The Age of Intelligent Algorithms: AI at PTV

PTV experts discuss about AI

Artificial intelligence (AI) is the theme of Science Year 2019 as announced by the Federal Ministry of Education and Research. No wonder: Intelligent IT systems have an impact on all areas of our lives, offering a multitude of opportunities. This topic is also on PTV’s agenda – intelligent algorithms and innovation have always been part of the company’s DNA. In order to explore this complex issue, we met with the PTV experts in this field for an interview: Klaus Nökel, Head of Innovation, Christian Mähler, Principal Software Engineer, Werner Heid, Director Methods and Axel Gußmann, Director Data answered our questions about the potential of AI.

Compass: AI is currently on everyone’s lips. But the subject itself is not that new, is it?

Klaus Nökel: That’s right, I did my doctorate on the subject of AI in 1989 and I have been dealing with this topic for many years now. AI experienced the first hype cycle in the early 1990s. At that time, research had pushed ahead too far with its vision on AI and most of the ambitious goals could not be met. So AI fell out of the limelight for a few years. About ten years ago it could regain momentum, especially in the last two years AI has attracted a lot of attention.

Compass: Why is that?

Klaus Nökel: With an unswerving dedication to driving AI forward, a small group of researchers continued to work on the methodological basics. Improved methodologies and faster computers allowed them to implement some of the ideas that only existed on paper in the 1980s.

Compass: How would you define artificial intelligence?

Christian Mähler: For me, AI is the attempt to imitate intelligent, human behaviour using a computer.

Axel Gußmann: In principle, it is a collective term for different sub-areas, which include robotics, sensorimotor skills, machine learning and empathic intelligence.

Werner Heid: There are different requirements that an AI system must meet. Helsinki University, for example, defines adaptiveness as a key requirement, i.e. the ability to adapt without human intervention. And autonomy, which means that the system can run autonomously over a longer period of time and make decisions independently. It should also be able to handle probabilistic information, i.e. probabilities.

Compass: What role does machine learning play in this field?

Christian Mähler: Machine Learning is a technique that is part of AI, it’s a sub-area that has increasingly gained ground in recent years. Here, artificial knowledge is generated from experience. IT systems are enabled to recognize patterns in data and then draw conclusions from them.

Klaus Nökel: In principle, you feed as much relevant data as possible into your model and train it. The algorithm learns from this, determines correlations and uses them solving for new problems. However, the whole thing only works with a large number of training examples and sample solutions. 100 or 1000 examples are not enough, the computer needs millions of records.

Axel Gußmann: Big data systems provide an important basis for this type of learning. Therefore it is so helpful that in today’s age of digitisation people reveal their behaviour and leave behind a lot of digital traces, for example through apps or the like. These data tracks provide what the machine learning structures urgently need, namely huge amounts of training examples.

Compass: There is also a lot of talk about the dangers and risks of AI. One of the most prevalent fears people have is that of losing control. What do you think of that?

Klaus Nökel: The biggest danger I see is that the world is getting more and more AI-friendly. I have already heard scientists talking about the need to clean up cities in order to prepare them for autonomous driving. So no more trees on the roadsides, everything paved etc. Today, many people can no longer imagine buying an item that is not in Amazon’s assortment. Our consumer behaviour is constantly steered in a certain direction, channelling demand towards items that are digitally recorded. Processing thus becomes easier. But I don’t think that’s a good idea.

Christian Mähler, Principal Software Engineer, Klaus Nökel, Head of Innovation, Werner Heid, Director Methods und Axel Gußmann, Director Data (f.l.t.r)

Compass: How does PTV use artificial intelligence?

Christian Mähler: Our software PTV Optima already uses machine learning algorithms for prediction calculation. These algorithms also help improve the accuracy of the model. And there are many other aspects where we deal with AI. Together with external scientists, we are researching various AI-related issues. One of our partners is the FZI Research Center for Computer Science here in Karlsruhe. I also see a lot of potential for our products in the future.

Werner Heid: We are particularly interested in AI methods for determining demand for new ridepooling services, or for the use of autonomous vehicles, or for driving time prediction, for example in the field of dynamic route planning.

Axel Gußmann: Our data department has recently developed a method for mode detection of movement data, i.e. the appropriate allocation to the means of transport. So our next step is to expand the use of AI across our range of products.

Klaus Nökel: With the algorithms and methods that we have used over the past 40 years until today, we are always very close to what machine learning does today. And we will continue to do so. We are at the forefront of exploiting new developments such as AI. Artificial Intelligence will help us to develop our transport models faster, to calibrate and optimise them even further and to make our forecasts more realistic.