Toward effective combination of off-line and on-line training in ADP framework

Abstract

We are interested in finding the most effective combination between off-line and on-line/real-time training in approximate dynamic programming. We introduce our approach of combining proven off-line methods of training for robustness with a group of on-line methods. Training for robustness is carried out on reasonably accurate models with the multi-stream Kalman filter method (Feldkamp et al., 1998), whereas on-line adaptation is performed either with the help of a critic or by methods resembling reinforcement learning. We also illustrate importance of using recurrent neural networks for both controller/actor and critic