Introductory Tutorial Workshops, Saturday June 29, 1996, 8:00 to 18:00
I-1: Identification and Robust Control Interplay
I-2: Petri Nets in Industrial Automation: Modeling, Performance, Scheduling, and Control
I-3: Computation and Control
I-4: Control Using Digital Signal Processors
I-5: Sliding Mode Control: Theory and Applications
I-6: Robust Adaptive Control Designs for Linear and Nonlinear Plants
The workshop addresses recently emergent theories which tie in the modeling of systems with their use in a subsequent robust control. Robust control design approaches require models in which system uncertainty is described by bounded perturbations. In order to use this theory, the onus is on the designer to produce sensible estimates of a bound for the unmodeled dynamics in addition to estimating a nominal mode. Standard identification techniques currently do not lead to such bounds. In addition, even if such bounds are available, standard methods of model validation also do not address such "uncertainty" models. This identification/validation/design mismatch has hindered the application of robust control and has lead to an increased interest in the connections between the identification and control design phases. The methods presented in this workshop focus on the potential synergistic coupling between these problems jointly to produce a better performance-oriented control design.
The workshop includes tutorial introductions to identification in open and closed loop and to robust control based on prescribed model uncertainty. This will then be developed to consider how a prescribed model uncertainty can be validated and/or estimated from experimental data, and further, how this validation/estimation may be turned towards an ultimate control usage. Iterative schemes will then be discussed which combine the modeling for control paradigm successively with the controller design accounting for modeling error. Such an approach links the robust control methods to experimental determination of suitable models. Methods which admit adjustment of the controller objective criterion to accommodate improved knowledge of the model will also be presented. Connections between these iterative schemes and more normal adaptive control will be considered. Experimental and simulation examples will be used to illustrate the approaches and issues.
This workshop will introduce Petri nets as a tool-methodology to advance industrial automation and present the latest developments-applications in performance evaluation, scheduling, and discrete-event control of automated manufacturing systems.
Global competition has forced industrial companies to design production systems which are computer-integrated, flexible and agile. The growing complexity of these systems presents tremendous challenges to researchers and developers for these systems' design and implementation. An advanced methodology based on Petri nets and related graphical and mathematical tools promises to address various R&D issues in advanced automation. Together with computer tools, it has been used in industry to reduce system development time drastically. The resulting systems also exhibit better performance, flexibility and agility to meet varying needs.
This workshop presents the fundamentals of Petri net modeling methods as well as recent development of Petri nets and their applications in industrial automation. The emphasis is on performance evaluation, scheduling, supervisory control synthesis, discrete event controller design, benchmark studies of industrial systems, and CAD tools.
Organizers and Main Lecturers:
This workshop is devoted to computation; particularly computational complexity and its impact on control. The workshop focuses on basic issues in control-oriented computation, and is primarily directed to industry or academic engineers who have little exposure to computational issues. The treatment will be tutorial, but it exploits recent insights which shed new light on standard results from a computational point of view. Emphasis is given to robust control and system identification, but other areas are also considered.
Computational complexity gives new insight to many historical "big" results in control, from basic theorems in linear systems, to LQG, H-infinity, least squares for identification, to the many results from Doyle, Khammesh, Megretski, Shamma, Packard, etc. about when sufficient robustness conditions in terms of LMIs are also necessary for specific uncertainty structures. The historical evolution of computation will be traced as an issue in control, and the key ideas will be illustrated with examples such as the path to the 2 Riccati equation solution to H-infinity, and the research program attempting to generalize Kharitonov's results for parametric uncertainty.
With respect to historical "big" results, in each case problems which are apparently in one computational complexity class are shown to in fact be in an easier class. In particular, problems which as stated may seem to be unsolvable in polynomial time, can be solved in polynomial time by reducing to linear algebra. What is now clear is what characterizes these problems: either the system has special structure which can be exploited, or the uncertainty is relatively unstructured.
A unified framework will be presented for modeling, identification, analysis, and synthesis which makes computational issues clear; in addition, an informal introduction will be provided to the basic ideas of computational complexity. Polynomial time problems and the basics of computation will then be presented: Numerical linear algebra, Least squares, Linear programming, Riccati equations, and LMIs. Both theoretical and practical aspects of algorithms will be discussed.
Real-time and data aspects of identification and control requires one to go beyond conventional computational complexity theory. We must have not only off-line computation (e.g. real parameter robustness, controller synthesis) which can often be fit naturally within conventional complexity theory, but also online computation (e.g.controller implementation) and data requirements (e.g. how much data is needed to identify a plant). Practically motivated examples will be used to show how each of these issues arise.
This workshop addresses practical implementation of control systems using Digital Signal Processors (DSPs) and evaluates the role of DSPs in control system design. The DSP's architecture is analysed, and it is shown why DSPs are so useful in control system design and implementation. Practical issues are emphasized, such as how to code fixed-point and floating-point DSPs, numerical considerations, and development tools. Two design examples are presented for illustration: an AC motor drive and a magnetic bearing. Working demonstrations will be presented during the workshop. Two design methods for the AC Induction motor control are discussed: the widely used vector control or field oriented method, and a nonlinear state feedback decoupling method. Implementation and performance of these two methods will be analysed and experimental results presented for adaptive torque and speed control of AC Induction motors. The task of a DSP in generating Pulse Width Modulated (PWM) waveforms to control motor drives is also discussed, as well as implementation of PWM techniques, such as Sinusoidal and Space-Vector methods.
This workshop demonstrates to both academicians and practicing engineers why enforcing sliding modes in systems with discontinuous control leads to beneficial properties such as decoupling the design procedure and low sensitivity with respect to system uncertainties (plant parameter variations, disturbances). Design methods are discussed in the context of applications to show the advantages of the sliding mode control methodology. The versatility of the application examples by their physical nature and the goal of the control tasks has been selected to be helpful for potential attendees to utilize practical sliding mode control. To overcome implementation difficulties, special attention will be paid to suppression of chattering caused by unmodeled dynamics and the use of digital controllers. Finally prospective and recent research topics will be surveyed.
Sliding mode control enables separation of overall system motion into independent partial components of lower dimensions and low sensitivity to plant parameter variations and disturbances. These properties make sliding mode an efficient tool to control high order dynamic plants operating under uncertainty conditions which is common for control in a wide range of modern technology processes. The potential of control resources may be used to the fullest extent within the framework of nonlinear control methods since the actuator limitations and other performance specifications may be included in the design procedure.
The scope of sliding mode control studies embraces heterogeneous problems (mathematical methods, design principles, applications). Major attention will be paid to sliding mode control design for finite dimensional systems, governed by ordinary differential equations. Recent developments for infinite-dimensional and discrete-time systems will also be provided.
The application part of the workshop will be dedicated mainly to control of electrical and mechanical systems (electric drives, power systems, vehicles, robots and manipulators).
Organizers:
The purpose of this workshop is to discuss the ideas and principles behind the design, simulation and implementation of robust adaptive systems for identification and control.
The workshop will be divided into three parts: The first part will deal with the design of robust adaptive schemes for linear plants with constant and time varying parameters. The basic techniques for designing a wide class of robust adaptive controllers will be presented and their performance properties and deficiencies will be clarified. Participants with no prior knowledge of Adaptive Control will be able to design and simulate adaptive controllers at the end of the lectures.
In the second part, recently developed nonlinear design tools for adaptive control will be combined into recursive design procedures for nonlinear systems with unknown parameters and unmeasured states. The resulting adaptive controllers will be shown to possess stability, convergence and transient performance properties which are superior to those of some traditional adaptive schemes.
Finally, the last part of the workshop will demonstrate the design and application of the adaptive systems presented earlier using recently developed Adaptive Control Toolboxes. Applications include electronic braking, adaptive cruise control, active suspension, heavy duty vehicles, jet engines, etc.
Introductory Tutorial Workshop # I-1 (Saturday, June 29)
Identification and Robust Control Interplay
Robert Bitmead, Australian National University, Australia
Michel Gevers, University of Louvain, Belgium
Robert Kosut, Integrated Systems, Inc. & Stanford University, USA
Roy Smith, University of California, Santa Barbara, USA
Introductory Tutorial Workshop # I-2 (Saturday, June 29)
Petri Nets in Industrial Automation: Modeling, Performance, Scheduling, and Control
Frank DiCesare, Rensselaer Polytechnic Institute
Frank Lewis, University of Texas at Arlington
L. E. Holloway, University of Kentucky
Mengchu Zhou, New Jersey Institute of Technology (Organizer)
Introductory Tutorial Workshop # I-3 (Saturday, June 29)
Computation and Control
John Doyle, Caltech
Munzer Dahleh, MIT
Introductory Tutorial Workshop # I-4 (Saturday, June 29)
Control Using Digital Signal Processors
Herbert Hanselmann, dSPACE GmbH, Germany
Issa Panahi, Texas Instruments Incorporated
Introductory Tutorial Workshop # I-5 (Saturday, June 29)
Sliding Mode Control: Theory and Applications
U. Ozguner, Ohio State University
V. Utkin, Ohio State University
K-K.D. Young, Lawrence Livermore National Laboratory
Introductory Tutorial Workshop # I-6 (Saturday, June 29)
Robust Adaptive Control Designs for Linear and Nonlinear Plants
Petros Ioannou, University of Southern California
Petar Kokotovic, University of California at Santa Barbara
Return to the IFAC '96 Tutorial Workshops Page