Autonomous driving will bring fundamental changes to the way traffic works. For instance, in the future we may have to take detours to help all road users reach their destinations faster and more safely. We spoke with traffic researcher Sven Henning from Paderborn University in Germany.
Sven Henning, your research looks at the traffic of tomorrow and beyond. Will we all soon be driven around by autonomous cars?
I’d estimate that autonomous vehicles will be a normal sight on the roads in 20 years’ time, or perhaps even sooner. It’s hard to predict when the technology will become truly widespread. That will depend on how forcefully the manufacturers push this type of transport, and also on the relevant legislation.
You recently published a paper on autonomous traffic management with the complex title "A model-based study of the reliability of algorithm-detected critical points in road networks”. Can you sum up the findings for us?
The idea is to view the road network as a static topology – a bit like an urban landscape, but without road users. Taking the road network, the routes followed by individual transport axes, and the geography of intersections, it is possible to identify critical points in the road network. These could be, for example, the places where traffic jams occur more often purely because of the topology of the roads. I have developed various algorithms here that enable us to make relatively reliable traffic forecasts without observing actual traffic. Of course, these results are particularly useful for autonomous driving – for calculating the best route, for instance, for improving driver-assistance systems and, ultimately, for keeping the traffic moving.
Don’t most drivers already choose the best route, for instance when travelling to work in the morning?
Not necessarily. There are usually numerous different ways of getting from A to B. Many people always take the same route because they believe it is the shortest. It is possible to collate this information in an algorithm and relate it to a specific road network. This may reveal, for example, that drivers choose a particular road much more frequently than others because of a particular topology – because of the course taken by the road. Autonomous vehicles could avoid such routes in future to reduce traffic levels there.
So you’re out to change people’s habits...
As we know, people are fairly unpredictable when they are driving. Every driver has certain preferences, their own personality, and moods that change from one day to the next – these are all factors that influence the way people drive. By contrast, an autonomous driving system is completely predictable. I can get every vehicle to drive the way it needs to for the whole system to work. This opens up whole new possibilities, such as intersections without traffic lights.
Intersections without traffic lights? How would that work?
I call the whole thing “active intersection management”. At every intersection there is an “intersection manager” that has an overview of all the vehicles approaching the junction at any time. The system calculates the ideal path for each vehicle and the optimum speed to ensure that everyone can negotiate the intersection without stopping or colliding with another vehicle. In my model, the system registers the car at around 100 meters from the intersection and might slow it down to 40 km/h for a distance of 30 meters, for instance, and accelerate it back up to 50 km/h for the rest of the way, until it passes the junction through a gap in the cross-traffic.
Paderborn is not London, Paris or New York. Is this city really a good base for your observations?
Paderborn is where I live and work and where I experience traffic first-hand every day. But apart from my personal impressions, Paderborn also has many of the characteristics of a typical city. These include road networks that have developed organically, beltways and a dense downtown area. I don’t limit my research to Paderborn though – I include other cities as well, and I have also developed imaginary road networks of my own. This way I can concentrate on a particular characteristic of a road network and find out more about it.
Does your topology model really work?
Initial studies show that the model does indeed work. I was able to demonstrate in my simulation that pinpointing neuralgic points in the road network helps reroute traffic and increase traffic flow. That is very promising. Another implication of the model is that individual vehicles may have to make a small detour occasionally so that the traffic as a whole is not disrupted.
Is this what’s meant by “social traffic management”?
Let’s say that it’s no longer about each individual focusing solely on their own destination, which they obviously want to reach as quickly as possible. The system takes account of the destinations of the other road users as well. The aim is for everyone to reach their destination as quickly as possible. But that also means that your journey might have to take two minutes longer than usual to ensure that the traffic as a whole keeps moving. In this system, the egotistical view of the individual driver is replaced by a group perspective. To that extent, we are indeed heading for social traffic management.