Any designs for a cheap x-y position sensor?

[Robert Macy]

need a cheap, low power [read that as don't want to pay much, and need
it portable] x-y position sensor that tells me where I am when I ask..

specs
1. RELATIVE position once START is ok. Then can RESET and start again
from another marker.
2. over an area of approx 100 ft by 100 ft
3. update when requested, every 50, up to maybe 100mS. just need to
mark where I'm at on a grid.
4. absolute accuracy [over relative field] 'gentle' distortion of over
6 inches be ok.
5. relative accuracy for adjacent readings, better than +/- 1/4 inch,
probably get by with +/- 1 inch, but should be monotonic.

Note: absolute accuracy means a gentle stretching of the grid relative
to the perimeter is ok. but needs to be reproducible. Like exactly
where a mark is can be off by 1 foot, but it is always possible to
find the EXACT location within the tighter relative accuracy.
Monomotonic is like no backlash.

I could get away with wheels like those used by the realestate agents,
but the less paraphernalia/clutter the better plus likely to be in
rough terrain.

Be nice to simply stick a post in the ground and know where I am in
polar coordinates away from that stick when I ask.

I thought of GPS, relative, but expensive? power hungry? and updates
only every second ...sometimes. I thought of simple g force MEMS
devices double integrate, but their drift may preclude using that.

Anybody have a brilliant approach?

 

[[email protected]]

need a cheap, low power [read that as don't want to pay much, and need
it portable] x-y position sensor that tells me where I am when I ask..

specs
1. RELATIVE position once START is ok.  Then can RESET and start again
from another marker.
2. over an area of approx 100 ft by 100 ft
3. update when requested, every 50, up to maybe 100mS. just need to
mark where I'm at on a grid.
4. absolute accuracy [over relative field] 'gentle' distortion of over
6 inches be ok.
5. relative accuracy for adjacent readings, better than +/-  1/4 inch,
probably get by with +/- 1 inch, but should be monotonic.

Note: absolute accuracy means a gentle stretching of the grid relative
to the perimeter is ok. but needs to be reproducible. Like exactly
where a mark is can be off by 1 foot, but it is always possible to
find the EXACT location within the tighter relative accuracy.
Monomotonic is like no backlash.

I could get away with wheels like those used by the realestate agents,
but the less paraphernalia/clutter the better plus likely to be in
rough terrain.

Be nice to simply stick a post in the ground and know where I am in
polar coordinates away from that stick when I ask.

I thought of GPS, relative, but expensive? power hungry? and updates
only every second ...sometimes.  I thought of simple g force MEMS
devices double integrate, but their drift may preclude using that.

Anybody have a brilliant approach?

Laser scanner, I believe that is what they use at building sites
these day

Laser and rotating mirror measuring angles and maybe distance to
reference points I assume

-Lasse

 

[hamilton]

On 9/17/2012 3:11 PM, Robert Macy wrote:

Anybody have a brilliant approach?

Maybe you can extend this:

http://hackaday.com/2012/09/15/robotic-odometry-from-an-optical-mouse/

 

[Ecnerwal]

In a controlled environment (evidently not applicable) large scale
(billboard+) printers have been built using a spark gap (for acoustic
generation) on the print head and microphones at the edges. Read about
it, don't know the details, I'd assume the controller makes a spark and
only counts the first impulse at each microphone. IIRC the use of a
spark was in part to get a nice "sharp" sound that was not easily
confused with background.

Some variant on laser rangefinding would do the job, but cheap and 100ms
might not coincide too well. Park some retroreflective prisms at the
corners and go to town (look at what surveyors use these days - a "total
station" - good to 1/100th of a foot anyway at moderate ranges.)
Depending on available optics, moving the prism around and parking the
station would also work. For actual available tech, two bodies, and much
longer than 100mS updates (well, you have to walk the prism and park it
to get a reading - I don't know how fast the TS can track it as it
moves), one guy to point the TS at the prism and one guy to walk the
prism around the field is a typical approach, at least until the TS
needs to move in order to see the prism. That also gets you 3D.

I don't have a TS to play with, but TS and GPS are the reason I have a
transit that "used to be" reasonably nice and is now (and has been for
decades) "obsolete."

 

[Owen Roberts]

Optical version of a Aircraft VOR. Diode lasers are cheap. Mount them
at the corners and sides of the field on a rotating stage. Fan out the
beam with a cylindrical lens. You transmit a sync pulse, either
optically or by other means, when the laser is at true north. Look
for the laser light using a bandpass filter. Find the leading edge of
the beam as it passes over a detector. This gives you the angle at a
known time. Any two of them gives a fix. There are plenty of
papers on the net on multilateration.

If you have the leading and trailing edges of the beam, and its
diverging, you also have a measure of the distance in the far field of
the source.

Cheaper then time of flight.

Steve

 

[Jasen Betts]

Laser scanner, I believe that is what they use at building sites
these day Laser and rotating mirror measuring angles and maybe
distance to reference points I assume

Scanning lasers are only used for level (altitude) AFAIK

 

[Jasen Betts]

need a cheap, low power [read that as don't want to pay much, and need
it portable] x-y position sensor that tells me where I am when I ask..

specs
1. RELATIVE position once START is ok. Then can RESET and start again
from another marker.
2. over an area of approx 100 ft by 100 ft
3. update when requested, every 50, up to maybe 100mS. just need to
mark where I'm at on a grid.
4. absolute accuracy [over relative field] 'gentle' distortion of over
6 inches be ok.
5. relative accuracy for adjacent readings, better than +/- 1/4 inch,
probably get by with +/- 1 inch, but should be monotonic.

Note: absolute accuracy means a gentle stretching of the grid relative
to the perimeter is ok. but needs to be reproducible. Like exactly
where a mark is can be off by 1 foot, but it is always possible to
find the EXACT location within the tighter relative accuracy.
Monomotonic is like no backlash.

I could get away with wheels like those used by the realestate agents,
but the less paraphernalia/clutter the better plus likely to be in
rough terrain.

Be nice to simply stick a post in the ground and know where I am in
polar coordinates away from that stick when I ask.

I thought of GPS, relative, but expensive? power hungry? and updates
only every second ...sometimes. I thought of simple g force MEMS
devices double integrate, but their drift may preclude using that.

Anybody have a brilliant approach?

A bunch of cheap USB cameras, looking down on the scene from various
angles, and a bunch of software.

100 feet resolved to 0.25 inch is 4800 points. In X and Y that's a 4800
x 4800 grid. A 23 Mpixel camera would do it. Or four carefully
overlapped 8Mpixel cameras.

Subpixel resoulution should be possible, so it may be doable with
lower resolution cameras.

 

[Martin Brown]

need a cheap, low power [read that as don't want to pay much, and need
it portable] x-y position sensor that tells me where I am when I ask..

specs
1. RELATIVE position once START is ok. Then can RESET and start again
from another marker.
2. over an area of approx 100 ft by 100 ft
3. update when requested, every 50, up to maybe 100mS. just need to
mark where I'm at on a grid.
4. absolute accuracy [over relative field] 'gentle' distortion of over
6 inches be ok.
5. relative accuracy for adjacent readings, better than +/- 1/4 inch,
probably get by with +/- 1 inch, but should be monotonic.

Note: absolute accuracy means a gentle stretching of the grid relative
to the perimeter is ok. but needs to be reproducible. Like exactly
where a mark is can be off by 1 foot, but it is always possible to
find the EXACT location within the tighter relative accuracy.
Monomotonic is like no backlash.

I could get away with wheels like those used by the realestate agents,
but the less paraphernalia/clutter the better plus likely to be in
rough terrain.

Be nice to simply stick a post in the ground and know where I am in
polar coordinates away from that stick when I ask.

I thought of GPS, relative, but expensive? power hungry? and updates
only every second ...sometimes. I thought of simple g force MEMS
devices double integrate, but their drift may preclude using that.

Anybody have a brilliant approach?

Two surveyed wooden posts along a ~100' baseline and an ultrasonic tape
measure. You will have to put a repeater on top of the posts to send
back the pulse since in clear air there will not be enough echo.

Realistically you would probably want something to do the maths for you.

 

[Robert Macy]

I think that George guy who was asking about radio front ends a few
weeks ago was probably doing something like that.  Three transmitters
and a bit of signal processing will give you three intersecting
hyperbolas for a given set of phase differences.

Acoustics has much shorter wavelengths, but is more vulnerable to wind
and multiple reflections.

Another approach would be mechanical triangulation: two stakes, some
very thin steel cable, two spring-loaded reels, and two rotary encoders.
  Stretch one piece of cable between the stakes, so you know the
baseline, and connect the other ends of the cables to a third stake.
You get the two radii from the encoders, and can triangulate to your
position.

Acoustics is interesting, but I've got the soundcard tied up. Didn't
think of it, but could back off and go after 8 channel 44100 rates.

Interesting approach, a 'mechanical' solution. I can picture the
'tether' method. Like two tape measures attached to two poles.
Accuracy when directly between the two poles is the pits, but best
around 45 degrees from each. The 'field' easily extends past the two
poles also. Sadly, being in a remote area precludes lugging a lot of
weight up into the region. Plus, being tethered seems fraught with
physical breakage problems. But still, a somewhat unusual approach,
which might have applications for another project - inspection sites,
where the whole thing could be a fixture.

 

[Robert Macy]

Laser scanner, I believe that is what they use at building sites
these day

Laser and rotating mirror measuring angles and maybe distance to
reference points I assume

-Lasse

I think you might be right to just use a robust surveying tool.
Doesn't Zircon out of Campbell, CA make a laser tape measure for cheap?

 

[Robert Macy]

On 9/17/2012 3:11 PM, Robert Macy wrote:
<<snip>>




Maybe you can extend this:

http://hackaday.com/2012/09/15/robotic-odometry-from-an-optical-mouse/

Wow! didn't think of that at all! Thought of a mouse-like rolling
thingy, but the approach in that URL is much more self contained and
actually fits our existing software!

We have a prize winner!

 

[Robert Macy]

Near electromagnetic field strength measurement could return coordinates in
100 x 100 ft square with better then 1 ft accuracy.
Put a source of a field at unknown location, and two receivers at known
locations; solve for the position.
It could be done vise versa: two transmitters at known location and a
receiver at unknown location.
Of course, this method is prone to EMI and distortions due to large
conductive or ferrous objects; however it is simple and it could be more
robust then acoustics, optics or EM trilateration.

Vladimir Vassilevsky
 DSP and Mixed Signal Consultant
 www.abvolt.com

Great approach. Firt thing I thought of, too. but alas, I have some
fields already in existence around the system that the introduction of
any new field is likely to cause problems.

Just a footnote to readers: Distance measuring using magnetic fields
is really cheap and fairly accurate [relative terms]. I designed a
portable distance measuring system for a Medical Product which
essentially measured displacement along 16 axes, simultaneously. We
built a simple two axis unit for measuring displacement from 6 inches
to over 27 inches. Cost less than $5. No coil winding. Power
consumption less than 5 mA per axis at 5Vdc. The exciting thing was
that the MEASURED noise at a displacement of 8 inches was a mere 1 mil
rms!!

 

[Robert Macy]

In a controlled environment (evidently not applicable) large scale
(billboard+) printers have been built using a spark gap (for acoustic
generation) on the print head and microphones at the edges. Read about
it, don't know the details, I'd assume the controller makes a spark and
only counts the first impulse at each microphone. IIRC the use of a
spark was in part to get a nice "sharp" sound that was not easily
confused with background.

Some variant on laser rangefinding would do the job, but cheap and 100ms
might not coincide too well. Park some retroreflective prisms at the
corners and go to town (look at what surveyors use these days - a "total
station" - good to 1/100th of a foot anyway at moderate ranges.)
Depending on available optics, moving the prism around and parking the
station would also work. For actual available tech, two bodies, and much
longer than 100mS updates (well, you have to walk the prism and park it
to get a reading - I don't know how fast the TS can track it as it
moves), one guy to point the TS at the prism and one guy to walk the
prism around the field is a typical approach, at least until the TS
needs to move in order to see the prism. That also gets you 3D.

I don't have a TS to play with, but TS and GPS are the reason I have a
transit that "used to be" reasonably nice and is now (and has been for
decades) "obsolete."

Acoustics, I haven't considered very thoroughly, but could have some
effectiveness.Will think about that one.

The prism idea is interesting for its accuracy. But I think Ineed this
to be 'single' operator system.

 

[Robert Macy]

A bunch of cheap USB cameras, looking down on the scene from various
angles, and a bunch of software.

Hadn't considered optics much. But there is a lot of power in using
'many' cheap units and sorting things out in SW

 

[tm]

On 9/17/2012 3:11 PM, Robert Macy wrote:
<<snip>>




Maybe you can extend this:

http://hackaday.com/2012/09/15/robotic-odometry-from-an-optical-mouse/

Wow! didn't think of that at all! Thought of a mouse-like rolling
thingy, but the approach in that URL is much more self contained and
actually fits our existing software!

We have a prize winner!

=============================================

Good luck getting it to work over rough ground.

 

[Robert Macy]

Optical version of a Aircraft VOR.  Diode lasers are cheap. Mount them
at the corners and sides of the field on a rotating stage. Fan out the
beam with a cylindrical lens.  You transmit a sync pulse, either
optically or by other means, when the laser is at true north.  Look
for the laser light using a bandpass filter. Find the leading edge of
the beam as it passes over a detector.  This gives you the angle at a
known time.      Any two of them gives a fix. There are plenty of
papers on the net on multilateration.

If you have the leading and trailing edges of the beam, and its
diverging, you also have a measure of the distance in the far field of
the source.

 Cheaper then time of flight.

Steve

Sheer genius! I love USING the weakness - the divergence of the beam,
as a feature, rather than fighting it!

Now if I can just figure out a way to make a non-mechanical rotating
beam. pulse when goes by north, time around to find angle divergence
finds distance voila! polar coordinates! plus the resulting position
data should be monotonic. have some distortion have some error, but
never overlap. perfect.

Single pole stuck in the ground don't get too far away. battery
consumption to power a beam to overcome sunlight? snow light?
This technique is worth thinking about for the lawn robot!

 

[Robert Macy]

100 feet resolved to 0.25 inch is 4800 points. In X and Y that's a 4800
x 4800 grid. A 23 Mpixel camera would do it. Or four carefully
overlapped 8Mpixel cameras.

I don't know how flat a field of view they give you. Stitching four
fields together, and keeping the cameras sufficiently rigidly aligned to
keep the stitching stable, might well be challenging.

I don't know how often a camera supplies an image. But, stitching
shouldn't be a problem. I've done spline fit and stitching software
for some really 'distorted' images caused by views from two different
positions and it is surprising the results possible. - if you're
willing to stretch the final image, too.

 

[Robert Macy]

Scanning lasers are only used for level (altitude) AFAIK

Luckily, x-y should be all that's necessary. If I get into x, y, z
I'll also get into 'tilt' and end up with 6 axes to deal with! x-y
simulates the 'flat' earth, so should be sufficient. Better than not
knowing where you're at or how fast you're moving.

 

[Robert Macy]

Subpixel resoulution should be possible, so it may be doable with
lower resolution cameras.

--

good catch.

It would be easy to put a 'marker' on the unit and as the Operator
moves around, spot it in the camera's field of view. Live with the
camera rep rate and simple interpolate to find where the system is at
WHEN the system asks.

 

[Robert Macy]

Pole with webcam looking down at you?
The rest is software...

starting to lean that way, too. gross, numbers but getting pretty
cheap. Plus have a record of EXACT terrain for archives.