that nature is the best blueprint for the development of modern automobiles.
    Johannes Barckmann
    > Story


    people all over the world feel a bit surer when they drive off.
    Ralf Fröhling, Patrick McFadden, Robert Gastl.
    > Story


    that the pedestrian safety not only comes from the road but from the car also.
    Maren Finck.
    > Story


    of technologies to get the best out of design and function in order to develop a more enjoyable driving pleasure.
    Hans Haug.
    > Story

The person in the machine.

A learning-based concept for automated driving.

Highly automated driving, ADAS, autonomous driving – terms that have been in people's minds for a long time now. Only the experts know the exact difference. But when we hear them, we all understand more or less the same thing: hands off the steering wheel! Suddenly, "keep your eyes on the road", something we all had to learn, no longer seems quite as important. We expect it to bring us more comfort, so that we are relaxed on arrival, creativity instead of wasting time at the wheel.

But there is a long and stony road to travel before we get there. Every further innovation brings with it unresolved questions and seemingly insurmountable hurdles. One of these hurdles is that the car of the future will need to be able to reliably recognise the behaviour of other road users. A hurdle that has now been tackled, thanks to EDAG's experts.

"Forward-looking algorithm" is how Alexander Hirschle modestly describes it. However, what the electrical engineer graduate is working on is nothing less than the core elements of automated driving. Even if the vehicles of the future are fitted with better and better sensors, from mono cameras and stereo cameras through to radar, each car will nevertheless require a "brain". An algorithm that defines what to do in what situation. A set of rules subject to continuous opmisation.

Empirical values for the machine

Take the example of driving on a dual carriageway or the motorway. To the right, you see the acceleration lane for traffic joining the road you are on. You know that a vehicle in this lane that is driving parallel to you will very shortly change to your lane. Instinctively, you change lane or brake to leave room for the car to make this manoeuvre. This is a basic rule that you have learnt. Your driving experience means that you are familiar with and able to correctly assess just about any situation. For you, there is practically no difference between a normal road and an acceleration lane to your right – regardless of what the road and your surroundings look like. Or to take another equally familiar example. You notice that, behind a heavy goods vehicle some way ahead, there is a vehicle that is constantly "twitching" over towards the left. The driver seems to want to pull out into the other lane and overtake the HGV, but is prevented from doing so by passing traffic. The closer you get to the driver, the stronger you suspect that he might soon pull out in front of you. Your logical reaction is to slow down in good time and be ready to apply your brakes.

Hard work for vehicle technology

As simple as this might sound because in reality it really does feel simple to us, this is very hard work indeed for vehicle technology. What is needed is a very finely thought-out algorithm, as capable as we human beings are of correctly assessing driving situations. More in fact - capable of predicting situations. A "human in the machine" if you like. 

The algorithm developed by Alexander Hirschle and his team in Ulm becomes all the necessary data from vehicle sensors in the form of an environmental model and object list. During the first stage, the individual vehicles in the immediate vicinity are checked for striking features or anomalies. These "features" should provide information on driving behaviour. The "twitching" over towards the left described above would be one of these features. During the second stage, the current driving situation is calculated on the basis of this information. The statistical analysis then predicts the most probable driving behaviour in the cars in the vicinity. In the final stage, the car then plans the ideal route, already reacting to other vehicles.

Being part of the future

What looks like a "brain" in a car is in fact not so very far removed from the original itself. Like the human equivalent, the EDAG experts' system is always learning something new. Every new driving situation is re-evaluated and individually calculated. "The algorithm makes it possible to have a kind of chauffeur that already drives autonomously here and there. A further key component in the journey towards the autonomous vehicle. It is a good feeling to have created part of the future," said Alexander Hirschle.

Your contact for this subject

Alexander Hirschle
E/E Embedded Systems
Tel.: +49 731 140 595 00