Uncertain variables and learning algorithms in knowledge-based control systems

This paper concerns a class of knowledge-based control systems with unknown parameters in the knowledge representation describing a plant. For the deterministic case, the logic-algebraic method of determination of the control decisions has been developed. The purpose of this paper is to present an extension of this method for the case with unknown parameters. In the first part, an approach based on uncertain variables is presented. In the second part, the method and algorithms of current knowledge updating in the learning system are proposed. The main idea of this approach consists of a step-by-step estimation of the unknown parameters in the knowledge representation using the successive values of the control decisions and their results. This concept may be considered as an extension of the known ideas of identification and adaptation for the traditional case. This work was presented, in part, at the Third International Symposium on Artificial Life and Robotics, Oita, Japan, January 19–21, 1998