The digital twin project in York involved the creation of two models: a conventional strategic transport model and a real-time Optima model. The conventional model is a standard transport model, which is almost like a reference case, showing what we think would be happening on a typical day in the morning peak hour, the evening peak hour and then the average hour during the day. We have lots of sensors, from automatic traffic counters (ATC) and counters on some of our traffic signals, so we can see how many cars are turning left, right, etc.

We also then get real-time signal data and TomTom journey time data, so we know how long it's taken to travel. We've also recently added in some Highways England count data. The programme works out what is happening, and we can use that to forecast up to an hour in advance. It also works on a rolling horizon basis, so the data, the latency between the data and the software in real-time models are updated every five minutes with the citywide simulation.

On top of that, each year we calibrate the real time and conventional models. We collect all that data every minute of the day, every day, so we build a picture of what's been happening in a typical month to calibrate the real time model. We then play back the recorded data through the model to check its forecasting is sufficiently accurate. It’s important to update the model every year because the Highway Network and travel patterns change. 

Real-time, quickly

The project came about as a way of making the best use of our existing infrastructure. Rather than building an expensive new infrastructure, which takes a long time to do, we’re squeezing everything that we can out of our existing infrastructure.

York is a city of manageable size, and this would be a much bigger challenge in a bigger city. We're lucky that we had a lot of the sensors in place and a lot of ATC data. We had good coverage, so we were well placed to do that. Lots of things were coming together. We were updating our signals, so we could put in live communication and see what's happening in real time. That's quite rare: often, to get signal data, depending on how new or old your installation is, you sometimes have to go with a laptop and physically plug it into the signals. Or sometimes you have to request it from a supplier and get them to transfer the data.

60-minute predictions

The main thing that we use the real-time traffic model for is managing our network day to day, so it's a traffic operations tool. It means we can now predict up to an hour in advance what's happening. We now have everything built in: actual journey times, what's happening with the signals, what's happening with the real volume of traffic, not just a sample. That gives us more accurate data, so if we see a corridor getting more congested than normal, we can try and step in and change the signal times to try and relieve congestion. We've got much richer data than we had before.

One of the most important things we’ve learned is being able to get the data from traffic signals. That's a really big thing, especially if you've got lots of signals that are running dynamic strategies. If they're changing a lot, you need to know what's happening, getting good coverage of either ATC or other ways of counting vehicles and thinking how you can use things you already have.

Chris Cuckson was in conversation with Craig Thomas 

York’s real-time model was awarded the CIHT 2022 Transport Planning award. Read more about the model and the partners including the Wood Group who led the consortium here.

Register your interest for the 2023 CIHT Awards: Get ahead of the competition and be the first to know about the awards by clicking here.