EVOLUTIONARY COMPUTING
at the Department of Intelligent Systems,
Jozef Stefan Institute, Ljubljana, Slovenia
(EvoNet associate academic node)
This document provides the following information:
-- About the Node
-- Research in Evolutionary Computing
-- Researchers
-- Projects in the Area of Evolutionary Computing
-- Educational Activities
-- Contact Information
-- Bibliography
ABOUT THE NODE
Jozef Stefan Institute (JSI), Ljubljana, is the leading research
organisation in Slovenia. With total staff of around 700 it covers
basic and applied research in natural sciences and technology. The
Institute cooperates with numerous research institutions worldwide
and represents a training ground for many young researchers. A large
number of its scientific staff are also teaching at the Universities
of Ljubljana and Maribor.
Department of Intelligent Systems at JSI, founded in 1979 as the
Artificial Intelligence Group, currently has a research staff of
about 20. They participate in basic research projects funded
by the Slovene Ministry of Science and Technology, as well as in
international research projects, such as COST and ESPRIT. Research
at the department is conducted in the areas of machine learning,
knowledge-based systems, qualitative modelling, decision making,
logic programming, natural language processing, and evolutionary
computing.
RESEARCH IN EVOLUTIONARY COMPUTING
Research in evolutionary computing at JSI started in 1991. It covers
the following topics:
-- genetic algorithms,
-- genetic programming,
-- search and optimisation,
-- production scheduling,
-- data mining.
The emphasis is on genetic algorithms and their applications in
engineering optimisation problems. Issues relevant for practical
applications of genetic algorithms are addressed, such as:
-- constraint satisfaction,
-- qualitative optimisation criteria that are hard to handle
with traditional optimisation methods,
-- incorporation of heuristic knowledge into genetic search,
-- hybridisation with other stochastic search algorithms.
Characteristic of our work is the orientation towards solving
real-world problems. The application domains, where most work has
been done until now, are production scheduling, optimisation of
process parameters in industrial processes, and dynamic systems
control.
In production scheduling, a genetic-algorithm-based scheduling
system has been developed to tackle with a subclass of problems
where schedules are to be optimised with respect to resource
allocation. The system involves an adjusted recombination operator
for constraint satisfaction, and hybridisation of the involved
genetic algorithm. Application oriented research based on this
approach involves scheduling operations in a production unit
of a textile factory and scheduling activities in ship repair.
In an ongoing applied project, process parameters of continuous
casting of steel are being tuned by means of genetic search.
This is a multidimensional, multiobjective, and computationally
demanding optimisation task. Results obtained so far for the Acroni
Jesenice steel works indicate possible improvements of the production
process.
In dynamic systems control, genetic algorithms have been applied to
synthesise a control rule and to optimise the parameters of a control
rule with a predefined structure for the testing problem of the
inverted pendulum control. Genetic algorithms have also been used
to tune the controller of a laboratory device, where the optimality
criteria preferred by human operators were considered. We are currently
employing the developed techniques in planning and optimising the
container crane working-cycle. Most of this work is performed through
evolutionary experimentation on realistic model devices rather than
their simulators.
In addition, genetic algorithms were employed in enhancing a signal
interpretation procedure for two-phase flow measurements in fluid
dynamics. Other areas, where evolutionary computing has also been
involved, but to a smaller extent, include attribute-based machine
learning, inductive logic programming and machine discovery of
equations. The latter was done using the genetic programming approach.
RESEARCHERS
The following researchers of the JSI Department of Intelligent Systems
work on evolutionary computing:
-- Dr. Bogdan Filipic, Researcher
-- Dr. Tanja Urbancic, Researcher
-- Dr. Saso Dzeroski, Researcher
-- Viljem Krizman, M.Sc., Assistant researcher
-- Matija Drobnic, M.Sc., Assistant researcher
PROJECTS IN THE AREA OF EVOLUTIONARY COMPUTING
-- Evolutionary Computation in Optimisation and System Identification
(1995--97). Research project funded by the Slovene Ministry of
Science and Technology. Principal investigator B. Filipic.
-- ESPRIT IV European Network of Excellence in Evolutionary Computing
(EvoNet). Started in 1996. Jozef Stefan Institute is an EvoNet
associate academic node.
-- Modelling of continuous casting of steel. An applied project for
Acroni Jesenice, the largest Slovene steel works. Started in 1996
and carried out at the Faculty of Mechanical Engineering, University
of Ljubljana. We contribute to this project by optimising the
parameters of a continuous caster using a genetic algorithm.
EDUCATIONAL ACTIVITIES
-- Introductory course on evolutionary computing and genetic algorithms
(3 to 5 hours) is given by B. Filipic at the University of Ljubljana
since 1994. The course is regularly held at the Faculty of Computer
and Information Science, and occasionally at the Faculty of Mechanical
Engineering. Certain number of students work on evolutionary computing
for their research projects.
-- Courses on advanced information technologies for participants from
industrial, business, research and government institutions are being
organised by the Center for Knowledge Transfer in Information
Technologies at JSI. Some of these courses include presentations
of evolutionary computation and it application potentials given
by the JSI EvoNet team members.
CONTACT INFORMATION
Dr. Bogdan Filipic
Department of Intelligent Systems
Jozef Stefan Institute
University of Ljubljana
Jamova 39
1000 Ljubljana
Slovenia
E-mail: bogdan.filipic@ijs.si
URL : http://www-ai.ijs.muzej.si/filipic/home.html
Phone : +386 61 177 3352
Fax : +386 61 125 1038
BIBLIOGRAPHY
Journal Papers
--------------
I. Zun, B. Filipic, M. Perpar, A. Bombac.
Phase discrimination in void fraction measurements via genetic algorithms.
Review of Scientific Instruments, 66 (10): 5055--5064, October 1995.
B. Filipic.
An evolutionary approach to scheduling in a large-scale production system.
Electrotechnical Review, 62 (3-4): 217--223, 1995.
A. Varsek, T. Urbancic, B. Filipic.
Genetic algorithms in controller design and tuning. IEEE Transaction on
Systems, Man, and Cybernetics, 3 (5): 1330--1339, September-October 1993.
Book Chapters
-------------
B. Filipic, D. Juricic.
A genetic algorithm to support learning fuzzy control rules from examples.
In F. Herrera, J. L. Verdegay (Eds.), Genetic Algorithms and Soft Computing,
(Studies in Fuzziness and Soft Computing, Vol. 8), Physica-Verlag,
Heidelberg, 1996, pp. 403--418.
B. Filipic.
A genetic algorithm applied to resource management in production systems.
In J. Biethahn, V. Nissen (Eds.), Evolutionary Algorithms in Management
Applications, Springer-Verlag, Berlin, 1995, pp. 101--111.
T. Urbancic, I. Bratko. Learning to Control Dynamic Systems.
In D. Michie, D. J. Spiegelhalter, C. C. Taylor (Eds.),
Machine Learning, Neural and Statistical Classification,
Ellis Horwood, Chichester, 1994, pp. 246--261.
B. Filipic.
Enhancing genetic search to schedule a production unit.
In J. Dorn, K. A. Froeschl (Eds.), Scheduling of Production Processes,
Ellis Horwood, Chichester, 1993, pp. 61--69. Also in Proceedings of the
10th European Conference on Artificial Intelligence ECAI-92, Vienna,
Austria. John Wiley, 1992, pp. 603--607.
Papers in Conference Proceedings
--------------------------------
T. Urbancic, B. Filipic, V. Krizman.
Controller tuning with genetic algorithms: A case study in container crane
control. Proceedings of the CESA'96 Symposium on Control, Optimization and
Supervision, Lille, France, 1996, Vol. 2, pp. 803--808.
B. Filipic, I. Zun, M. Perpar.
Signal interpretation in two-phase fluid dynamics through machine learning
and evolutionary computing. Proceedings of the Ninth International Conference
on Industrial and Engineering Applications of Artificial Intelligence and
Expert Systems IEA/AIE-96, Fukuoka, Japan, 1996, pp. 461--466.
B. Filipic.
A survey of evolutionary computation methods. Proceedings of the
Fifth Electrotechnical and Computer Science Conference ERK '96,
Portoroz, Slovenia, 1996, Vol. B, pp. 31--34 (in Slovenian).
I. Zun, M. Perpar, B. Filipic.
Probe measurements of local carrying gas fraction in trickle-bed reactor.
Proceedings of the First International Symposium on Two-Phase Flow
Modelling and Experimentation, Rome, Italy, 1995, pp. 675--682.
S. Dzeroski, L. Todorovski, I. Petrovski.
Dynamical system identification with machine learning.
Proceedigs of the Workshop on Genetic Programming: From Theory to
Real-World Applications, Tahoe City, California, 1995, pp. 50--63.
B. Filipic, A. Srdoc.
Task scheduling and resource management in ship repair using a genetic
algorithm. Proceedings of the 8th International Conference on Computer
Applications in Shipbuilding, Bremen, Germany, 1994, Vol. 2,
pp. 15.17--15.28.
S. Dzeroski, I. Petrovski.
Discovering dynamics with genetic programming. Proceedings of the 7th
European Conference on Machine Learning, Springer, Berlin, 1994,
pp. 347--350.
B. Filipic, D. Juricic.
An interactive genetic algorithm for controller parameter optimization.
In R. F. Albrecht, C. R. Reeves, N. C. Steele (Eds.),
Proceedings of the International Conference on Artificial Neural Nets
and Genetic Algorithms, Innsbruck, Austria. Springer-Verlag, 1993,
pp. 458--462.
B. Filipic.
Inductive machine learning with genetic algorithms. Proceedings of
the Second Electrotechnical and Computer Science Conference ERK '93,
Portoroz, Slovenia, 1993, Vol. B, pp. 181--184 (in Slovenian).
T. Urbancic, D. Juricic, B. Filipic, I. Bratko.
Automated synthesis of control for nonlinear dynamic systems.
Preprints of the IFAC/IFIP/IMACS International Symposium on Artificial
Intelligence in Real-Time Control, Delft, The Netherlands, 1992,
pp. 605--610.
B. Filipic.
Optimizing production processes with genetic algorithms. Proceeding of
the First Electrotechnical and Computer Science Conference ERK '92,
Portoroz, Slovenia, 1992, Vol. B, pp. 167--170 (in Slovenian).
A. Varsek, T. Urbancic, B. Filipic.
Synthesizing control knowledge with genetic algorithms. Proceedings of
the ISSEK Workshop, Bled, Slovenia, 1992, 8 pages.
B. Filipic.
Genetic optimization of controller parameters. Proceedings of the 14th
International Conference on Information Technology Interfaces ITI-92,
Pula, Croatia, 1992, pp. 173--178.
A. Srdoc, B. Filipic.
Optimizing resource exploitation in ship repair using genetic algorithms.
Proceedings of the MIPRO Conference, Rijeka, Croatia, 1992, pp. 3.95--102
(in Croatian).
B. Filipic.
Scheduling with genetic algorithms. Proceedings of the 13th International
Conference on Information Technology Interfaces ITI-91, Cavtat, Croatia,
1991, pp. 161--166.
B. Filipic.
A comparison of genetic operators on the traveling salesman problem.
Proceedings of the 35th Conference on Electronics, Telecommunication,
Automation and Nuclear Engineering ETAN, Ohrid, Macedonia, 1991,
pp. 257--264 (in Slovenian).
Invited lectures
----------------
B. Filipic.
Genetic algorithms for industrial scheduling applications.
Department of Communication and Computer Engineering, Fukuoka Institute
of Technology, Fukuoka, Japan, 31 May 1996.
B. Filipic.
Learning and optimizing phase discrimination in gas-liquid flow.
Faculty of Engineering, Kobe University, Kobe, Japan, 29 May 1996.
B. Filipic.
Genetic algorithms in measurement and control. Second Symposium
on Artificial Intelligence in Measurement and Control, Zagreb,
Croatia, 5 November 1993.