From: dstamp@watserv1.waterloo.edu (Dave Stampe-Psy+Eng)
Subject: Re: TECH: Options for position sensors???
Date: Sun, 5 Apr 1992 23:41:10 GMT
Message-ID: <1992Apr5.234110.22314@watserv1.waterloo.edu>
Organization: University of Waterloo



Last time this discussion was held, there seemed to be the following
classes of position/angle detectors:

Magnetic: Polhemus is a good example.  Has problems with metal nearby
at distances greater than 2 ft. or so.  Polhemus trackers have long
(200 mS) in-system delays, but reasonable update rates.

Passive magnetic:  Headmounted fluxgate compasses, for example.  Those
who have tried them say: LOTS of problems with metal, and very bad if
the head is not held level.  Not very accurate (repeatablilty, not
resolution).

Mechanical:  Hanging arms from the ceiling, displays on long,
counterbalanced arms, strings and springloaded reels.  They'd
be cheap, but are prone to vibration and flexing in arms, and
add inertia or weight (or just plain limit range).

Ultrasonic:  Whether Powergloves, Logitech wands or those funny
headgear that used to be sold for the Mac (moved the mouse with
your head, eh?  Sounds like VPL is infringing a patent there
themselves), these have the same problems and characteristics.
Speed is not too great, as you really need more than one transmitter
to get more than 3D (position or angle, but not both) and you
have to fire the transmitters in sequence.  It takes about 10-30 mS
for the room's echoes to die down after each pulse, so the rates are 
10-30 samples/sec.  Cheap transducers are prone to glitches, expensive
ones to noise.

Simple Infrared:  Put special sensors (with geometric masks) on
the helmet, blink IR sources in different parts of the room, 
use brightness levels at the sensors to compute angle and 
position.  Could be noise and reflection sensitive, since it's
non-imaging.  Other method: pairs of sensors, and sweep the room with
narrow IR beams (laser of incandescent source.  Sensors tell when
the beam crosses both of them: 3 beamers and 2 sensors give you
angle and position.

IR cameras:  Cheap IR LED sources, cheap B&W cameras and an inexpensive
video processor.  Many waysa to extract angle and position from the
LED positions that the camera(s) see (2 cameras for easy depth thru stereo).
Put LEDs on the helmet for good positions sensing, or put the LEDs on 
the wall and the camera on the helmet for best angle sensing.

Myself, I've fooled around with all of these.  My money is (literally)
on the camera system-- I've been developing the hardware over the
past year.  Should be about $500 or so for the camera-card combination.
Speed: about 1 mS per frame per card/camera in CPU time, 60 fps, up
to 10 points (more points just take a little more CPU time). 

Have I missed out anything?


--------------------------------------------------------------------------
| My life is Hardware,                    |                              | 
| my destiny is Software,                 |         Dave Stampe          |
| my CPU is Wetware...                    |                              | 
| Anybody got a SDB I can borrow?         | dstamp@watserv1.uwaterloo.ca |
__________________________________________________________________________