Summary: Robust Reinforcement Learning
with Static and Dynamic Stability
with Peter Young, Douglas Hittle,
Matt Kretchmar, Michael Anderson, Jilin Tu,
Chris Delnero, David Hodgson
Colorado State University, Fort Collins, Colorado
Supported by NSF Grants CMS-9804757 and 9732986,
Siemens Building Technologies, Colorado State University
Reinforcement learning agent in parallel to engineered controller.
Potential for combining reinforcement learning and robust control theory.
Very brief reviews of
small gain theorem
integral quadratic constraints (IQCs)
Integral Quadratic Contraints for neural network
static nonlinearity (tanh)