The art of getting systems to speak to each other

Paul Harvey is a TransiT co-investigator and expert in autonomous and adaptive systems based at the University of Glasgow. We asked him about his work.

What’s your role at transit and what does this involve?

I am the work package lead for ontologies at TransiT. My job is to bring together the different academics who are actively working on, developing and using ontologies. I also work closely with TransiT’s demonstrator teams, to make sure we are all in harmony and able to communicate with each other clearly and consistently.

What are ontologies?

Ontologies are formal, mathematically-described representations of knowledge.

A very simple example is, if I say ‘car’ and you say ‘car,’ do we mean the same thing? My car might not have any information about the fuel source it uses. And your car might have zero information about how many passengers it expects to hold – but we still use the same word.

An ontology is a formal representation of this type of information, so that people and machines can unambiguously describe it and understand it.

As well as describing things, they enable something called data interoperability. This is where systems in one domain can clearly interact and understand what systems in another domain mean. Ontologies are the glue that enables this interoperability.

We can also use ontologies to reason. We can ask them questions and they can give us back answers that are completely understandable and completely explainable in a way that we don’t really see with things like machine learning models.

Why are they important to TransiT?

Digital twins are essential for the transport sector to decarbonise. But our transport system is not as joined up as, for example, a national energy system or telecoms network.

It consists of many different stakeholders from logistics providers and local councils developing roads, to the Department for Transport, regulating and setting the scene at the national level.

Different stakeholders like these might use different digital twins. But to decarbonise the transport system as a whole, we need to find a way for individual digital twins in different sectors to all share information, context, knowledge and decisions in a way that is comprehensible and consistent. This is what ontologies provide.

What’s your background?

I’m a computing scientist by background with a PhD in distributed and adaptive systems. After my PhD, I got a fellowship to go to Japan for a couple of years. I then joined a Japanese company called Rakuten Mobile, where we created a research lab looking at autonomy and how it can be applied in complex adaptive (networked) systems. We started looking at ontologies as a way to ground the decisions that would be made by artificial intelligence in a way that was clear, consistent and explainable. That’s when I started working with ontologies as the kind of formal knowledge base to bring all these self-adaptive processes together.

Then I came back to the School of Computing Science at the University of Glasgow. My focus here is the technologies and skills needed to both design and operate fully-autonomous systems in safe, meaningful and sustainable ways.

I’m also vice chair of a telecoms standards focus group called the Focus Group on AI-Native Networks.

What do you hope to achieve at TransiT?

The key success would be to demonstrate how ontologies make federated digital twinning possible. This is where multiple digital twins are linked together in a network to work together towards the same goals.

It would be fantastic to develop a transport ontology output that could be picked up and built upon as the years go by, for example by the Department for Transport.

Another goal would be to improve the visibility of ontologies as a queryable entity – in other words, they can answer complex questions.

This highlights how ontologies can themselves be a form of digital twin – but with very low computational overhead and high degrees of accuracy.

On Thursday 3 September at One Great George Street in London, our event Transport Ontologies: State of the Nation will explore the critical role of data in decarbonising transport. Find out more on our Eventbrite page here.