Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!samsung!zaphod.mps.ohio-state.edu!swrinde!ucsd!rutgers!netnews.upenn.edu!grip.cis.upenn.edu!ulrich
From: ulrich@grip.cis.upenn.edu (Nathan Ulrich)
Newsgroups: comp.robotics
Subject: Re: Robots in our Future?
Message-ID: <29011@netnews.upenn.edu>
Date: 30 Aug 90 16:40:39 GMT
References: <J?=%^5|@ads.com> <1990Aug28.234809.15660@portia.Stanford.EDU> <1990Aug30.014817.8794@portia.Stanford.EDU>
Sender: news@netnews.upenn.edu
Reply-To: ulrich@grip.cis.upenn.edu (Nathan Ulrich)
Organization: University of Pennsylvania
Lines: 56

In article <1990Aug30.014817.8794@portia.Stanford.EDU> boehlke@sunrise.stanford.edu (Dan Boehlke) writes:
>I see the next breakthrough in robotics being the 
>introduction of very high accuracy manipulators--
>say an order of magnitude (or more) better than
>any systems of the 80's.  There are plenty of potential
>products that simply cannot be assembled today anywhere 
>outside of a laboratory.
>
>Ultra-precision manipulators will be "enabling" technology.  
>They will make it possible to produce consumer products that 
>are impossible to manufacture on a mass-production basis today. 
>Once the technology exists, new products will be developed 
>which exploit it.  Electronic minaturization will go far beyond
>where it is today, quality will get better, and cost will go down.

I strongly disagree.  If you want "ultra-precision" position control, then
use hard automation or any of a variety of NC machines which are capable of
accuracy better than a millionth of an inch (much better, in some cases).
By my definition, these are not robots.  Of course, from my point of view,
99% of "robots" used today are simply reconfigurable assembly machines, not
real robots---they are never called upon to react to their environment in
any significant way, which I would call the first prerequisite for a true
robot.

But then again, this is just a problem of definitions.  I think that to
really make robots versatile and useful outside of factory or hazardous
applications, you need to enhance several things.  First, their force-control
capability has to be drastically expanded.  The arm that Bill Townsend
developed at MIT has demonstrated that this is within the capabilities of
current technology--its transmissions are better than 96% efficient with
zero backlash.  Second, the ratio of system weight to system payload has
to be increased.  Again, this just requires intelligent design that is not
tied to the requirements of industrial automation.  Third, the sensing
capability, especially tactile and vision, has to be improved.  I think many
of the limitations of robots stem from their mechanical design (that's what
I do, so I'm biased), but there are clearly problems that can only be
resolved from the AI community.

If we want to bring robots out of their current narrow area of application,
then they must have some of the abilities that humans possess.  I don't
believe that we should build anthropomorphic robots, but we can learn
from the way biological systems solve problems.  Planes don't fly by flapping
their wings, but their wings and tails use principles found in birds.

Humans have terrible position control capability (ever tried to put your finger
in a precise point in space--you can't), but have remarkable force-control and
sensory feedback abilities.  I think we can learn from this.  Of course, our
incredible brains could just be compensating for lousy mechanics, but watch
an ant or a spider sometime; they certainly don't have much in terms of
high-level reasoning.

Just my opinion.

Nathan Ulrich                           "If it was easy, someone would have
ulrich@grip.cis.upenn.edu                done it already..."
DoD #0080 - GT750 pilot