Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sdd.hp.com!zaphod.mps.ohio-state.edu!uwm.edu!src.honeywell.com!msi.umn.edu!sctc.com!smith
From: smith@sctc.com (Rick Smith)
Newsgroups: comp.robotics
Subject: Re: "Easy" way to put "AI" in realtime embedded systems?
Message-ID: <1991Apr16.213122.8043@sctc.com>
Date: 16 Apr 91 21:31:22 GMT
References: <5478@mindlink.UUCP> <kube.671722611@menaik>
Organization: SCTC
Lines: 36

>Nick_Janow@mindlink.UUCP (Nick Janow) writes:

>>Forth might be a good choice for adding AI to embedded systems.  Forth is
>>widely used in embedded systems, and is also used in robotics and AI
>>research.

This thread would make more sense if the buzz-acronym "AI" had been
replaced by something more specific. Personally, I think that "AI"
in robotics consists of trying to connect the quantitative world of
sensors and control with a more qualitative world of goals, meanings,
and intentions. But other people think it means other things (expert
systems, usually).

I don't think there *is* an "easy" way to put AI in an embedded system.

You can probably take any specific technique that is/has been assocated
with AI and implement it in C or Forth or Lisp, or whatever, but then
it's a question of the "easy way to put software in an embedded system".

On the other hand, if you're doing exploratory work in sensor
interpretation, you don't necessarily know a priori what the system
needs to do to solve the problem.

Of course, classical FORTH has the embedded dictionary and interpreter
that give it some LISPish features missing in C, making it easier to
extend in novel, unexpected ways. That's a plus in such work.

And kube@cs.UAlberta.CA (Ron Kube) writes:

>Do you have an example of Forth's use in AI research?

There is/used to be a lab full of FORTH hackers at UMass doing robotics
research. I don't know if any of their work was AI work.

Rick.
smith@sctc.com    Arden Hills, Minnesota