Under Control


I just got back from Santa Fe, where I discussed with a number of like- minded individuals - in an informal conference setting - chaos and complexity theories, especially as they relate to issues in manufacturing. Nearby was the Santa Fe Institute, a multi-disciplinary research and graduate education enterprise involved in studies of the principles that determine the dynamic behavior of complex systems. Models of complex systems seem to have application in many fields, including biology, computer science, economics, physical and political science - hence the multi-disciplinary approach. Besides institute members, attendees included representatives of Rockwell International and Allen-Bradley, a Rockwell company, an others.

The kind of ideas we talked about go by many names - agent-based systems, emergent programming, chaos theory, nonlinear dynamic systems, and artificial life. Basic to many of them is the notion that total systematic behavior can arise from the union of localized systems.

For example, the classical way of thinking about something like an ant colony is to imagine that somewhere there is a master plan that governs it. But, in reality, no one "programs" the ant colony as a totality. What really happens is that the relationships of each ant to all other objects - to include its environment and the other ants - taken together define the colony. Individual ants have rules to follow, and the interactions of the rule-governed individuals end up determining colony behavior.

The sample applies to human societies. Rather than imagine there is one essential "invisible hand" that guides the market system, it may be more productive to envision any number of "hands" that interact in a somewhat characteristic fashion to form an economic system of such complexity that there is a single, right answer to question of group behavior can be very dangerous.

What's true for macroeconomics systems is also true for manufacturing systems. To apply these theories to automated production, what we do is replace scheduling techniques that treat the manufacturing system as a totality - in which all variables must be accounted for - with techniques for parallel scheduling of random and chaotic processes. In Santa Fe, we at Flavors Technology presented our emergent system approach to scheduling paint booths at a General Motors auto plant.

In this application, the booths are the defined objects and the programming code describes the booth's relationship to its environment. A "schedule" per se is not programmed, but only the objects and the rules of the interaction. The result: a flow of trucks emerging from the autonomous agent called the paint booth. Each time we run the simulation, the results are different.

Use of these type scheduling techniques should be considered whenever a system is complex, i.e., large, complicated and interactive and intractable. Most complex systems are too... well, too complex to be wholly conceptualized. This alternative approach focuses on operational variables and is much better able to deal with unforeseen eventualities. Practitioners continue to be surprised by the results they get. Systems seem to get better in startling ways. Considerable savings in software development are also possible.

As with anything new, acceptance in the engineering community has been tepid at best. Apparently though, Japan is an exception. Recently, Japan hired some of the top researchers from the Santa Fe Institute to study chaos theories in Japan.

At the conference, institute representatives gave the attendees a solid overview of some of the new techniques. We also visited the institute during afternoon tea and wound up giggling a lot. Presentations by Dove, Patenak, Roth and others updated us on work currently being done in applications in neurons, global system behavior, adaptive learning and social interaction.

In general, the attendees were enthusiastic and interested. In fact, we elected to work late on evening rather than frequent the bar. We also decided to do the whole thing again next year and to start a news letter. An E-mail net is being set up by Ken Carter of Control Technology Corp. on INTERNET [ken@cthulhu.control.com]. He has information available.

As appeared in Manufacturing Systems Magazine February 1993 Page 40

Manufacturing Systems Magazine Articles
References - Table of Contents

Send mail to rmi.info@barn.org for more information.
Please send mail to webmaster@barn.org regarding web site structure.
Copyright © 1996-2002 R.Morley Inc. All Rights Reserved

R. Morley Incorporated
614 Nashua Street, Suite 56
Milford, NH 03055-4992 USA
Tel: 603-878-4365 FAX: 603-878-4385