> From:         llile@SALTONUSA.COM[SMTP:llile@SALTONUSA.COM]
> Sent:         Tuesday, November 11, 2003 4:35 PM
> To:   PICLIST@MITVMA.MIT.EDU
> Subject:      Re: [EE:] GPS vs LPS

> OK, we all know that GPS is good at locating positions down to some
> accuracy that is good enough to tell if you are on Mount Everest or not,
> but generally not accurately enough for many tasks.  For instance, a
> surveyor needs to know where he is within 2 cm, a robot lawnmower needs to
> know where it is within 6 inches maybe.  An automated farming machine, say
> a robot 4 bottom plow or a robot planter needs to know where it is
> withinhalf a meter maybe.  Any of these would benefit from a Local
> Positioning System, something that would cover just a few thousand square
> feet with immensely better relative accuracy.

> This might allow a robot lawnmower to employ efficient mowing algorithms,
> instead of random (efficient in terms of processor power) blundering
> about.

> How would one go about cooking this up?  I have scratched my head about it
> a little.

> 1.  Have a central GPS reciever at a stable position, say on a pole or one
> the roof of a house. ...

> 2. Duplicate GPS on a small scale.  Have 3 or more transmitters near the
> corners of your property/farm field/Yard at stable positions.  Continually
> broadcast signals from the three and your robot needs to figure out how to
> decode them into GPS-like distance signals.  Unfortunately, the GPS
> transmitters are synched in a very accurate way linked to the atomic
> clock, something that would be hard to manage in a small, cheap, homemade
> system.

> 3. Duplicate GPS with ultrasonic waves.  ...

It isn't really feasible to compete with the GPS system, at least not at its level
of complexity. This also eliminates radio methods, which might have to be licensed
and which have synchronization problems. You have the advantage that this is
for your own use (not for commercial sale), and will only be used over a small area.

Triangulation would be the way to go. First, consider this ideal example. Set up three
differently colored poles around the edge of the region of interest. On top of the
vehicle, put a rotating sensor which spins continuously in the horizontal plane. Each
time one of the poles comes into the sensor's center of vision, the sensor puts out a
pulse, along with a signal which identifies the color of the particular pole. The timing
of these three pulses could be analyzed to give the vehicle's position.

Now eliminate color as the determiner of pole identity. Put a flashing IR lamp on each
pole, with flash frequency different for each pole. This would also allow operation at
night, and would make it easier to filter out background light (look only for an AC
signal, within one of three frequency bands). Each lamp would have to have a
omnidirectional light pattern, which would rule out LEDs (or would it?). Use an
array of LEDs shining from within on an external diffuser?

The general idea is to use a simple mechanical scanner to perform the sampling of
the reference points. How to couple the signal between the fixed and rotating parts?
Don't. Use a rotating mirror to reflect the field of view down onto a stationary photo-
detector. The biggest problem will be discriminating the individual fiducial points.
A good solution, if it were possible, would be to place the reference points such that
there would be no ambiguity in identifying them. For example, if #1 were at 0 degrees,
#2 were at 45 degrees, and #3 were at 200 degrees, and if the "shorter" spacing
between #1 and #2 always held up, there would be no doubt which was #1. In the
continuous pulse train from the detector, #1 would always be the first in the pair
closest together. It would take a considerable amount of analysis to find 3 positions
which would always offer unambiguous identity and also allow unambiguous
position detection, but once found, the system would be easy to build.

John Power

--
http://www.piclist.com hint: To leave the PICList
mailto:piclist-unsubscribe-request@mitvma.mit.edu