Controlling the path

"Ken O"

Well, everything is possible, given unlimited time and budget. :-)


I think the biggest problem here is finding the appropriate set of sensors
that will provide you enough accuracy.

For example, one could implement such a robot using a wheel encoder and a
gyroscope. In a perfect world, you'd just input the room dimensions and the
robot would make perfect 90 degree turns after running for a certain
distance.
On our less than perfect world, you will find that gyros drift, and wheel
encoders accuracy are dependent on the wheel material and type of surface.

On the other hand, there are autonomous agricultural machines that are
guided by expensive GPS systems (precision smaller than one foot) that are
programmed to work a plantation field in a similar fashion you described
above.

Cheers

Padu

  I thought about implementing  some beakons around the room, the robot
could communicate with them and triangulate itself in a given room size.
would that be possible?
I am confuses on how it will determine the distance.

ken

"Ken O"

Yes it is possible. There are a couple of commercially available solutions
such as tge NorthStar (expensive though)
http://www.evolution.com/products/northstar.masn

I don't know if 3.5mm is within their accuracy figures.

Cheers

Padu

Looks like that only gets accuracy down to a few CM - which is very good,
just not what Ken seems to be looking for.

--
Curt Welch                                            http://CurtWelch.Com/
curt@kcwc.com                                        http://NewsReader.Com/

would it be possible to tweak it to get accuracy??

ken

No clue.  I was just reading the specs from their site.  I don't even know
the details of how it works.

--
Curt Welch                                            http://CurtWelch.Com/
curt@kcwc.com                                        http://NewsReader.Com/

Sensors capable of <3.5mm resolution are rare, though not unheared of.
It becomes a question of money. Some laser sensors might be employed for
such a task.

Determine what type of sensor could be used to obtain approx. 1/8"
resolution over 8-10+ feet, consistently, and there's your answer. The
rest of the project will be easy. A mechanical equivalent to what you
want to do is the farm maypole, where the tractor or mule was attached
to the maypole via a rope. Every complete circuit the rope got a little
shorter, so the circle got smaller. Basically you're replicating a old
maypole farming technique, but without the rope.

-- Gordon

Doh!  I was just gonna suggest the rope!  Actually, fishing line tied to a
rod 1mm in diameter, with the other end tied to a motorized toy car.  ;^)

Seriously, though, what about a large turntable with a linear actuator mounted
on top (like the spoke of a wheel)?

Jeff.
--
Jeff Shirley
spam-hockey@mindspring.com
"Bill Gates is filthy rich, but that doesn't mean I want to be married to him."

Maybe you should consider the maypole trick?

Put a fixed stationary weight in the center of the room to act as a pivot
and run a cable from the object to your robot.  Put a spool in our robot
with an accurate barrel with an encoder on it and gearing to allow you to
make it unwind slowly but accurately.  The robot drives in circles around
the room tied by the cable to the center pivot.  To determine the radial
angle, you could do a few things.  You could attempt to use dead reckoning
by having the robot counting wheel rotations.  If you combine that with
some way for the robot to recognize when it has completed a full circle
(tape mark no the floor - fixed laser beacon extending out from the center)
then that might produce an accurate enough answer.  Or you could put some
type of encoder in the center post that would spin around with the cable
that would broadcast it's angular position to the bot.

Depending on the accuracy you need and the distance you have to work with,
you might have to put some type of constant tension device on the cable and
then do some calculations to adjust for the drooping of the cable as it got
farther out like surveyors have to do when using tape measures.

Using a mechanical system like that might save you some money over the cost
of the remote sensors you would need to get the type accuracy you seem to
want.

--
Curt Welch                                            http://CurtWelch.Com/
curt@kcwc.com                                        http://NewsReader.Com/

I will eventually need to evaluat he corners, not covered by this system.
Also, I do not want people to stomple on the wire, i would like to leave the
area around the robot free.

k

Ok.  Is this something you are going to set up for a single room you use
for testing or something you want to be able to take to different
locations?

If you are building a single room for testing, you might also think about
using a large X-Y positioning system attached to the celling like a crane.
It would be straight forward to get the accuracy you want with a fixed
system like that.

Doing it with a mobile bot seems like it might be tough.  You said the
spacing of the points should be 3.5 mm (right?) but you didn't say what the
accuracy needs to be.  Is it +- 3.5 mm?  Or do you need something like +-
.1 mm?

And how large of a room are you talking about?

How accurate are those laser range fingers used by surveyors these days?
Do they get down to that level of accuracy?  Maybe you can adopt that
technology so that it tracks and plots the location of the bot as it moves
around the room (and transmits the location to the bot if need be)?  That
way, you only need a reflective mirrors on the bot.  Just more food for
thought...

--
Curt Welch                                            http://CurtWelch.Com/
curt@kcwc.com                                        http://NewsReader.Com/

I want to take it to any room, Although i think of putting a laptop on it to
do all the calculations, so changing room will not be a problem


basically the robot will run lines 3.5mm apart, so probably a precision
error of 0.1mm to 0.5mm would be acceptable


could be 20 by 30 feet in area, ceiling of about .. i guess 12feet


that is an idea, I will have to look up into that

ken

For the fun of it, I've been doing some digging on the net to learn about
that stuff since I really don't know much of anything about that field.

The instrument surveyors use for measuring angles (though a scope) is a
theodolite.  They don't measure distance.  The digital theodolites have
encoders on them to electronically measure the position of the instrument.
A theodolite with built in EDM (electronic distance measurement (with a
laser rangefinder) is called a total station.  Their accuracy seems to be
on the order of 1 or 2 mm for distance and a few seconds for angles.
Accuracy no doubt a function of what you spend.  Costs seem to be $10K and
up but it's hard to find prices on the net for them.

I suspect they are normally designed to be aimed manually and don't include
automatic tracking.  But I think some might include automatic tracking to
allow a single operator to use them.  I don't know if they can track a
moving object or not.  Probably not.

So I think the accuracy of the technology might be in the rage you need,
but you probably won't find one that could be adapted to what you are
looking for.

However, in researching that, I found there's another type of produce
called a laser tracker that looks like it would work in that application.
They are designed for use in manufacturing to accurately measure points in
3D space.  They are designed to be mounted in a stationary location, and
then measure the distance using a laser rangefinder to different points in
the 3D space around them.  They seem to have multiple applications, from
measuring the volume of a container it's placed in to simply taking
accurate distant measurements.

In a deferential mode (which counts light interference pulses as you move
the target) they can produce measurements accurate to something like 3
microns.  The are used for example to check the accuracy of milling
machines.  But in the absolute distance mode (which simply uses the speed
of light to measure the distance) they are still accurate to something in
the range of .01 mm.  So, I'm guessing, you could find a device that would
accurately track your bot as it moved around, and allow you to hook it to a
computer to transmit the data to your mot, or to just record it.

However, I couldn't find any prices on line, and they look expensive.
($10K to $50K maybe?)

Here's some examples of the devices:

http://www.attinc.com/trackers.htm

http://www.leica-geosystems.com/corporate/en/products/laser_tracker/lgs_353
17.htm

Here's a video of the device in action:

http://www.leica-geosystems.com/common/shared/downloads/inc/downloader.asp?
idf13

It shows how in addition to the laser tracker, they sell a hand held laser
scanner that in the video is used to digitize the surface of an airplane
wing.  They just scan the wing with the hand held scanner (looks kinda like
a check-out UPC reader) while the laser tracker measures the distance to
the hand held scanner.  I assume the scanner has some sort of inertial
positioning system because it seems to know which way it's pointed.  Or
maybe the laser tracker is doing something like tracking multiple points on
the hand held scanner to determine it's position in space?

From looking at this technology, I'd say it would work fine for your
application of tracking your robot as it moved about the room in three
dimensions.  The only requirement is that you have line of sight from the
tracker to the bot (you don't even need to put the tracker in the center
because they have ranges to a hundred feet or so). But it's going to cost
you.

And from looking at that technology, I doubt you are going to get
resolutions better than a CM or so with any cheap technology.  If you
actually need resolution below a mm, you are probably going to have to go
with something like these laser trackers.

--
Curt Welch                                            http://CurtWelch.Com/
curt@kcwc.com                                        http://NewsReader.Com/

For the fun of it, I've been doing some digging on the net to learn about
that stuff since I really don't know much of anything about that field.

The instrument surveyors use for measuring angles (though a scope) is a
theodolite.  They don't measure distance.  The digital theodolites have
encoders on them to electronically measure the position of the instrument.
A theodolite with built in EDM (electronic distance measurement (with a
laser rangefinder) is called a total station.  Their accuracy seems to be
on the order of 1 or 2 mm for distance and a few seconds for angles.
Accuracy no doubt a function of what you spend.  Costs seem to be $10K and
up but it's hard to find prices on the net for them.

I suspect they are normally designed to be aimed manually and don't include
automatic tracking.  But I think some might include automatic tracking to
allow a single operator to use them.  I don't know if they can track a
moving object or not.  Probably not.

So I think the accuracy of the technology might be in the rage you need,
but you probably won't find one that could be adapted to what you are
looking for.

However, in researching that, I found there's another type of produce
called a laser tracker that looks like it would work in that application.
They are designed for use in manufacturing to accurately measure points in
3D space.  They are designed to be mounted in a stationary location, and
then measure the distance using a laser rangefinder to different points in
the 3D space around them.  They seem to have multiple applications, from
measuring the volume of a container it's placed in to simply taking
accurate distant measurements.

In a deferential mode (which counts light interference pulses as you move
the target) they can produce measurements accurate to something like 3
microns.  The are used for example to check the accuracy of milling
machines.  But in the absolute distance mode (which simply uses the speed
of light to measure the distance) they are still accurate to something in
the range of .01 mm.  So, I'm guessing, you could find a device that would
accurately track your bot as it moved around, and allow you to hook it to a
computer to transmit the data to your mot, or to just record it.

However, I couldn't find any prices on line, and they look expensive.
($10K to $50K maybe?)

Here's some examples of the devices:

http://www.attinc.com/trackers.htm

http://www.leica-geosystems.com/corporate/en/products/laser_tracker/lgs_353
17.htm

Here's a video of the device in action:

http://www.leica-geosystems.com/common/shared/downloads/inc/downloader.asp?
idf13

It shows how in addition to the laser tracker, they sell a hand held laser
scanner that in the video is used to digitize the surface of an airplane
wing.  They just scan the wing with the hand held scanner (looks kinda like
a check-out UPC reader) while the laser tracker measures the distance to
the hand held scanner.  I assume the scanner has some sort of inertial
positioning system because it seems to know which way it's pointed.  Or
maybe the laser tracker is doing something like tracking multiple points on
the hand held scanner to determine it's position in space?

From looking at this technology, I'd say it would work fine for your
application of tracking your robot as it moved about the room in three
dimensions.  The only requirement is that you have line of sight from the
tracker to the bot (you don't even need to put the tracker in the center
because they have ranges to a hundred feet or so). But it's going to cost
you.

And from looking at that technology, I doubt you are going to get
resolutions better than a cm or so with any cheap technology.  If you
actually need resolution below a mm, you are probably going to have to go
with something like these laser trackers.

--
Curt Welch                                            http://CurtWelch.Com/
curt@kcwc.com                                        http://NewsReader.Com/

Just out of interest, what's this for?

Michael

Measuring EM waves activites in a room up to 8 Giga Hz, so the robot need to
be very accurate and precise. I want to ba able to map  out a room in 3D
with Em acivities.

ken

If it is an emply test room, then I do think the May Pole idea could
get into the corners by using a few tricks;

Put an encoder on top of the May Pole, which can sense the direction of
the fishing line. Also a touch sensor on the outside front corner of
the robot. The data from the encoder in the center, and data from the
robot both need to be recorded, so perhaps you would prefer a fiber
optic cable instead of the fishing line. Of course, a wire data line
would work, but perhaps you mentioned fishing line because metal wire
would interfere with the measurements.

The robot spirals out until it senses the wall, where it runs along the
wall. You don't need your sensor data there, so don't record data while
the touch sensor is active. The fishing line will go limp and be
dragged a bit.

When the robot pulls the fishing line taught again, it will again
follow the circle. The touch sensor disengages, and data recording
begins again.

Regarding people tripping over the line... you don't want people in
there while you are taking readings anyway, right?

Joe Dunfee

You could put two or three SICK scanning laser rangefinders on a platform,
initially map the room and refer to subsequent readings to correct your
course.  I think these have +/- 1mm accuracy.  They are not small, but you
don't mention a size or weight constraint.

Mitch

Intersesting idea,  a bit too big, but I'll look up into it.

ken