From Robot Modeling to Physical AI: Toward Interpretable and Safe Robot Intelligence (Talk by Henrik Krauss).

Date: 2026-09-09
Time: 13:00-14:00
zoom: https://riken-jp.zoom.us/j/95043228399?pwd=aWUSqsfPkGw49BKmI1dtD20VqOX4mY.1

Team: Robot Learning Team
Speaker: Henrik Krauss, The University of Tokyo
Title: From Robot Modeling to Physical AI: Toward Interpretable and Safe Robot Intelligence

Abstract:
How do we make robots intelligent without making it impossible for humans to understand their inner working? In this seminar, I will share my international research journey from my master’s studies to my PhD, following my transition from conventional robot modeling and sensing to learning-based approaches for intelligent robots. I will introduce my work on soft robot sensing, physics-informed learning and model predictive control, before focusing on my doctoral research on learning interpretable robot dynamics from video and using them for robot control. I will also discuss my work on human visual attention and human attention-informed artificial agents, and how these different directions connect to my vision of safe and interpretable Physical AI. Beyond the research itself, I will share lessons from studying and conducting research in Germany, Hong Kong, and Japan, including experiences with international collaboration, finding creative research directions, and building an early academic career abroad.