Path: utzoo!utgpu!water!watmath!clyde!rutgers!cmcl2!nrl-cmf!ames!pasteur!ucbvax!VAXA.ISI.EDU!smoliar
From: smoliar@VAXA.ISI.EDU (Stephen Smoliar)
Newsgroups: comp.ai.digest
Subject: Re: table tennis playing robot
Message-ID: <8801251658.AA21776@vaxa.isi.edu>
Date: 25 Jan 88 16:58:19 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 23
Approved: ailist@kl.sri.com

Since I was at the MIT Artificial Intelligence Laboratory during the 60's,
I can flesh out some of Rob Elkins comments.  When I first arrived, there
was a major effort to build a robot which could play table tennis.  (I think
this was the summer of 1966.)  There was an "arm" which was an AMF manipulator
which operated in cylindrical coordinates and a simple vidicon "eye."  When
I arrived, a basket had been attached to the end of the arm, and development
was concerned with getting the arm to catch the ball based on the trajectory
tracked by the eye.  Because of the relatively primitive environment, there
were all sorts of problems.  For example, the initial version really didn't
work in three dimensions.  Thus, for a successful test, one had to throw
the ball in a specific plane;  and the arm could only adjust itself on
the up-down and forward-back axes.  As I recall, there was a joke to the
effect that, after six months of intensive development of both hardware and
software (the MIDAS assembled on a PDP-6), all researchers had learned how
to throw the ball in such a way that the arm didn't have to move!

I was not there when the arm tried to catch Marvin Minksy's (bald) head.
So I don't know if the story is really true.  Given the intricacies of
setting up a test, I suspect it is at least a slight exaggeration.
Attention subsequently shifted to an arm with four (I think) hydraulically
controlled flex-extend joint.  However, I do not think this arm ever caught
any flying objects.  (As a matter of fact, I'm not sure it was ever
controlled effectively for any purpose.)