Okyay Kaynak
Electrical and Electronic Engineering Department, Bogazici University, Bebek, 34342, Istanbul, Turkey
kaynak@boun.edu.tr

Abstract
Variable Structure Systems (VSS) theory based Sliding Mode Control (SMC) is well known with its characteristics of robustness against disturbances and uncertainties that inevitably exist in any control system. A recent tendency in the field is to exploit the strength of the technique for on-line learning in computationally intelligent systems.
In the first part of the tutorial, several variable structure systems theory based online training strategies for neural networks will be discussed. In contrast to the conventional backpropagation procedure, the presented algorithms for weight adaptation control the error dynamics of the neural structure. It is described by a differential or difference equation for the error, which is the difference between the desired and the actual network outputs. The learning parameters are adjusted to force the error to satisfy this stable error equation, rather than to minimize an error function like the backpropagation algorithm does. In other words, the proposed adaptation rules can be viewed as a controlled dynamical system where the network weights can be considered as the states of the system, the correction terms to the present values of the weights can be viewed as the control signal and the error signal is interpreted as the output signal.
In the second part of the tutorial several approaches from the literature for the design of stable neuro-adaptive systems will be discussed. In two of the approaches, namely the sliding mode feedback-error-learning approach and the sliding mode control with a task-specific output error approach, the analyses presented aim to extract the conditions for establishing an equivalence between the sliding mode control of the plant and the sliding mode learning in the neuro-controller. Dynamic parameter adaptation laws are derived and the applicability of the algorithms is discussed. In another method that will be presented, the controller tuning is based on an estimate of the command-error on its output which is determined by calculating the Jacobians and the one-step-ahead predicted tracking error of the system obtained using a neural predictive model. A robust on-line learning algorithm, based on the direct use of SMC theory is applied to both the controller and the model. The general characteristic of the approaches described is to allow handling of the plant-model mismatches, uncertainties and parameters changes.
The performances of the methods developed are assessed on several nonlinear systems the dynamic equations of which are assumed to be unknown. Several simulation studies are presented that demonstrate the alleviation of the adverse effects of observation noise and abrupt changes in system parameters.

Speaker
Okyay Kaynak received the B.Sc. degree with first class honors and Ph.D. degrees in electronic and electrical engineering from the University of Birmingham, UK, in 1969 and 1972 respectively.
From 1972 to 1979, he held various positions within the industry. In 1979, he joined the Department of Electrical and Electronics Engineering, Bogazici University, Istanbul, Turkey, where he is presently a Full Professor. He served as the Chairman of the Computer Engineering Department for three years, of the Electrical and Electronic Engineering Department for two years and was the Director of Biomedical Engineering Institute for one year. Currently, he is the UNESCO Chair on Mechatronics and the Director of Mechatronics Research and Application Centre. He has hold long-term Visiting Professor/Scholar positions at various institutions in Japan, Germany, U.S. and Singapore. His current research interests are in the fields of intelligent control and mechatronics. He has authored three books and edited five. He has also authored or coauthored almost 200 papers that have appeared in various journals and conference proceedings. He was elevated to fellow grade in IEEE in 2003.
Dr. Kaynak served as a President of the IEEE Industrial Electronics Society during the years 2002-2003 and he is currently the Editor-in-Chief of IEEE Transactions on Industrial Informatics.