Research
The Ego-AI Lab at Ulm University, led by Prof. Xi Wang, researches egocentric vision and embodied AI: understanding human intent from first-person video, learning world models, and building robots that perceive, reason and act.
First-person video captures the world as people experience it: their gaze, their hands and the objects they interact with. We develop models that recognise activities, anticipate intentions and infer the goals behind everyday interactions, bringing human behaviour patterns and common sense into machine perception.
Acting in the world requires anticipating it. We learn generative world models that predict how objects move and how scenes change when people or robots interact with them, for example goal-conditioned 6-DoF object motion generation in 3D scenes, and use them for planning and control.
Egocentric cameras move constantly and observe a world that moves with them. We work on visual SLAM, dynamic reconstruction and 4D scene representations that recover geometry, motion and semantics over time, forming the spatial backbone for embodied reasoning.
Perception is only useful when it leads to action. We connect multimodal reasoning with robot control, study dexterous manipulation of everyday objects and explore how multiple agents, human and artificial, can collaborate in shared environments.