Humans + machines

Professor Jürgen Leohold

Piloted driving will soon be perfectly normal in heavy highway traffic. And the automobile of the future will also look for its own parking space once the driver has got out. What we will definitely not see, however, at least not in the next ten years, are self-driving cars capable of navigating every traffic situation fully automatically. We are not talking about a revolution here, but rather the systematic evolutionary development of driver assistance systems that are already fitted into our vehicles today.

Professor Christian Gerdes

After all, we shouldn’t forget the amazing cognitive achievements that human beings are capable of, particularly in complex situations. We engineers can’t duplicate that yet, so we have to begin by simplifying the problem. We can do this, for instance, by limiting the car to low speeds or certain stretches of road or automating individual driving tasks step by step.

Professor Jürgen Leohold

Of course, we researchers are also working on the underlying requirements for automated driving in urban traffic. This sort of vehicle would have to be able to deal with every situation that might arise in real life – not only in fine weather, but also when it snows, for instance. What’s more, machines have to learn the kind of communication that we humans take for granted, for example how we use eye contact and gestures to signal to each other who should move first into a crossing.

Professor Christian Gerdes

I am convinced that everybody would be happy to hand over the steering wheel in certain situations, such as stop-and-go traffic. We just have to make sure that people on board an automated vehicle feel safe.

Professor Jürgen Leohold

That’s why we are conducting intensive research into how the display and controls in these vehicles should look. They should be as simple and intuitive as possible. The driver must always be able to tell from the display exactly what the vehicle is doing at that moment. For example, all maneuvers will have to be indicated before they are carried out.

Professor Christian Gerdes

Incidentally, the Volkswagen Electronics Research Laboratory (ERL) of the Volkswagen Group of America in California and Stanford University are also looking together at how automated driving should feel. This includes analyzing what professional chauffeurs do to make their customers feel safe.

Professor Jürgen Leohold

We are looking not only at acoustic or optical signals, but also at how customers experience longitudinal and lateral acceleration. This will determine whether people trust a technical system or not. It may even be possible for an autopilot to adapt itself to a customer’s individual driving style.

Professor Jürgen Leohold

We are only automating those functions where we can say with absolute certainty that this will improve traffic safety – for all road users.

Professor Christian Gerdes

We believe that, one day, it will be possible to design vehicles that avoid every accident, at least as far as the physics will allow. Much lighter vehicles could then be designed because they would need fewer safety features, which in turn would cut fuel consumption.

Professor Jürgen Leohold

A great deal of research and development is still necessary, however, for example in the areas of sensor technology, software and testing methods.

Professor Jürgen Leohold

An automated vehicle is always also a networked vehicle, for example because it has to be able to access the latest map updates at any time. But networking can also help automobiles to make the right decisions within milliseconds, because they can take into account information from other vehicles.

Professor Christian Gerdes

This sort of self-learning computer program is no longer science fiction. The question for us as researchers is how to organize this learning. In the future, automobiles will probably be connected to a server that will allow them to benefit from the experiences of other vehicles linked up to the same network.

Professor Christian Gerdes

Many people are prepared to supply certain information in return for an easier life or real added value. We see that, for example, with social media users.

Professor Jürgen Leohold

What’s important for Volkswagen is that customers can decide for themselves whether they are willing to supply information from their vehicle, as well as how much information they make available. It’s about transparency and trust. Anonymized vehicle data will be used exclusively to improve traffic safety. The car won’t turn into a hungry data monster.