CAREER: Toward Real-Time, Constraint-Aware Control of Complex Dynamical Systems: from Theory and Algorithms to Software Tools

University Of Vermont & State Agricultural College, Burlington VT

Investigators

Abstract

This Faculty Early Career Development (CAREER) grant will fund research that advances knowledge of real-time automatic control with application to complex, safety-critical systems in power, transportation, and manufacturing infrastructure, thereby promoting the progress of science, and advancing the national prosperity. Complex systems are characterized by the coupling of many subsystems, unmodeled components, significant variability, and environmental uncertainty, which prevent effective use of traditional model-based techniques, and render current data-driven approaches computationally intractable. This project addresses these shortcomings by building a new theoretical framework for creating data-based algorithms that are easy to tune by non-experts, real-time feasible, and accompanied by guarantees of safety and performance. A comprehensive plan for pedagogy and outreach aims to integrate the autonomous technology ecosystem in Vermont, establish and strengthen relationships between industry, academia, and government, disseminate research results through sharing of open-source software and accessible video content, and engage with students from underrepresented and underserved communities, including at rural high schools in Vermont. This research aims to develop the foundations for a new data-driven control framework for complex systems, which decouples the problems of tracking and constraint management through a marriage of the Internal Model Control and Reference Governor model-based design techniques with the Behavioral Systems Theory data-driven approach. By inheriting many desirable properties of the model-based approaches, such as robust stability, constraint enforcement, and finite-time convergence to constraint-admissible setpoints, this framework overcomes the limitations of data-driven approaches within a modular structure that simplifies analysis and design, and allows for easily tunable, computationally tractable algorithms. To this end, the project will investigate the system-theoretic underpinnings of the new control framework, characterize its properties and fundamental limitations, and develop generalizations to nonlinear, time-varying, and unstable systems. Informed by collaboration with Beta Technologies and the Vermont Electric Power Company, VELCO, the control framework will be validated on an electric motor experiment, a quadrotor swarm transporting a complex payload, and a hardware-in-the-loop simulation of a power system with significant penetration of distributed energy resources. This project is jointly funded by the Dynamics, Control and Systems Diagnostics program and the Established Program to Stimulate Competitive Research (EPSCoR). 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.