Artificial Intelligence (AI) has the potential to solve some of society's most important challenges, for example the mounting pressure on healthcare, while at the same time enriching and improving many other aspects of our society, such as transport. In short, the impact of AI is enormous, and it may even become the foundation of our future. Therefore, it is extra important that we are sure that all developed AI systems and techniques are completely reliable. The ROBUST consortium, in which TU Delft plays an essential role in the field of fundamental and applied AI knowledge, investigates how AI-based systems can be made safer and more reliable – and thus better contribute to social issues.
A growing ecosystem
ROBUST has been pre-selected as one of the first two Long Term Projects of the NWO, and builds on the network of the Innovation Center for Artificial Intelligence. This collaboration has already produced 30 flourishing labs, which are now reinforced with seventeen new labs. Arie van Deursen, involved in ICAI as a member of the advisory board, is excited to see the project grow:
The ICAI lab formula, in which five PhD students collaborate closely with companies or societal organizations for five years, has proven a great success, also for TU Delft. Within the ROBUST long term program we will create many more of those fruitful collaborations: in total with over 50 different partners. There is critical mass and cross fertilization — the perfect way to address the crucial challenge of trustworthy AI.Arie van Deursen, scientific director of one of the existing labs and co-applicant of the ROBUST proposal
Collaboration with industry is central to ROBUST. They not only provide a third of the financial contribution, but also the relevant, interdisciplinary research questions, access to relevant data, domain expertise and the context in which solutions can be tested.
A good example of this is the ROBUST-RAIL Lab, in which TU Delft is collaborating with Utrecht University, ProRail and NS. Railway infrastructure is extremely expensive, especially the marshalling yards near stations, for example, which are often located in the middle of town. A short lead time of the logistic planning process is of great importance to make the best use of this scarce infrastructure, to make the transport process more robust and to gear the transport volume even better to the expected passenger flows. Together with Utrecht University, NS has already made an enthusiastic start with a number of pilots that have demonstrated the potential of algorithmic support. It is one of the many examples of how AI techniques contribute to affordable and sustainable social developments. And that knowledge can ultimately be used for larger logistical puzzles, but also in the interaction between the human and the algorithmic planner. Or what about the question of how AI can make our railway network more agile, specifically in dealing with uncertainties or disruptions?
The great thing is that we kill two birds with one stone: we get to work on challenging algorithmic puzzles that contribute to the big questions surrounding the use of AI techniques – and at the same time we contribute to efficient public transport.Mathijs de Weerdt, leader of this ROBUST-RAIL lab
We have experienced that practical applications are an effective driver for research and that the results are actually implemented, as is currently happening at NS, when universities and companies work together for a long period of time in accordance with the idea behind the LTP.
Better understanding of complex matter
Another example is the lab in the field of collaborative knowledge engineering. In it, specialists from TU Delft in the field of ontology engineering and human computing work together with colleagues from Royal DSM, Biomax and Maastricht University. This collaboration of science and industry develops methods to enable people to construct, maintain and integrate structured knowledge, such as ontologies (hierarchical ordering in and between different data sets), in order to be able to use them on a company-wide scale. By using data in this way in an even smarter way, and enriching it with human computing, the power of all that knowledge comes within reach of industry and science. This way, companies can make better use of their data in two ways: the datasets are more accessible for their employees, but they also offer much more useful knowledge and insight. This technique could first of all lead to a breakthrough in food and biotech science. Something that could eventually lead to a clearer and more fundamental understanding of how complex knowledge systems can be made accessible in a reliable manner.
For us, this is a great opportunity to test our long experience in data management and human-in-the-loop systems in the real-life use cases of our partners. In this way, we will obtain even better AI methods for reliable knowledge and contribute to scientific developments in important sectors.Christoph Lofi, researcher in collaborative knowledge engineering
In one year's time, the elaboration of the proposal by the ROBUST consortium will be assessed by the NWO, after which a final contribution will complete the research budget of 95 million euros. After that, scientists and companies can work for ten years on a reliable foundation for a society in which AI is central.
It is exciting to see how scientists from TU Delft will be able to provide research and innovation contributions to the important scientific and societal challenges of this long-term programme. It shows how the TU Delft approach to combining research in-AI and with-AI is an excellent basis for impact in society.Geert-Jan Houben, Pro Vice Rector Magnificus AI, Data and Digitisation
ROBUST is made possible by Prof. Maarten de Rijke of the University of Amsterdam and ICAI. The co-applicants are Prof. Dr. Arie van Deursen (Delft University of Technology), Prof. Dr. Mark van den Brand (Eindhoven University of Technology), Prof. Dr. Bram van Ginneken (RadboudUMC), Dr. Eva van Rikxoort (Thirona), Prof. Dr. Clarisa Sánchez Gutiérrez (Amsterdam University) and Prof. Dr. Nava Tintarev (Maastricht University).