates that Operational Design Domain (ODD) categorisation11. That means when discussing with any supplier and we can, for the first time, refer to a set of conditions as a specific ODD rather than long-winded paragraphs trying to describe the difference between locations.
Now that the ODDs have a common language, we need a method to determine if the CAVs can operate safely. For a highways authority, this is a critical question: are these CAVs ready for the road? Specifically, are they ready for our roads? The answer to that question hints on whether more testbed development or real-world trialling is needed.
Thus, a different kind of test was needed: a simulation that would assess the performance of CAVs. That simulation also needed to be representative of the ODDs and independent from any supplier.
It is astonishing that we now have a proof of concept of this simulation only a little more than two years later: the OmniCAV project. Created collaboratively, OmniCAV combines a digital twin with traffic models, historical data, ML pedestrian movements and links to the National Scenario Database12. RACE and Thatcham Research supported the testing; while Admiral, Arrival, and OCC provided data and are potential end-users. It has a first-in-the-world integration between different simulations systems with traffic management and mapping data sources and, crucially, it links test facilities and the real world. It is, of course, based in Oxfordshire. CAV developers can test their performance in the different ODDs found, using different scenarios and select the most appropriate trial locations according to their capabilities and use cases. They can also compare simulation and testbed performance in RACE.
Our work in OmniCAV has mostly supported the link between the digital and physical: from providing accident and traffic flow historic data, to feedback on the local conditions, and upgrading our infrastructure to enable machine learning through CCTV cameras. In the process, we have come to appreciate the variety of different ways of mapping information between platforms (such as a “lane”) and became the first local authority to join ASAM.
From a local authority perspective, OmniCAV represents a crucial step in delivering the benefits of CAVs to the real world: it enables an independent assessment of the vehicle’s performance. Of course, every model is an approximation. Therefore, it will never be entirely accurate on its assessment or representation. At best, it can indicate the risk as part of a due diligence process. However, the road loop that OmniCAV is based on was chosen to span urban, peri-urban, highways and rural areas to bring down the barriers for all locations for inclusive mobility. Thus, OmniCAV can make it easier and safer for CAV suppliers to assess and progress their development in Oxfordshire both digitally and physically.
We need independent systems such as OmniCAV to enable CAV developers to answer the critical question whether asked by a highways authority, an insurer, export partner or a transport operator: “How do you know you are ready for this road?”.