DTU Conducts Crash Tests on AI-Based Products
In the past, artificial intelligence (AI) helped us with simple things like correcting spelling or cleaning with a robotic vacuum cleaner. AI has now become such a large part of our daily lives that the technology lies at the heart of solutions to optimize everything from our health and energy consumption to urban planning and infrastructure. However, very few AI technologies are tested under real-life scenarios before they enter the market. This has resulted, for example, in several accidents involving self-driving vehicles.
As part of a new billion-kroner European investment, Denmark is now leading the way in influencing the use of AI so that the technology is used in a responsible way. European companies, public authorities, and research and education institutions are being given access to test AI-based products in areas such as health, production, agriculture, mobility, and energy.
Sound development
The aim is to save time and money for companies and give investors and customers such as municipalities, regions, and national authorities extra security that they are purchasing proven and safe products. DTU coordinates European cooperation on smart cities in a test and experimentation facility (TEF) called CitCom.ai, which deals with issues such as energy and mobility. Small and medium-sized enterprises can test software and hardware products and robots here in physical and virtual environments. These could be self-driving cars, algorithms to secure data, or municipal robotic tractors.
“There are currently no processes for ensuring that AI and robots are safe and comply with the standards that are acceptable to Denmark and the EU. There is simply a gap from the time a company, authority, or researcher has a good idea until it enters the market. It is a bit like building a rollercoaster without ensuring that it is safe, or releasing health products on the market without testing them. That is why we are crash testing AI and robots now, so they can be safely integrated into society,” says Martin Brynskov, who heads DTU Connecting Communities, a research group under the Section for Dynamic Systems at DTU.
Green and digital Europe
He works at DTU Compute, where he heads and coordinates CitCom.ai, which focuses on power, mobility, and connecting devices, systems, and people to the Internet or each other. The EU project has 32 partners in 11 countries. In Denmark alone, these include the Confederation of Danish Industry, Force Technology—a technology, consultancy, and service company—and several municipalities.
The project is part of the EU’s Digital Europe programme, which seeks to facilitate the broad roll-out and use of digital technologies in society. The aim is to help strengthen Europe’s potential to compete globally and create a greener and more digital Europe. This makes the project an important element in regulating AI in the EU.
“The project is the first of its kind in the world, and involves us having a quite large portfolio. What Europe is doing is also of interest to the world. Europe is a leader in the practical regulation of AI and is the largest common market in the world. The rules which apply here are therefore significant,” says Martin Brynskov.
Test site with 12 km of roads and paths
Together with the other project partners, DTU advises companies on the ethical, safety, and economic factors associated with using AI. For example, they examine whether they are in compliance with the law and collecting data lawfully.
The technologies are tested at a wide range of facilities throughout Europe. One site is the DOLL Living Lab in Albertslund, where intelligent traffic lights to reduce vehicle CO2 emissions and congestion are tested. The living lab is one of Europe’s largest outdoor laboratories and has 12 km of roads and paths. Another test facility, Aarhus City Lab, is testing an autonomous robot that uses image recognition to find and pick up cigarette butts in the urban space. The long-term aim is to minimize waste in the city.
Can it all break down?
As part of the testing and experimentation facility, DTU researchers are involved in several projects testing energy systems together with companies. One of these is Energy Cool in Fredericia, which seeks to make the cooling processes in small and medium-sized data centres with up to 500 servers as climate-friendly as possible. This is done in part through AI-based control, whereby surplus heat is passed on to the district heating grid or used internally.
“More decentralized flexible and robust systems will be needed in the future to control both energy and mobility. Not only for climate reasons, but also for safety, economic and practical reasons. Our infrastructure is currently made up of central systems that span regions. It is vulnerable in a time of hybrid war and unstable energy supplies. At the same time, the use of AI also presents some dilemmas: Are we becoming more or less robust by adopting modern technology? Will it all break down if we are hacked? It is extremely complex and exciting. It is therefore important to crash test systems and products to ensure that we are as secure as possible,” says Martin Brynskov.