A Control Algorithm Base on Adaptive Critic Designs

This paper focuses on studying a class of on-line learning control based on adaptive critic designs. It can be classified as action-dependent dual heuristic programming (ADDHP). The principle of adaptive critic designs is introduced. We design an on-line learning controller, and provide details about the learning of critic networks and action network used in the present algorithm. In our simulation study, we employ the cart-pole balancing problem to show the applicability of the on-line learning control algorithm.