Skip down to main content

Better and locally sensitive use of data could cut traffic jams

Better and locally sensitive use of data could cut traffic jams

Published on
6 Nov 2019
Written by
Chico Camargo, Jonathan Bright and Scott A. Hale

Ever been stuck in queuing traffic which seems to have no explanation? Have you ever encountered a roundabout which creates more congestion than can be relieved?

No surprise there: city planning is a difficult job. Quite often traffic jams will be down to poor road design, or simply that new buildings and places have been developed and local road capacity hasn’t caught up.

When trying to understand the impact of a new road or a construction site, urban planners use a variety of data modelling tools to predict human mobility. These tools often rely on demographic information such as the working population in different parts of town, and help planners decide where to put new roads, transport systems or facilities such as a new Park & Ride. These methods tend to work adequately in large cities, where we have vast amounts of data about what places are where and why people are likely to want to visit a certain part of town.

Even though these tools work well for predicting commuting within large cities, in smaller towns and villages these methods tend to fall short. This is because they fail to take into account “what” is in a place – such as popular shops, green spaces, leisure facilities schools and hospitals.

In our new paper, ‘Diagnosing the performance of human mobility models at small spatial scales using volunteered geographical information’ published today in Royal Society Open Science, we look at how existing models predict traffic flows in the county of Oxfordshire in the UK, and show what kinds of trips they fail to predict. We do that by using data from OpenStreetMap, a collaborative project
to create a free map of the world, incorporating local knowledge of what’s in a place: what kind of buildings, what open spaces, how many shops, houses, etc.

We found that most existing human mobility models gave poor predictions of traffic volume when applied to Oxfordshire. For example, we found that even though there are twice as many trips from Botley and Sunningwell Ward to Drayton Ward than to Thames Ward, none of the existing models could predict that accurately. This is essentially because Thames Ward and Drayton Ward have approximately the same population, and are roughly at the same distance of Botley and Sunningwell Ward — and most tools for traffic prediction don’t take anything else in consideration when predicting the traffic volume between one place and another. We can’t say for sure if this difference in traffic volume is down to the popularity of Milton Park, Drayton Golf Club, and the Drayton Waste and Recycling Centre among residents of Botley and Sunningwell, but my guess is that they are better data reference points for potential journeys to the Drayton Ward than overall size or population.

In summary, this means that using more local knowledge, such as the OpenStreetMap data we used, could enable us to better predict how many people will travel from A to B, and whether infrastructure is able to support these journeys, not just in Oxfordshire and the UK, but across the world.