Speaker
Description
AI models are helping robots learn how to move and handle objects, but they still struggle to explain why they make certain choices or how they can apply what they’ve learned to new materials. To fix this, we are creating a system that connects a robot's brain to a digital "knowledge map." This map acts like an evolving rulebook of the physical world, tracking how objects interact and how actions lead to specific results.
Our research focuses on a humanoid robot with five-fingered hands learning to handle tricky, loose materials like seeds, grains, and dry soil. These everyday tasks—like pouring, scooping, and weighing—are actually highly complex because the robot must understand hidden physics, like how the material flows and slips. We train the robot by recording humans doing these tasks using motion-capture suits and high-tech gloves, matching their movements with video data and spoken instructions. Ultimately, we believe that giving robots this structured, map-based way of thinking will make them much better at adapting to new tasks while allowing them to explain their actions clearly and reliably.
| Pracovisko fakulty (katedra)/ Department of Faculty | KAI |
|---|---|
| Tlač postru/ Print poster | Budem požadovať tlač /I hereby required to print the poster in faculty |