ITU researchers secure prestigious Villum Experiment grants

Projects in infant cognition, robotics, and privacy-preserving AI receive funding for early-stage research.

Alessandro BruniAndrzej WasowskiAske MottelsonResearchartificial intelligencealgorithmsawardsrobots

Written 2 October, 2025 07:24 by Theis Duelund Jensen

Aske Mottelson, Alessandro Bruni, and Andrzej Wąsowski from the IT University of Copenhagen have received major grants from the Villum Experiment programme. The funding supports early-stage projects with high risk and breakthrough potential.

Rethinking infant communication through data
Associate Professor Aske Mottelson from ITU’s Human-Computer Interaction and Design section has been awarded 2.49 million kroner to investigate the developmental significance of infant crying.

Using customised firmware in commercially available baby monitors, the project enables large-scale data collection from families in exchange for feedback on their child’s crying patterns. This approach addresses key limitations in the field – namely, the lack of scalable datasets and inadequate audio representations – and will lead to the development of a neuro-inspired signal processing technique to explore whether infant cries contain developmental information.

Teaching robots to doubt
Professor Andrzej Wąsowski from ITU’s Software Engineering section has received 2.49 million kroner for a project that enhances robotic pose estimation. Current methods often produce either imprecise or overly confident results, risking collisions or equipment loss. By adapting Stein mixture inference – a Bayesian method previously limited to offline analysis – for real-time use, the project combines robotics and statistics to enable robots to assess their own uncertainty and navigate more safely in unpredictable environments.

Secure AI collaboration across Boundaries
Associate Professor Alessandro Bruni from ITU’s Theoretical Computer Science section has been awarded 1.97 million kroner to explore how multiple parties can collaboratively train machine learning models on sensitive data without compromising privacy. The project combines neuro-symbolic learning – which integrates data-driven models with logical constraints – with secure multi-party computation, allowing for robust and logically consistent learning in domains where data sharing is not possible.

Further information

Theis Duelund Jensen, Press Officer, phone +45 2555 0447, email thej@itu.dk