NSF Foundational Research in Robotics (FRR) for AI-Integrated Autonomous and Embodied Robotic Systems

Foundational Research in Robotics is NSF's core robotics program and the primary federal home for embodied intelligence research that is not defense-directed. Its definition of a robot is worth quoting because it functions as the eligibility test: intelligence embodied in an engineered construct, with the ability to process information, sense, plan, and move within or substantially alter its working environment. Every proposal must clear three specific bars - it must focus on a robot or a class of robots, it must endow robots with new capabilities or significantly enhance existing ones, and it must address a fundamental gap in robotics research. NSF is explicit and unusually blunt about the consequence: projects that are better suited to other NSF programs may be declined without review. That is the single most important operational fact about FRR. It sits in a deliberately drawn boundary with NSF's Robust Intelligence program, which funds foundational AI but explicitly does not fund robotics and redirects robotics proposals to FRR; conversely, an AI-heavy proposal that uses a robot as a demonstration vehicle rather than advancing robotics science will be sent the other way. Getting that framing right before writing is more consequential than any other single decision. The program is jointly managed by the CISE and ENG directorates, which is why it accommodates both learning-and-perception-led and mechanism-and-control-led work, and it encourages meaningful experimental validation on a physical platform - a real differentiator from purely simulation-based programmes. FRR accepts full proposals at any time, with no submission window, under program description PD 20-144Y. Eligibility is unrestricted as to entity type. NSF program officers explicitly recommend pre-submission consultation, and given the no-deadline structure and the without-review declination risk, that conversation is the highest-value step an applicant can take.

Foundational Research in Robotics (FRR) is NSF's core robotics program, run jointly by the CISE and ENG directorates, and its framing is unusually explicit about the AI connection: the program defines a robot as intelligence embodied in an engineered construct, with the ability to process information, sense, plan, and move within or substantially alter its working environment. That definition is the gate. Proposals that are purely algorithmic, with the robot as an afterthought, tend to be redirected to other CISE programs, while proposals that treat perception, learning, planning, and physical embodiment as inextricably interwoven are the intended fit. NSF asks every FRR proposal to satisfy three elements simultaneously: focus on a specific robot or class of robots, endow that robot with genuinely new capabilities, and address a fundamental gap in robotics science rather than an engineering integration problem. Meaningful experimental validation on a physical platform is encouraged, which in practice means simulation-only work needs a strong justification. The practical advantage of FRR over most NSF programs is the submission model - there is no annual deadline. Full proposals are accepted at any time, which removes the usual scramble but also removes the forcing function, so teams should budget for a review cycle rather than a deadline. Declined proposals carry a minimum one-year moratorium before resubmission, so a weak first attempt is expensive. EAGER, RAPID, and RAISE proposals may also be routed through FRR at any time, but PIs must contact the cognizant program officer before submitting those. For embodied-AI groups - large behavior models, robot foundation models, learned manipulation, legged locomotion - FRR is the most direct federal non-defense funding line in the United States.

FRR is the core NSF vehicle for robotics research that is genuinely about robots rather than about algorithms that happen to be demonstrated on one, and the program's definition of its subject is the sharpest statement of embodied intelligence in U.S. federal funding language: a robot is intelligence embodied in an engineered construct, with the ability to process information, sense, plan, and move within or substantially alter its working environment. Jointly led by the CISE and ENG directorates under program description PD 20-144Y, it supports research on robotic systems exhibiting significant levels of both computational capability and physical complexity - the conjunction is deliberate, and proposals weak on either side tend not to survive review. NSF requires every proposal to address three specific points: it must focus on a robot or a class of robots, it must endow robots with new capabilities or significantly enhance existing ones, and it must address fundamental gaps in robotics research. The program explicitly prioritises foundational advances spanning engineering and computer science, and states that meaningful experimental validation on a physical platform is encouraged, which is a real filter on purely simulation-based work. The operational feature that makes FRR unusually valuable is the absence of a deadline: full proposals have been accepted anytime since August 1, 2020, so a team can submit when the preliminary results justify it rather than racing an annual date. The one timing constraint that bites is on the way out - declined proposals face a minimum one-year moratorium before resubmission, which makes a premature submission genuinely costly. NSF strongly encourages potential investigators to discuss projects with an FRR Program Officer before submitting, and given the no-deadline structure there is no reason not to.

The NSF Computer and Information Science and Engineering: Future Computing Research (Future CoRe) program (NSF 25-543) supports foundational computing research and education with a strong emphasis on AI and machine learning. The program funds research that advances the foundations of computing including AI/ML theory, algorithms, systems, and applications. Future CoRe supports small to medium-sized research projects that can have significant impact on computing foundations. The program has target dates of February 5, 2026 and September 10, 2026 for proposal submissions (note: these are target dates, not deadlines, meaning proposals may be submitted at any time but will be reviewed in batches around these dates). Investigators may not serve as PI, co-PI, or Senior/Key Personnel on more than two proposals submitted within any consecutive 12-month period across all Future CoRe programs. This is one of NSF's primary mechanisms for funding fundamental AI and machine learning research at U.S. academic institutions, covering areas such as machine learning theory, natural language processing, computer vision, robotics foundations, and human-AI interaction.

The ONR Long Range Broad Agency Announcement (N00014-25-S-B001) is the Office of Naval Research's primary mechanism for soliciting research proposals across all naval science and technology priority areas. The BAA accepts proposals on a rolling basis through September 30, 2026 and covers ONR's full spectrum of research interests with particular emphasis on AI-related topics including autonomous maritime systems, human-machine teaming, machine learning for sensor fusion, cooperative autonomous swarm technology, undersea autonomy, and AI-enabled decision superiority. Proposals can be funded through multiple mechanisms including individual investigator grants, the Young Investigator Program (~$510K over 3 years for early-career faculty), and Multidisciplinary University Research Initiative (MURI) awards ($1.5M/year for 3-5 years for research teams). ONR recommends contacting relevant program officers before submitting to discuss alignment with current research priorities. The BAA supports basic research (6.1), applied research (6.2), and advanced technology development (6.3) across the full range of naval-relevant science and engineering disciplines.

DE-FOA-0003600 is the DOE Office of Science's FY 2026 open, rolling solicitation for financial assistance, providing roughly $500 million across seven program areas: Advanced Scientific Computing Research, Basic Energy Sciences, Biological and Environmental Research, Fusion Energy Sciences, High Energy Physics, Nuclear Physics, and Isotope R&D and Production. AI and machine learning research is supported directly through Advanced Scientific Computing Research, while Biological and Environmental Research funds atmospheric process research, environmental systems process research, and earth-energy systems modeling — making this a major channel for AI-enhanced climate and earth system modeling work. DOE anticipates 200 to 350 new awards ranging from $5,000 to $5 million each, with project periods from six months to five years. The FOA opened September 30, 2025 and accepts applications on a rolling basis through the end of FY 2026.