Searching \ for '[EE]: Positioning' in subject line. ()
Help us get a faster server
FAQ page: www.piclist.com/techref/index.htm?key=positioning
Search entire site for: 'Positioning'.

Exact match. Not showing close matches.
'[EE]: Positioning'
2001\01\14@174403 by

I am looking at an application requiring multiple objects (say up to 10 max)
to be positioned in about a 40cmx40cmx40cm area. I would be looking at
achieving a usable accuracy of 7bit minimum but would quite like to have
9-10bits of usable range. Hence I'd probably be looking at obtaining rough
8-12bit input which can then be filtered. Ideally the sample rate would be
moderately high, say at least 1kHz per object. However such specs are not
final.

The main method I've come up with is capacitative ranging from multiple
points fixed on a base unit, then mapping the various capacitances
(distances) to cartesian coordinates. However I'm not sure:
1) About the effectiveness of such a unit with multiple objects to track.
2) The distances this will be effective over. I'm yet to calculate the level
of capacitance you would need to measure in such a situation.
3) The speed achievable with such a method.

Other possible methods I thought of were:
1) Angular determination. If the angle from fixed base points could be
determined triangulaion could be used.
2) Some other form of distance measurment via signal intensity. I'm not sure
what signals would be effective. Maybe by using RF transmission techniques
and power measurement the sensitivity required could be reduced.

I would imagine there would be a moderate amount of DSP involved in
converting sensor readings to cartesian coordinates and hence would like to
use DSP to improve resolution as much as possible (e.g. Digital background
noise removal, multiple object handling via frequency division).

TIA,
Tom.

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.

Brandon, Tom wrote:
{Quote hidden}

There is a lot of good stuff on the net re using
a cheap CCD camera for object detection. Your
requirements may suit this via top down camera
and/or side camera? Search some of the uni
sites, and even the robot soccer, there are
algorythms and hardware designs and stuff.
-Roman

--
http://www.piclist.com#nomail Going offline? Don't AutoReply us!
email listservmitvma.mit.edu with SET PICList DIGEST in the body

Sorry should have given an idea of object size\use. Would ideally like to be
able to track an object less then 1cm cube (this could\will be connected to
a larger device with most of the circuitry so could just be an 'antenna').
This object should be able to be attached to the hand (or other objects
ideally irrelevant of object (may have to be fairly non-conductive say)
hence a CCD system would prob. make it hard to isolate the target from the
hand (esp. as the object may be visually concealed by the hand or another
object. Maybe a couple of CCD's but then it gets harder, also, I'd like to
avoid having a structure around the area. Ideally only a base would be
required.

Thanks anyway,
Tom.

{Original Message removed}
Can your objects be smart enough that they could be polled so that only one
at a time would be actively responding to the base?

----- Original Message -----
From: Brandon, Tom <tomPSY.UNSW.EDU.AU>
To: <PICLISTMITVMA.MIT.EDU>
Sent: Monday, January 15, 2001 6:33 PM
Subject: Re: [EE]: Positioning

> Sorry should have given an idea of object size\use. Would ideally like to
be
> able to track an object less then 1cm cube (this could\will be connected
to
{Quote hidden}

> {Original Message removed}
Sorry, yes that is certainly possible. I would expect the devices to be
cabled to the device unit, so while the devices may not be smart enough they
will be in constant contact with the base unit. I am happy with the base
unit performing large portions of the processing. My only desire is that a
single base unit can handle a moderately variable number of devices, hence
I'd rather not have too much base circuitry devoted to individual devices.

I would expect to use some form of time slicing for noise elimination (e.g.
switch off object 'transmitter', take background noise reading, switch on
device, take positional reading). Hence multiple objects could use time
slicing.

I'd probably slightly prefer non-time based discrimination as it would
increase sampling rate. e.g. If I could sample once and seperate devices by
frequency, only one set of readings would be needed for 5 devices (or 10
devices) instead of 5 in a time based system. And in systems such as
capacitative ranging I would prob. use some from of bandpass filtering for
noise elimination, hence a single Fourier transform could be used to
ascertain readings for a solely bandwidth\computation limited number of
devices.

Tom.
{Original Message removed}
Ultrasound is the simplest solution.

Sources on the 4 verticies of your base (6 is better), and small
crystals
(microphones) with AGC preamps on your objects. You crank up the gain
with time since it means you're father from the sensor and so the signal
will be weaker when you finally hear it.

You would normally fire each source
in succession, but since you want want a 1khz sample rate (and your
volume is too big for this with sound traveling at 750ft/s/1000=0.75
of a foot) you would want to fire them simultaneously, but with
different
assigned frequencies groups for each sensor, stepping through the
groups, and then using a DSP on each sensor to pull out the true
time of flight for each transmitters tone sequence.

IOW you fire a sequence of 4 different tones at a 4Khz rate through
each sensor to get coverage of the volume you've spec'd.
You sample the sensor, extracting each tone and it's time, and then
compute your position. By using 4 transmitter you can compute a 'closure
error', and add as many objects as you want, each with it's own
processor (or time multiplexed at very high rates (determined by your
resolution requirements)).

You could also fire the sources sequentially, with each
burst being a different frequency, and short duration, but you don't
gain anything since you can still have signal overlap at any one sensor,
depending on where it is wrt the transmitters.

Don't forget to compensate for temperature and
humidity effects on the velocity of sound.

You would NOT use amplitude/distance since this is unrelable,
particularly

One of the researchers here has a commercial unit that uses a couple
of 2mm square crystals to measure rat lung motion/displacement over many
cm separation.

I also have a 2D 'sonic digitizer' which used a pen which generated an
audible
spark, which was then picked up by two linear microphones. Worked quite
well, but your high sample rate requirement precludes using that
technique.

You could also do this with RF and multiple channels of phase detector
(or time of flight but you'd need Ghz sample rates).
Basically, each corner has a small antenna radiating RF pulses (to avoid
reflection problems) at different frequencies. Your objects have small
dual loop antennas (to avoid cancelation effects)
that pick up the emissions and do a phase measurement of
the burst. At 100Mhz you have a 3 meter volume to work in without any
'special' processing. Using RF allows your
sample rates to be incredibly high. If your objects and immediate area
are
none conductive, then you could get away with CW RF.

You might also want to look at commercial 6 DOF sensor systems as used
for motion capture for video games and movies.
Polhemus Systems, and Ascension Technologies are two \$\$ commercial
turnkey systems that will do what you want. I recommend Ascention's
'Flock of
Birds' since their technology is immune to metalic reflection effects
(eddie current) and allows for high sample rates, and their sensors
are also much smaller (on the order of a 1" cube).
The active space for both is on the order of 3 Meters. Extra
transmitters increase your working area. Both use magnetic technologies.

http://www.ascension-tech.com/
http://www.polhemus.com/trackers

A full list of input devices at
http://www.billbuxton.com/InputSources.html
look under 'motion capture'

Now where do I send the bill for consulting fees <G>?

"Brandon, Tom" wrote:
{Quote hidden}

> {Original Message removed}

More... (looser matching)
- Last day of these posts
- In 2001 , 2002 only
- Today
- New search...