Newsgroups: comp.robotics
Path: brunix!uunet!wupost!micro-heart-of-gold.mit.edu!uw-beaver!rice!rusty
From: rusty@masc.rice.edu ( Robert Kelley)
Subject: Re: The Real World
Message-ID: <1992Mar8.063826.24616@rice.edu>
Sender: news@rice.edu (News)
Organization: Dept. of Math. Sciences, Rice University
References: <1992Mar6.113154.349@cs.nott.ac.uk> <33425@ttidca.TTI.COM>
Date: Sun, 8 Mar 1992 06:38:26 GMT

 
] The Polymath (aka: Jerry Hollombe, M.A., CDP, aka: hollombe@soldev.tti.com)
] 	writes:
] 
] In article <1992Mar6.113154.349@cs.nott.ac.uk> pre@cs.nott.ac.uk
	(Phillip Edwards) writes:

] }I subscribe, provided it is interpreted with flexibility, to the RIA
] }/ BRA definition:
] }
] }"A robot is a reprogrammable, multifunctional manipulator designed
] }to move materials, parts tools or specialised devices, through variable
] }programmed motions for the performance of a variety of tasks."
] 
] To that I'll add that, sometimes, the robot only has to move itself.
] I suppose that could come under "tools."

			*********************

I see what you mean about "interpreted with flexibility."  With all due
respect to its perpetrators, this does seem a little narrow.

I know a number of reasonably successful attempts have been made to offer
useful definitions of robotics.  I further know that no one effort can
be said to have been completely successful, any more than have been
definitions of "learning" or "intelligence".

While I have absolutely no wish to offer a suggestion as to what might
comprise an, ah, ultimate definition, *AND* no wish to reopen a flame war,
I will offer the following, by way of contrast, and because I think it
contains some elements, and excludes others, which may tend to make it
more useful.  Its chief virtue and weakness is, of course, its simplicity:

A robot is a mechanical device capable of performing mechanical work with
incomplete or inexact work specifications.

Note what this definition does *not* include:

--Anything to do with "programming."  This leaves the door open to analog
computing approaches and devices such as neural nets based on learning by
example.  Some *intelligence*, however, is implied by the incomplete/inexact
work specifications.

--Mechanisms by which the work is performed; wrenches, detachable tools,
pneumatic tools, water wheels.  I believe most definitions tend to be too
narrow, while this opens the door to constructs which might perfectly well
fit the definition of "robots" but are not commonly perceived in that
category.  For instance, while self-regulation machinery is not commonly
thought of as robotic, self-regulating houses, (climate, lights, locks, etc.),
perhaps are.

As to the obvious keyword, "mechanical" (you may substitute "partly mech-
anical" if you wish), I submit it as my opinion that this is, in fact, the
central property of the class of devices we are talking about.  In short,
machines-- which use machines-- without excessive handholding by humans.

	RS Kelley, Ph.D.	rusty@rice.edu		Rice University
