L.INT research groups: CWI and HvA connect theory and practice through L.INT research group

In Dutch, L.INT stands for “lectureships and institutes”. It enables a lecturer, who is a PhD researcher at a university of applied sciences and often has their own research group, to collaborate with a research institution for a period of four years. The sustainable collaboration between universities of applied sciences and research organisations that this creates strengthens research throughout the entire innovation chain.

Industrial Digital Twins

Computer scientist Jurjen Helmus has been a lecturer in Industrial Digital Twins at Amsterdam University of Applied Sciences (HvA) since September 2024. Together with CWI and a diverse group of students, he creates digital versions of physical objects. ‘I digitise the physical world and bring the results to life with scale models in my lab. I take that physical aspect with me to CWI: this is unique in an institute that is best known for its many computers and bright minds. I really try to be a link between practice-oriented research at our university and the theoretical knowledge of CWI. That is my L.INT lectureship.’

His HvA colleague Nanda Piersma (now scientific director at the Centre of Expertise Applied AI) was an L.INT lecturer herself and advised Jurjen to submit the L.INT application. "She believed I should get support for this. CWI has solutions for hypothetical problems. I have practical problems for those solutions and connect industrial parties from the region to CWI." The collaboration between the institute and the university of applied sciences is going well. Jurjen: "I work with Tim Baarslag's Intelligent and Autonomous Systems team for one to one and a half days a week. They are exceptionally well-versed in theory, and I use their knowledge to solve industrial problems on a project basis. CWI colleagues also come to my lab at the university of applied sciences. That connection has great added value for all of us."

Portrait of Jurjen Helmus

The enterprising lecturer wants to build bridges between CWI, vocational and academic education, government and industry — including in the field of smart robotics. In this interdisciplinary project, students from secondary vocational education, higher professional education and university education are working together to develop an intelligent robot system that automatically recognises boxes, picks them up and places them optimally in a logistics environment.

Vocational students are the builders. They make the end effector (the robot's “hand”) and build a test site to simulate the robot's ability to hold and move boxes. HBO students are the programmers. They teach the robot arm how to move (from calculating the correct angles to planning the fastest route) and how to “see” and “feel” its surroundings using sensors.

Master's students in Applied AI learn how to teach the robot arm to choose between different boxes. Even if the robot does not know exactly which box will come next, it must determine how to place the box stably, safely and efficiently.

CWI can once again play a key role in this use case. The logistics environment is a so-called online decision problem, in which the robot only has information about the next box and therefore has to deal with uncertainty in a smart way. This ties in seamlessly with the type of issues described in the paper, such as online 3D bin packing, robot constraints, stability conditions and explainable reinforcement learning. CWI is developing new RL models that not only learn what the arm should do, but also why—for example, measuring the effect of choices between two boxes and processing information about which boxes are coming up the pipeline.

According to Jurjen, it is precisely this contribution from CWI that is crucial: "A robot arm cannot carry endless computing power. You need models that deal with uncertainty intelligently and still make robust decisions. CWI calculates for us which algorithms are feasible, how we can explain them to operators, and what the optimal balance is between computing power, reliability, and real-time applicability."

Language models for a virtual service engineer

Another project he involves CWI in, is the virtual service engineer. "We are creating an ecosystem around language models (the technology behind AI systems) that will enable us to transfer the knowledge of experienced engineers to young engineers in the field. With smart manuals and instructional videos generated by these language models. We are developing the virtual service engineer and CWI is supporting us in this. In many projects, I try to be the link between CWI's algorithms and the practical application of a social problem."

Follow-up L.INT missing

Jurjen hopes that more institutes will collaborate with universities of applied sciences, although he finds the funding limiting. "I think that fundamental and practice-oriented research are growing closer together. I think it would be healthy for future lectureships if the Ministry of Education, Culture and Science were to provide more funding for this. Until then, as a lecturer, you have to be creative and entrepreneurial in your approach to your lectureship: work hard and pioneer.‘ He therefore has another tip. ’Try to secure more job security now by making agreements for the period after your L.INT lectureship. A follow-up L.INT would help enormously."