Research Topics

Occupation kernels and Liouville operators

Operator theoretic methods for data-driven identification, verification, and control synthesis.

Model-based reinforcement learning

Capitalizing on recent developments in reinforcement learning in continuous time and space, we develop novel model-based reinforcement learning methods that vastly improve data efficiency and usefulness for online optimal feedback control.

Details and publications

Data-driven modeling

Motivated by the need to adapt to changing models and reduce modeling errors in model-based control, we developed data-driven adaptive system identification methods to identify unknown parameters in the system model.

Details and publications

Inverse reinforcement learning

Based on the premise that a cost (or reward) function fully characterizes the intent of the demonstrator, we developed methods to learn the cost function from observations for linear and nonlinear uncertain systems.

Details and publications

Stability of nonsmooth systems

Extensions of the LaSalle-Yoshizawa Theorem to nonsmooth nonautonomous systems and switched systems utilizing Filippov solutions of nonsmooth differential equations and nonstrict Lyapunov functions.

Details and publications