21/05/2026
๐๐๐ฟ๐ผ๐ฝ๐ฒ, ๐๐ต๐ฒ ๐๐ผ๐บ๐ฒ ๐ผ๐ณ ๐ฅ๐ผ๐ฏ๐ผ๐ ๐๐๐น๐๐๐ฟ๐ฒ ๐ช๐บ
The article by Riccardo Oldani published in ๐๐ ๐
๐๐๐๐ก๐ explores some of the key challenges in European AI and robotics research through the lens of euROBIN project, the Horizon Europe-funded network of excellence aiming to build a shared ecosystem of robots, software, and methodologies.
Among the Italian contributors featured in the article are Alberto Finzi, head of the AI and Cognitive Robotics research line at PRISMA Lab, and Riccardo Caccavale, both involved in euROBIN activities through the CREATE Consortium.
One of the main goals of the project is โto make it possible to transfer and reuse technologies developed in laboratories, overcoming the fragmentation that still slows down research today,โ explains Finzi. Robotic solutions often remain tied to the specific prototypes or platforms on which they were originally developed, making them difficult to replicate or integrate with other systems.
In this context, learning becomes a central element of the project, especially through approaches that allow robots to acquire and transfer skills more naturally across different platforms. One of the most relevant aspects discussed is learning by demonstration, which allows robots to observe human operators and emulate the sequence of actions required to complete a task. This approach could make robotics more accessible even to people without programming expertise.
On the research side, one of the main challenges concerns the transfer of robotic skills across different platforms. โWe want skill transfer to be as simple as possible,โ Finzi explains, โand to study techniques not only for training, but also for transferring behaviors and capabilities from one robot to another.โ
Finzi also stresses the importance of a โhuman-centeredโ approach, in which robots are expected to recognize human intentions and plan their actions accordingly, especially in collaborative environments where humans and machines share spaces and activities.
Riccardo Caccavale discusses the contribution of neurosymbolic systems, where reasoning capabilities and data-driven learning are integrated within the same robotic framework. โWe combine the ability to explicitly represent knowledge, typical of symbolic systems, with neural network learning,โ Caccavale explains. โThis allows us to obtain robots capable of โreasoningโ about their own actions.โ
The final part of the article focuses on cascade funding and on the coopetitions promoted within euROBIN: collaborative competitions in which research teams are evaluated not only on their own performance, but also on their ability to integrate and reuse software modules developed by other groups. An approach designed to encourage interoperability, knowledge sharing, and collaboration across the European robotics community.
