Model Reference Control Using CMAC Neural Networks Alpaslan Duysak 1 , Abdurrahman Unsal 2 , and Jeffrey J. Schiano 3 1 Computer Engineering, Dumlupinar University, Kutahya, TR aduysak@dumlupinar.edu.tr 2 Electrical Engineering, Dumlupinar University, Kutahya, TR unsal@dumlupinar.edu.tr 3 Electrical Engineering, The Pennsylvania State University, State College, Pennsylvania, USA jschiano@psu.edu Abstract. This paper demonstrates the use of CMAC neural networks in real world applications for the system identification and control of nonlinear systems. As a testbed application, the problem of regulating fluid height in a column is considered. A dynamic nonlinear model of the process is obtained using fundamental physical laws and by train- ing a CMAC neural network using experimental input-output data. The CMAC model is used to implement a model reference control system. Successful experimental results are obtained in the presence of distur- bances. 1 Introduction Due to uncertainty and complexity, accurate models of dynamic systems are of- ten difficult to obtain. As a result, it is a challenge to design a controller that causes the system to respond in a desired way. Biological control systems, on the other hand, are successful despite imprecise information and complex situ- ations. Therefore, there has been a great effort to model them. Recent studies in biology, neuroscience, and psychology have led to detailed theories regarding the anatomy and physiology of the cerebellum, which is the part of the brain responsible for learning and voluntary motions. Albus proposed a mathemati- cal description of how the cerebellum computes addresses where control signals are stored, organizes memory, and generates output signals. Based on this de- scription, he proposed a manipulator control system called the cerebellar model articulation controller (CMAC),[1]. CMAC gained more attention after Miller [2] used the CMAC network for real-time control of a full-scale multidegree-of-freedom industrial robot with con- siderable success. CMAC controllers are widely used in robotic applications. For example, CMAC is utilized to perform feedforward kinematics control of four- legged robot in straight-line walking and step climbing, [3]. Shiraishi [4] used a CMAC controller for fuel-injection system and experimentally evaluated the CMAC performance on a research vehicle in a configuration fully compatible J. Marques de S´ a et al. (Eds.): ICANN, Part II, LNCS 4669, pp. 670–679, 2007. c  Springer-Verlag Berlin Heidelberg 2007