UT Austin Wins $30M NSF Center for Human-Robot Co-Adaptation

A new NSF center will study long-term human-robot adaptation

The University of Texas at Austin announced on August 27, 2026 that it will lead a new National Science Foundation Science and Technology Center focused on human-robot co-adaptation: how people and assistive robots learn from one another and change their behavior over extended periods in real environments.

The five-year, $30 million award will support the Center for Human and Robot Co-Adaptation, bringing together 39 researchers from six universities plus industry collaborators. This is a newly funded research initiative, not a commercially released robot or foundation model.

The research moves robots beyond short lab demonstrations

Most robotics benchmarks and demonstrations focus on whether a machine can complete a task under controlled conditions. The new center is targeting a harder problem: whether robots can remain useful as a person's needs, habits, physical capabilities and environment change over months or years.

UT describes examples such as an assistive robot learning not only how to set a table but when a particular household is likely to need that task, or adapting its reach to a person's physical limitations during rehabilitation. Researchers will study both directions of adaptation because humans also change how they communicate and behave once robots become part of an everyday environment.

Real-world sites will include homes, hospitals and assisted living

The research will use a Human Environment with Robots, or HERO, facility network distributed across the partner institutions. Planned settings include houses, dormitories, cafés, a public museum, a rehabilitation hospital and elder-care or assisted-living environments.

That real-world emphasis is important for physical AI. A robot that performs well in a standardized test may still fail when user preferences shift, social expectations are implicit, or a task must be adapted to an individual's changing abilities. Long-duration deployments can expose those gaps in ways that short laboratory trials cannot.

UT says community members will be able to influence how and where robots are deployed. An internal ethics board will oversee the research and develop consent and opt-out procedures. Those governance mechanisms matter because the project involves machines operating around people in sensitive everyday settings.

Six universities and major robotics partners are involved

UT Austin will lead the center with partner universities MIT, Yale University, Indiana University Bloomington, the University of Utah and Tufts University. Industry collaborators named by UT include Amazon, Apptronik, Diligent Robotics, Google DeepMind, Hello Robot, MassRobotics, NVIDIA and Robust AI.

UT's Texas Advanced Computing Center will support simulation, machine learning, data storage and digital twins of the HERO sites. Researchers can use those virtual environments to test ideas before physical deployment.

NVIDIA's role includes accelerated computing, simulation and robotics platforms for training and evaluation, while researchers will also examine NVIDIA foundation models intended for human-robot collaboration and adaptation. Apptronik will contribute physical-robotics expertise for safe humanoid deployment studies.

The technical goal is adaptation, not only task completion

The center's central research question is broader than making robots perform more actions. It aims to understand how an AI system can learn changing preferences, recognize constraints, acquire new skills, discard outdated behaviors and operate within social rules that may never be stated explicitly.

That requires advances across machine learning, robotics, cognitive science and social science. It also creates evaluation challenges: a system that becomes more personalized over time needs to be measured for reliability and safety as its behavior evolves with the user.

Why this development matters

Robotics is shifting from isolated demonstrations toward systems intended to share homes, hospitals and workplaces with people. Long-term adaptation is one of the major unsolved problems in that transition. A robot that works only when its environment and user stay fixed is not truly ready for everyday deployment.

The new NSF center creates a coordinated, multi-institution program specifically around that problem, with real-world deployment sites, large-scale simulation resources, industry hardware and AI partners, and formal ethics oversight.

The correct status is: the center has been awarded five years and $30 million in NSF funding and is being established now. Its research agenda is prospective; the announcement does not claim that long-term co-adaptive robots are already solved or ready for broad deployment.