KAIST and Nvidia are joining forces to develop physical AI that learns human movement and physical intelligence, combining KAIST's robotics expertise with Nvidia's AI computing platform. The announcement is the institute's second major Nvidia collaboration in short order, following a partnership on agentic AI tailored to the Korean language and domestic industries.
KAIST announced Sunday that it will establish an Nvidia AI Technology Center (NVAITC) with Nvidia to pursue joint research in physical AI.
The partnership will take shape through the creation of a "Human Physical Intelligence Technology Center" between KAIST's Department of Mechanical Engineering and Nvidia. KAIST's Human Physical Intelligence Research Center will serve as the core hub for the joint research, with plans to gradually expand collaboration into all areas of physical AI — including wearable robots, humanoid robots, digital twins and manufacturing.
The Human Physical Intelligence Research Center was established around the laboratories of professor Gong Gyeong-cheol, who specializes in wearable robotics, and professor Kim Jeong, who focuses on biorobotics. The two professors serve as co-directors of the center.
KAIST has accumulated large-scale human motion data and gait and motion control technologies through its wearable robot and biorobotics research, and has experience commercializing that work into real products and industries. Gong's research team in particular has earned global recognition for designing and developing wearable robots that support the rehabilitation and daily lives of people with mobility impairments, validating their performance in real-world user environments. The team has also driven the commercialization of wearable robot technology through Angel Robotics, a spinoff company that translates the research into commercial products and services.
The joint research will build on accumulated human motion data and robot control technologies to lead development of a human motion foundation model.
On Nvidia's side, Charles Chung, a senior manager at the Nvidia AI Technology Center, will participate in the project, providing technical advice and developer support. Nvidia also plans to run research and education programs using its Omniverse and digital twin platforms.
The primary goal of the first phase of joint research is to build a "Human Motion Foundation Model."
A human motion foundation model is a generative AI-based model trained on large-scale human motion data that can understand, predict and generate a wide range of human movements. It is expected to serve as a core foundational technology enabling wearable robots and humanoid robots to read a user's intentions and movements more accurately and respond naturally.
The technology is expected to find applications not only in wearable robots that support the rehabilitation and daily lives of patients with mobility impairments, but also in humanoid robots working alongside humans, motor function analysis and digital healthcare.
Ultimately, the researchers aim to use an AI framework to uncover the principles by which humans plan movement and control muscles and joints, then build next-generation robot systems that can overcome mobility impairments and extend human physical capabilities.
The physical AI technologies developed through this research are expected to spread beyond humanoid and wearable robots into a range of national strategic industries, including manufacturing, shipbuilding, steel, logistics, construction, defense and healthcare.
"In the age of physical AI, competitiveness comes not from AI itself but from domain expertise and data rooted in a deep understanding of humans and robots," KAIST President Bae Choong-sik said. "By combining KAIST's human-centered research with Nvidia's world-class AI infrastructure and physical AI technology, we will develop physical AI that better understands and assists people, and grow into a hub that leads global physical AI research and industry — not just in Korea, but worldwide."
nbgkoo@heraldcorp.com
