CAREER: Towards a data-driven systems and control theory

This NSF CAREER project aims to create new ways to control complex systems using data. Modern technologies such as autonomous vehicles, intelligent robots, advanced energy systems, and smart infrastructure must make decisions in real time while operating in uncertain and rapidly changing environments. Traditional control methods rely on detailed mathematical models of how a system behaves, but building such models is often difficult, expensive, or even impossible for today’s highly complex systems. This project will bring transformative change by developing a new scientific foundation that allows engineers to design reliable control strategies directly from measured data, without requiring precise models. This will be achieved by creating mathematical tools that use time-series measurements to understand system behavior and guide decision-making. The intellectual merit of the project includes advancing the theoretical foundations of data-driven control, addressing fundamental open questions in modelling, analysis, and control design for complex dynamical systems, and developing methods that learn and adapt in real time while providing guarantees on performance and safety. The broader impacts of the project include improving the reliability and safety of next-generation autonomous technologies and integrating research results into education and outreach activities that prepare students and the broader community for a data-driven technological future. Despite growing interest in data-driven control, ensuring robustness under uncertainty and enabling real-time adaptation remain fundamental open challenges. Many existing approaches rely on fixed offline data or uncertainty models that are incompatible with control design, preventing their safe deployment in real-world environments. This project overcomes these limitations by adopting a behavioral systems framework, where the system is defined directly by its trajectories rather than intermediate parametric models. This unifying behavior-based perspective drives three coordinated research efforts. The first develops principled uncertainty modelling and robust control methods directly from data, ensuring rigorous safety and performance guarantees. The second incorporates streaming data into the behavioral framework to create adaptive algorithms capable of continuous learning and adaptation in complex, changing environments. The third validates this theory through high-fidelity simulations and robotics experiments. Together, these research efforts establish a unified, reliable foundation for the data-driven control of complex systems. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria. NSF Award ID: 2542634 | Program: 01002627DB NSF RESEARCH & RELATED ACTIVIT | Principal Investigator: Jeremy Coulson | Institution: University of Wisconsin-Madison, MADISON, WI | Award Amount: $570,000 View on NSF Award Search: https://www.nsf.gov/awardsearch/show-award/?AWD_ID=2542634 View on Research.gov: https://www.research.gov/awardapi-service/v1/awards/2542634.html