Research

Our research lies at the intersection of control theory, estimation, and machine learning. We develop methods for nonlinear dynamical systems, with particular attention to rigorous analysis and engineering applications.

Nonlinear and adaptive control

We study nonlinear, adaptive, optimal, and robust control, including feedback design under individual actuator constraints. A central interest is connecting controller design with stability guarantees and practical implementation.

Kernel methods and estimation

Reproducing kernel Hilbert space (RKHS) methods provide a framework for nonparametric estimation and adaptive control. Our interests include online estimation, persistence of excitation, convergence analysis, and kernel representations of nonlinear dynamics.

Learning for dynamical systems

Research interests include reinforcement learning for feedback control, adaptive dynamic programming, Koopman operator methods, and machine learning surrogate models for control and optimization.

Fusion plasma control

Applications include plasma density regulation, safety-factor and temperature profile control, disruption avoidance, and allocation of shared actuators. Sai Tej Paruchuri’s prior collaborative experimental work includes DIII-D, KSTAR, and TCV.