Model-Free control performance improvement using virtual reference feedback tuning and reinforcement Q-learning

摘要: This paper proposes the combination of two model-free controller tuning techniques, namely linear virtual reference feedback VRFT and nonlinear state-feedback Q-learning, referred to as a new mixed VRFT-Q learning approach. is first used find stabilising using input-output experimental data from process in model tracking setting. Reinforcement Q-learning next applied same setting input-state collected under perturbed ensure good exploration. The learned with batch fitted Q iteration algorithm uses neural networks, one for Q-function estimator controller, respectively. approach validated on position control two-degrees-of-motion open-loop stable multi input-multi output MIMO aerodynamic system AS. Extensive simulations independent channels AS show that controllers clearly improve performance over controllers.