Newsgroups: comp.robotics
Path: brunix!uunet!stanford.edu!CSD-NewsHost.Stanford.EDU!Xenon.Stanford.EDU!jek
From: jek@Xenon.Stanford.EDU (James E. Kittock)
Subject: Re: A sense of balance
Message-ID: <1993Feb1.002811.22121@CSD-NewsHost.Stanford.EDU>
Sender: news@CSD-NewsHost.Stanford.EDU
Organization: Computer Science Department, Stanford University.
References: <1993Jan31.085934.17072@adobe.com> <1993Jan31.161809.7936@sbcs.sunysb.edu>
Date: Mon, 1 Feb 1993 00:28:11 GMT
Lines: 60

In article <1993Jan31.161809.7936@sbcs.sunysb.edu> shane@cs.sunysb.edu (Shane Bouslough) writes:
[discussion of balance sensors]
>
>I have also heard of, but never seen (get this), microprocessor controlled
>carpenter's levels. I'm not sure what they do, but they're supposedly
>hot sellers, and if there is a CPU hooked up to the thing, maybe it's
>possible to use it as a sensor. Hmmm. I'm going to have to look into that.

Someone suggested mercury switches, and I can envision
a pretty simple mercury sensor that would measure tilt.
It would work somewhat like the "bubble window" in a
carpenter's level.  Basically, it would be a cube with
a conductive bottom.  The sides would have horizontal
wires which would form complete circuits with the
bottom when mercury connected them.

Thus, one face of the cube might look like (ugh, ASCII
graphics):

  |---------------|
  |---------------|
  |---------------|
  |---------------|
  |~~~~~~~~~~~~~~~|

and as the robot tilted towards this face, the mercury
(the top surface marked with ~~~~~~) would connect
circuits higher and higher up the face, giving an
indication of tilt.  Of course, if the robot is tilting
along both axes, the mercury on this face would look
something like:

  |---------------|
  |---------------|
  |----------~~~~~|
  |-----~~~~~     |
  |~~~~~          |

but by combining readings from both axes, there should
be a simple way to determine actual amount of tilt.

Our own balance mechanisms work something like this,
from what I recall. 

So anyhow, this is the basic idea, but I can see some
problems, especially: mercury would slosh about a bit
so during transitions in tilt you'd get sloppy readings
(there must be some way to correct for this, like
time-averaging or something).

Good luck,
--james



-- 
 james kittock :: stanford cs grad student :: duke '92 :: go blue devils!!
 "Adaptation is a recursive process in living systems." -- Alex. M. Andrew
 I'm tired of that quote, but I don't have a new one handy.  Such is life.

