Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Fuzzy Logic Operating Systems

2 views
Skip to first unread message

Ekrem SABAN

unread,
Feb 13, 2006, 9:53:52 AM2/13/06
to
.

=================
To incorporate the fuzzy idea into the operating system is not a devious
idea to me, as complex systems do not have necessarily crisp arbitration.
With *complex systems*, I do not only mean expert systems or other types of
AI, but relatively "simple" systems like OS. Most often, (OS) designs do not
incorporate uncertainties at all, and the computer languages used to
implement OSs are using crisp logic. For instance, a C expression will
either be matched (true) or not (false). If this is the way OS have been
designed, the question is if this the (best) right way to do so.

The idea is to use fuzzy logic in OS design is not new. On the one hand,
some FLOS products were already on the market (FuzzySoft AG), but the
products were not accepted by customers (see
http://www.dbai.tuwien.ac.at/marchives/fuzzy-mail95/0734.html ). On the
other hand, some of these OS's could even make it and were used in fuzzy
control systems (Bosch WET2820GB (BCH-WET2820GB) Built-In Washer Dryer,
WFE2021GB, etc.). And there exists research to use FLOS even in hard
real-time systems [1]!

The reason fuzzy logic is used are:
1. Complex systems cannot be pinned down to a set of formulas easily.
2. It is very difficult to write down the knowledge of an expert in crisp,
Boolean logic. The propositional logic cannot handle more complicated cases
like believe revision, which may be necessary if several experts are asked.
3. Last but not least, humans think fuzzy. We express ourselves using fuzzy
terms, called words. :) The effort to simplify the reality by mapping it to
Boolean logic results sometimes in problems that are later found and their
correction is expensive. And crisp complex systems do not behave optimal
when confronted with the reality.

OS are responsible to provide the limited resources of the system to the
processes. For real-time OS's, they have not only to assign the resources
but they have to guarantee a response time for each request as well. To
avoid problems like deadlocks, limits must be defined that determine when a
response time was too late, or whether there were other occurances that are
not allowed. These criteria can be defined with crisp logic, "But most
real-world application's knowledge is not so hard. The boundaries are more
of a softer kind. An insignificant violation of a constraint may cause a
crisp constraint satisfaction system to reject a schedule that could have
been the best )beside the insignificant violation) to find. And that is the
point where fuzzy theory is involved. Soft constraints are modelled by
flexible fuzzy constraints" [1].

Another point is the constraint satisfaction. There are cases where not all
constraints have to be satisfied. "Release dates and due dates of jobs and
operations are obvious candidates for fuzzification. An operation may start
a little earlier or later and a small due date violation may be acceptable
either."

Also, at a higher level, a fuzzyfication may make sense. For instance, if I
make a search for all files containing the word "radnom" in the file name,
the file system could ask whether this should be "random". As a person who
can understand five languages, it is also something I would appreciate alot
if a word or subject searched is also searched in other languages the user
knows. Also, a verb may have different declinations that are also
interesting in a search. Another point is when searching all files having
... that are greater than 10 MB, say. Are here files are only slightly
smaller than this minimum (256 bytes less than 10 MB) really uninteresting
to the user?

We all know some helps used in sophisticated tools, like MS Office, that
help alot when typing text. One of these features is the auto correction of
words. This could be a service supplied by the OS so that all the software
may make use of it. The price of making it available for the OS designer
would increase the costs of the OS, but encapsulate this feature to it and
will make all software better usable & with less code lines. So, the
software for this OS will become more stable. Oh sorry that I forgot to say
the link to fuzzyness concerning auto correction this time! But to decide
which language is to be corrected is a fuzzy concern. If there is a citation
of another language, the system may find out - as the reader also can find
out the original and the cited text from a different language - using fuzzy
logic.


I wrote these patches down only to start a discussion on this subject. The
questions that may be discussed are:
(1) Is it reasonable to fuzzify operating systems?
(2) Where can such fuzzification procedures give improvements? What are the
problems in crisp OS'es that would not be a problem for FLOS?
(3) What FLOS have been used designed since now? What are their outstanding
properties? What was the benefit of using fuzzy logic in design,
development, testing? How did these systems react when confronted with real
problems?
(4) What about using FLOS in (hard/soft) real-time systems? What are the
benefits? Where can fuzzy definitely not be used?
(5) What about the impact on cost of OS development/implementation/testing?
(6) ...


=============================================================
[1] « Fuzzy and Possibilistic Approaches towards Scheduling Applications»,
Vienna University of Technology, diploma thesis of Thomas Länger,
1994-04-26.


--

Kind regards/Mit freundlichen Grüßen,

D.I. Ekrem SABAN
Vienna University of Technology/TU Wien
Institute for Computer-Aided Automation/Institut für Rechnergestützte
Automation
Department for Automation Systems/Abteilung für Automatisierungssysteme
(+43-1)58 801-183 23

http://reopen911.org


Ekrem SABAN

unread,
Feb 14, 2006, 5:34:50 AM2/14/06
to
Hi there!

I am currently a research associate at the Vienna University of Technology &
responsible for the creation of new projects. Currently, I have a new
subject that involves fuzzy control and fuzzy logic operating systems.

What I want to discuss here is the reasons for using a fuzzy logic operating
system that makes some decisions using the theory of fuzzy logics.

0 new messages