Relative Positioning

The "time of flight" for wireless, as you propose, is going to be very
difficult because electro-magenetic waves (speed of light) are
extremely fast. So, I woudl abandon that idea.  But, sound is an
alternative.  If each transceiver takes turns emitting an ultrasonic
"ping", then each of the other transceivers can record the time it
takes for the sound to travel to them.  But, you also need to let each
transciever know when the pulse was emitted by either a wired or
wireless transmission (an IR beacon is common, but you may not have
line-of sight to do this)

Here is one link to the type of system I am describing, as commonly
used in robotics;
http://www.restena.lu/convict/Jeunes/beacon.htm

In your case, you might send an IR signal (similar to those used in a
TV remote control) to tell a particular beacon to "Ping"... at the
same time the other transcievers read the IR signal to know which
transciever is doing the Ping and start to listen for it.

But, in the above method, you might have issues with some transcievers
not getting the IR signal.  But, if you are in a small-ish room, you
can just flood the room with the IR, and it may work.  Still, a wired
system is going to be a lot easier to do.

Joe Dunfee

    Sure. Look up "magnetic motion capture".  It's not easy to do, but
it's done routinely.

                John Nagle

In case you are attempting to use this for motion capture, probably the
best place to start is the wide variety of technical articles regarding
motion capture systems used for such things as movie special effects,
medical research, etc.

Maybe you need all these transceivers, or maybe what you can really use
is a jogging suit with (for example) triggered LEDs or fluorescent pads
embedded in it, and shot by a high-resolution digital camera. Motion
capture software translates the markers into distance and movement, and
can export that data for such things as integration with Maya or another
character animation program. You get what you are looking for (distance
between two objects) but of course this system only works for objects
visible to the camera.

John mentions magnetic AC trackers, but there's also acoustic and
several other technologies in regular use. Depending on the resolution
you need you might be able to use inertial tracking, like what a Wii
controller does. The solution really depends on how accurate you need to
be, how many frames per second of motion you need to record, and how
many markers you need to monitor.

-- Gordon



mattrapoport@gmail.com wrote:

Thank you for the thoughtful responses.  I've looked into the
solutions mentioned and unfortunately I don't believe they will suit
my purpose.  I realize now how important it was for me to stress the
required accuracy.  It should be accurate to the millimeter or better
if possible.  It also must be entirely self contained.  No cameras.
Onboard processor.

Many Thanks.

Again, we are a bit vague about your intended purpose and exact
situation.  But, another idea is to use a "string pot".  This is a
spring-loaded spool that has a flexable wire wrapped onto the spool.
Inside the spool is a rotary potentiometer.  As the wire is pulled
out, you determine the distance by the value of the potentiometer.

The drawback is that it is only a measurement between two points.
Though you can add multiple units, I am sure there is a practical
limit.  Also, they are not that cheap.

A poor-man's stubstitute might be to use the key-chain spools (or the
smaller ones designed to hold I.D. badges) and use a wire with some
resistance.  As the wire is pulled out, the current is forced to
travel down more of the wire, and therfore the resistance will
increase.  I haven't done this, so I am unsure how reliable it would
be.

If you want more ideas, you will have to be much more specific about
your project.

Joe Dunfee

Sounds like a great R&D project in itself. Maybe there's something like
this that has the accuracy, is completely self-contained, and doesn't
require line of sight (or sound). But most sensors of one type or
another fail in one or more of these categories. The majority are not
self contained, and must relay their information to a central processor.

Do also consider that a single sensor cannot determine its own relative
position. It can determine its distance to another sensor, but not and
X/Y position. For this you need multiple sensors, and they need to talk
to one another. A coordinated, modulated RF signal may work for this.
The interal processing of the data will no doubt need to be your own
project.

-- Gordon

    That's a great project.  Those are better specs than the motion
capture industry can currently achieve. Do you have venture capital?

    This is not impossible, but it's not cheap or something that can
be done with off the shelf technology.  The magnetic motion capture
systems have trouble getting better than 2cm accuracy.  The multiple
camera video systems do a bit better, but you have to be within
the camera space and there are still occlusion problems.

    I could see something with GPS, pseudolites, phase angle
measurement, local Bluetooth links, accelerometers, and an
IK system for cleanup.  Nice project.  Five people, two years,
about $5M, I think.

    An interesting option would be to use cameras looking outward
strapped to the user's joints.  During recording, just record
all the video.  In postprocessing, use SLAM to extract positional
data.  Then it's mostly a software problem.  If you're in some
environment with lots of texture the SLAM matching can latch onto,
this could work.

                John Nagle
                Animats

If I take an RF transmitter (A) and an RF receiver (B) and I precisely
record the time it takes for the the signal to get from A to B, I
wonder how accurate the distance measurement would be.  If I moved A
1mm away from B, would the recorded time accurately represent the
change in distance?

Thanks,

Matt

Your 1mm distance would represent 3.33x10^-11 seconds at the speed of light
(elecromagnetic radiation, including radio waves).  That's about 0.033
nanoseconds (unless my math fails me).  Can you measure time differences
with enough accuracy to reliably measure that kind of difference?

Jim

Hi Matt,

in principle you are right, the time difference would represent the change
in distance divided by the speed of light. So 1mm difference would
correspond to approximately 3ps time difference.

There are several sources of inaccuracies:
- the speed of light depends on the medium. In air, it's close to vacuum but
not quite, and depends somewhat on temperature, moisture, pressure etc.
- the circuits/equipment you use to measure such small time differences will
not be perfectly accurate, depending on how much you would want to spend on
this. In addition to initial inaccuracy (which you could calibrate out) the
errors might also depend on temperature, supply voltage, etc.
- the phase shift in the cables that connect the antennas to the equipment
often depends on e.g. bends in the cable and so on (unless you want to spend
many hundreds of dollars per yard of cable). So moving the antennas (and
therefore the cables) around might influence the measurement results
- at short distances (in the order of one wavelength or less) the coupling
between the antennas becomes somewhat complicated and involves other modes
than TEM waves - which will result in other phase shifts/time differences
than you would expect based on simple radio wave propagation
- however, the most important source of inaccuracies is probably the
reflections of radio waves off other objects and so multiple copies of the
radio signal will reach the receive antenna along different paths (this is
called multipath) and with different delays - very similar to what happens
with echo's and sound.

So, in summary, yes it will work, but it is not quite so easy and/or
accurate as you might imagine - sorry...

   Peter

 Approximation:

At the speed of light (radio wave, infrared beam, gamma ray, whatever) , 1
foot (30cm) is 1 nano second. (10 ^ -9 seconds)
(I seem to remember 11.7 inches, but hey, who's counting?)

1mm is 1/300 of this distance.

You do the rest of the math(s).

How were you planning on measuring the propagation delay?

Deep.

You'd probably want to use a modulated RF signal, and compare the phase
of the returned signal (from another sensor) with the original. This is
how some surveyors transits work. The circuitry is simpler, as you are
not measuring speed of light. You could probably achieve 1mm accuracy
depending on the frequencies used. How to make all this so that each
sensor is stand-alone and independent is another matter. If you use more
than two sensors you need a way of timing the pulses from each. There
will need to be some coordination between the sensors. There are other
problems to overcome that are left to the OP.

-- Gordon

    I built a LIDAR that way once, around 1990.  Surprisingly similar to
an FM radio, IF cans and all.  But nobody does it that
way any more. (Well, the Swiss Ranger does, which allows them to use
LEDs as the light source).  Today, you usually just count, really fast.
Here's a Bosch laser range finder for $129:
"http://www.amazon.com/Bosch-DLR165K-Laser-Rangefinder-Kit/dp/B000T7LISM" .
Accuracy is 1-2 mm.

    The main problem with working in this area is that while the
parts are cheap (they're derived from cell phone technology),
the test instruments required are expensive.

    If you want to work in this area, you'll need shelves full
of RF engineering books, some really good RF test gear, and the
software for designing boards with gigahertz transmission lines.
There are advanced radio hams who do such things.

    Look up "A UWB based Localization System for Indoor Robot Navigation"
for some hints on current (2007) thinking on the subject.

    This is a basically good idea which takes significant engineering
effort to make work, but probably won't be expensive to replicate.

                    John Nagle

Lots more helpful responses.  It sounds like it will be very difficult
to achieve.  It seems strange that we can get accuracy to a meter with
transceivers flying 20,000 kilometers above the earth (GPS) but we
can't get accuracy to a millimeter in a system where the transceivers
are all a couple feet from each other.

Thank you for all the help,

Matt

On Fri, 9 May 2008 10:27:43 -0700 (PDT), mattrapoport@gmail.com


Not strange at all unless one continues to believe that apples =
oranges.

    You can get that accuracy.  It just takes the quality of engineering
that goes into GPS receivers.  There has not, to date, been enough of
a market to justify the engineering investment.

    Look into the GPS systems used for "precision farming".

                John Nagle

The comparison with GPS is irrelevent anyway as the OP insists that the
sensors be self-contained, for whatever reason. The bulk of the system
of a GPS network is the satellites. The receiver merely collects data.
Si is correct; it's apples and oranges, even before the investment
starts.

-- Gordon

On Sat, 10 May 2008 11:14:23 -0700, John Nagle


 "I think somebody should come up with a way to breed a very
large shrimp. That way, you could ride him, then after you camped
at night, you could eat him. How about it, science?" - Deep
Thoughts by Jack Handy

Hello,
please let me know which one is the right newsgroup that I can ask question
about industrial robotics shool in dallas texas area. thanks

Peter