INTRODUCTION TO INTERVAL ARITHMETIC AND ITS APPLICATIONS
Responsible Organizer
Dr. Josep Vehí, University of Girona
Speakers
Prof. J. Garloff, University
of Applied Sciences Konstanz, Germany.
Prof. V. Kreinovich, University
of Texas at El Paso, USA.
Prof. B. Barmish, University
of Wisconsin at Madison, USA.
22-23 February 1999, University of Girona, Girona, Spain
PRESENTATION
In coincidence with the workshop MISC'99
to be held on 24-26 February 1999 in Girona, a pre-workshop tutorial course
"Introduction to interval arithmetic and its applications" will be organized
by the University of Girona. The aim of the course is to bring together
researchers and doctorate students that work in systems and control engineering
so as to give them the opportunity to have a knowledge of the method of
intervals and to stimulate discussion of relevant problems and fertile
lines of future research.
Three prominent speakers, Professor
Jürgen Garloff, Professor Vladik Kreinovich and Professor Bob Barmish,
will provide an interesting and comprehensive state-of-the-art tutorial
on motivations, basic theories and applications of interval analysis to
systems and control. The past two decades have witnessed steadily increasing
recognition and appreciation of Interval Analysis in solving problems in
systems and control engineering. For instance, problems like analysis and
synthesis of robust controllers for uncertain plants or fuzzy interference
have been stated from the interval point of view. Recent developments have
generated a great deal of interests in emerging tools aimed at exploiting
qualitative, semiqualitative and interval simulation and their application
to fault detection and diagnosis and to system identification.
SCHEDULE
Monday afternoon 22 February 1999
15:30
- 19:00
Tuesday morning 23 February 1999 09:00
- 11:00 11:30 - 13:30
Tuesday afternoon 23 February 1999
15:30
- 19:00
PARTICIPATION FEE
The participation fee will be 15000
Spanish pesetas. Advanced registration is highly recommended (Registration
form: PostScript,
PDF)
CONTACT ADDRESS
Dr. Josep Vehí
Department of Electronics, Automatic
Control and Computer Engineering
University of Girona, Campus Montilivi,
Edifici P-II
E-17071 Girona, SPAIN
misc99@eia.udg.es
TOPICS OF THE COURSE:
INTRODUCTION TO INTERVAL COMPUTATIONS
1. Why intervals and why interval computations?
- Inexact knowledge of physical and technical constants;
rounding errors; approximation formulas of function,
derivative and integral. Sensitivity analysis
and tolerance problem
2. Basic properties of interval arithmetic
3. Rounded interval arithmetic
4. Enclosures for the range of a function
- Naive interval computations; mean value form and other
centered forms
5. Solution of systems of linear interval equations
- Linear interval equations; M-matrices and inverse positive
matrices; preconditioning; interval Gauss elimination;
interval Gauss-Seidel iteration .
6. Solution of systems of nonlinear interval equations
- Branch-and-bound methods (methods which do not use
derivatives). Newton-type methods. Automatic
differentiation; forward and backward. Optimal
bisection. Constraint propagation approach
7. Optimization
- Branch-and-bound methods (methods which do not use
derivatives). Newton-type methods. Optimization of
non-differentiable functions (method of generalized
derivatives)
APPLICATIONS OF INTERVALS
1. Error estimation for data processing - the basic application of
interval computations:
- why data processing (indirect measurements); why it
is important to estimate errors of data processing
- two opposite situations when we do not know probabilities,
only intervals: cutting-edge fundamental science
super-precise measurements); manufacturing applications
(crude measurements)
- linearization and generalized (affine) interval computations
- cases when interval computations are necessary
- in general, interval computations are intractable;
ways out: overestimation, Monte-Carlo methods
2. Additional problems:
- transforming an input-output relation into an explicit
algorithm; experimental determination of the input-output
relation (system identification); changing the
world: synthesis, design, control, and decision making
3. Transforming an input-output relation into an explicit algorithm:
real-life applications of solving systems of equations:
- systems of algebraic equations. Case study: computer
graphics
- systems of differential equations. Case study: particle
accelerators
- systems of integral equations. Case study: astrophysics
4. Interval-valued system identification:
- formulation of the problem. Typical situation: systems
of linear interval equations: Case study: non-destructive
evaluation of aerospace constructions
5. Synthesis, design, control, and decision making: real-life applications
of interval optimization
- problems with finitely many alternatives: how to compare
intervals
- problems in which an alternative can be characterized
by finitely many real numbers: interval optimization
- case study: chemical spectral analysis; image processing
in radioastronomy; detection of elementary particles
- problems in which alternatives can be characterized
by functions, e.g., in control: economical planning with
Leontieff's input-output model award-winning mobile
robot; telemanipulators (MIT/Utah robotic arm)
- general introduction to robust control: Kharitonov's
theorem; its applications
6. Beyond interval computations:
- intervals + probabilities
- intervals + expert knowledge
- first step: nested intervals and fuzzy logic
- second step: interval-valued degrees of belief
- case study: using expert knowledge to control a mobile
robot
7. Probabilistic robustness with intervals