Project Grant 2544200

This CAREER award from the National Science Foundation's Division of Computing and Communication Foundations under the Computer and Information Science and Engineering program (CFDA 47.070) supports research at the University of Texas at Austin on physics-based perception of three-dimensional human-world interactions. The project, awarded $349,323 on June 1, 2026, with completion targeted for May 31, 2032, develops a unified framework for learning physically grounded models of human-world interactions from real-world video. The research addresses current limitations in computer vision methods that produce visually plausible but physically impossible reconstructions, such as unsupported floating objects and body interpenetration. The deliverables include reconstruction algorithms that combine geometric reconstruction with dynamic simulation to estimate contacts, forces, and joint torques from monocular video, enabling more reliable analysis of human movement across applications in rehabilitation therapy, sports performance analysis, and robotic learning.

Beyond core research outputs, the project integrates education and knowledge dissemination through student training, curriculum development, open-source software release, and outreach activities connecting computer vision with biomechanics, sports science, and robotics. The scalable physics-aware perception framework is designed to advance assistive technologies, robotics, and embodied artificial intelligence systems while establishing a foundation for reliable deployment in health, safety, and robotic applications that depend on accurate understanding of physical human-world interactions.