Real-time iterative learning control. Design and applications. (Q625325)

The concept of Iterative Learning Control (ILC) theory is developed within the theory of repetitive processes due to its advantages over alternatives in many control-engineering problems. In the last two decades a series of books, survey papers or research studies have initiated and have completed the theoretical base of ILC. The present book aims to cover the need of an instrument containing a wide spectrum of ILC designs, case studies and illustrative examples for real-time ILC applications. As the authors state, the ultimate objective of this book is to provide readers with the fundamental concepts, schematics, configurations and generic guidelines in ILC design and implementations. The volume concentrates on the applications of ILC in a range of industrial technologies, including mechatronics, robotics, electrical machines, water-heating plants, furnace reactors, PID controller design. This monograph is structured in 11 chapters, the first one being an introduction which underlines the features and the advantages of ILC, as well as the complementary application-oriented character of the book with respect to the existing theoretical works in the ILC literature. Chapter 2 reviews the most important concepts, schematics and implementation issues related to real-time ILC. Five basic configurations of ILC for linear processes are presented, namely Previous Cycle Learning (PCL), Current Cycle Learning (CCL), Previous and Current Cycle Learning (PCCL), Cascade Learning (CL) and Incremental Cascade Learning (ICL). Two indispensable conditions are emphasized for the applicability of linear ILC algorithm to nonlinear processes: the global Lipschitz continuity condition and identical initialization condition (which is imposed by the fact that ILC tries to make a perfect tracking in a finite time interval). In Chapter 3 ILC techniques are used for the high-precision servo-control problem. Two optimal objective functions are proposed for the sampled learning control system design and three ILC algorithms (PCC, CCL and PCCL) are discussed under a unified optimal design. Chapter 4 presents a control problem associated with a class of discrete-time nonlinear systems and it shows that the ILC can compensate the nonsmooth and nonlinear factors in the system input, for instance deadzone, saturation and backslash. An application example of ILC to a piezoelectric motor is given. Chapter 5 deals with the use of ILC for enhancing tracking in batch processes through a filter-based ILC which employs the classical Bode-plot approach. The features of this ILC are illustrated by a water-heating process under a PI controller. Chapter 6 provides a frequency-domain method to design an ILC to improve the performance of the Smith predictor controller. A necessary and sufficient condition is given for the convergence of the iterative process for all admissible uncertainties and an ILC controller is designed. An application to a furnace reactor is used to illustrate the applicability of the presented method. In Chapter 7 two ILC algorithms are proposed, which are implemented (in the time-domain and in the frequency-domain respectively) to suppress the speed oscillations in a permanent-magnetic synchronous motor. Experimental results demonstrate that the frequency-domain ILC can perform robustly to measurement noise and other non-repeatable perturbations. In Chapter 8 iterative learning is used to develop a torque controller for a switched reluctance motor. For accurate current tracking, an ILC based compensation voltage is added to a P-type feedback current controller. Experimental results validate this cascaded torque controller. A new PID auto-tuning method is developed in Chapter 9. An objective function is minimized by searching the PID parameters iteratively and the iterative learning tuning provides an improvement in closed-loop control performance. Chapter 10 is devoted to an iterative learning identification and calibration method which estimates the constant parameters of micro-robots with highly nonlinear inverse kinematics. By virtue of the contractive mapping property of the iterative learning, geometric convergence speed can be achieved. The effectiveness of the proposed method is verified by an example with a multi-link closed-chain robotic manipulator. Chapters 5--9 are based upon some papers co-authored by the authors of this book. Some conclusions which are highlighted by the study of the real-time ILC design and applications are discussed in Chapter 11, as well as several future research directions in the ILC approach.