A Comparison of DHP Based Antecedent Parameter Tuning Strategies for Fuzzy Control

In the context of fuzzy control, antecedent parameters are used to provide a segmentation of the state space so that different regions can be modeled appropriately. In Adaptive Critic methodologies, two modules (the critic and the controller) must properly segment the state space to insure good performance. In this paper, we explore the effects of tuning antecedent parameters that are shared between these controller and critic networks (as opposed to tuning separate parameter sets). The results indicate that training shared antecedent parameters can be as effective as training separate antecedent parameters.