COMPETITION FOR N.1 RESEARCH CONTRACT, GROUP 09/IINF-05 INFORMATION PROCESSING SYSTEMS, DPT. COMPUTING SCIENCES, UNIVERSITÀ COMMERCIALE “LUIGI BOCCONI” DI MILANO

E4RL aims to turn videos of human activities into a resource for training robots, reducing reliance on manually designed simulations and costly robotic data collection. The project will convert first-person videos into egocentric 4D representations: 3D reconstructions of the scene and its evolution over time. It will recover the structure of the environment, the objects involved, and the person’s pose and motion, describing where elements are located, how they are used, and what role they play in completing tasks. These simple, interpretable, and efficient representations will be used to train control policies for deployment on a physical humanoid robot, using only onboard sensors and computing resources. The project will benefit from Francis Engelmann’s expertise (Università della Svizzera italiana) in compact 3D representations and Antonio Loquercio’s expertise (University of Pennsylvania) in robot learning and its application to humanoid robots.