It was exciting to see so many new ideas coming from the AI world in all forms of traffic management and new mobility at January’s Transport AI UK event in Manchester.

While I was there, I was also pleased to see, at long last, recognition that some of the tools being used in the UK and on the shelf are already AI, even though they are not being sold as such.

One key area I focused upon in my presentation, ‘The role of data in AI for connected and automated vehicles’, was the data used to train AI and then subsequently operate it in anger.

There is often an assumption from outside the traffic world that we have perfect and all-encompassing data, nicely curated and in standard formats, so AI software can just rock up and begin: “Making congestion a thing of the past.”

However, often this just isn’t the case. We might have at least some data, but we haven’t made it easy to find, or explained it well.

But, with my Transport Technology Forum hat on, I was able to at least make a start on fixing this gap, highlighting firstly that big data does not just mean a spreadsheet with a zillion rows.

Merging, triangulating and fusing data for different travel modes, sources and sensor types for separate policy aims is the key to adoption of AI.

Continuing my presentation, I then showed that connected vehicle data isn’t just from shiny new cars, or even just cars.

Connected vehicles can be driven by humans and are not simply self-driving cars. I shared examples of the RAC Foundation’s connected London to Brighton veteran cars run as a real-life data source, along with near miss and asset condition data from scooters and bikes from See.Sense’s rear light sensors.

I also showed that one-size data doesn’t fit all policy needs, hence the fusion I mentioned above, and the fact that data can be highly anonymised without losing its usefulness, even to meet GDPR.

Andy Graham of White Willow Consulting; credit: Andry Graham

AI-ready datasets

Some examples of data which is ripe for AI are: the origin and destination data from vehicles provided by TomTom and INRIX for measuring traffic signal delays; traction and harsh braking data from AISIN and NIRA Dynamics; road condition data from Gaist and emergency vehicle location information from HAAS Alert.

Close to my heart is the data from the eCall system already fitted to over 10 million UK cars and vans now being explored by VESOS Solutions, enabling lost vehicles whose airbags have gone off to be found in a few seconds.

I also pointed out nationwide datasets available and emerging, such as the Bus Open Data Service, the National Parking Platform, Digital Traffic Regulation Orders, digital traffic signal specifications and many more.

While these datasets are of use to future AI, they are also relevant to the DfT’s Intelligent Traffic Management Fund (ITMF). They aim to develop beacon sites in English local authority corridors, using AI and big data – as well as more traditional techniques – to achieve local policy aims.
 
Andy Graham was in conversation with John Challen.

Newsletter image: RAC Foundation’s connected London to Brighton veteran cars; credit: Andy Graham

Earlier this year CIHT took part in the Transport AI conference.