Toward Human-Centered Physical AI

Physical AI has attracted growing attention in recent years, yet we seldom address how such systems can truly empower humans and be deployed in the real world. In this talk, I introduce Human-Centered Physical AI, an emerging interdisciplinary field that integrates perception, learning, decision making, embodiment, and multi-agent physical and virtual interaction to advance the design, standardization, evaluation, and real-world deployment of Physical AI systems. By placing human needs at the core, this paradigm seeks to enhance safety, autonomy, trust, and quality of life. I will present assistive feeding as a use case and share recent findings that illustrate the principles and challenges of Human-Centered Physical AI in practice.