Project Grant 2541960

Duke University's Office of Research Administration received a $420,000 project grant awarded October 1, 2026, under the National Science Foundation's Computer and Information Science and Engineering program (CFDA 47.070), administered by the Division of Computing and Communication Foundations. The award supports a five-year CAREER project (completion date September 30, 2031) focused on developing a comprehensive theoretical and algorithmic framework for off-dynamics reinforcement learning—methods that enable artificial intelligence agents trained in simulated or source environments to reliably transfer knowledge and deploy effectively in real-world target domains with different transition dynamics.

The primary deliverables include foundational research addressing distributional shifts between training and deployment environments, algorithmic innovations organized around distributionally robust learning methods, and open-source software and data resources generated through the research activities. Additional project products encompass educational materials and workforce development initiatives, including student training at multiple levels and outreach activities designed to broaden participation in artificial intelligence. All research outputs, software tools, datasets, and educational resources will be made publicly available to maximize impact and benefit applications in healthcare, robotics, autonomous systems, and other critical domains where safe, data-efficient decision-making is essential.