Exact match. Not showing close matches.
'[EE] Beacon system for position location'
Anybody know of a website or links to a cheap beacon system
for postion location for a mobile robot ? I looked around
the net and everything was pretty expensive.
I was thinking of trying to use those cheap 315 mhz tx/rx
modules. I want to be able to postion an autonomous lawn
mower within 6 inches or so. I have an acre lot with a house in the
center so line of site is out.
I have a GPS receiver but that is only good to ten feet
with WAAS enabled.
Any other suggestions ?
Would a local version of WAAS work?
Use two GPS modules, one fixed and the other mobile. Fixed unit transmits location to the mobile, and the mobile unit calculates relative position.
Any drift in the fixed unit should be seen by the mobile unit as well. (Unless due to drift of internal calcs in the GPS module)
>>> On 30 Jul 08 at 10:25, in message
<shell.dca.net>, Kevin < Pine.LNX.4.64.0807301018280.32109dca.net> kben
That should work to some degree but not down to 6 inches unless you
use specialized differential GPS equipment.
I'm afraid that there is no cheap way to do what you want, Kevin. It
is a pretty hard problem, especially without line of sight.
What I'd be inclined to do is use odometry on the wheels most of the
time and just have some system for updating the position at some known
points to get rid of drift. For example, you could place several known
position markers in the yard and make sure that the robot goes past
them at least once per 100 feet of travel or somesuch. Also, you could
use a line of sight positioning method and just use odometry to handle
the times when there is no line of sight.
A final possibility would be to place some markers around the yard
which could be seen by a camera on the robot. Depending on which ones
it sees at which locations it could figure out is location.
On Wed, Jul 30, 2008 at 11:33 AM, Eoin Ross <chemstation.com> wrote: eross
what about just teaching it where to go? You would (well..guess with any system) make sure the yard was free of debris, but you put it into learn mode, take it around its path once, then have it repeat it each time. It would have to start at the same place each time but now you just have it programmed to go for x feet, turn so many degrees, go again x feet, etc.
Not as fancy, but might be easier to get the job done?
--- On Wed, 7/30/08, Sean Breheny <cornell.edu> wrote: shb7
I'm no RFID engineer, but along these lines, what about scattering
some RFID tags about the area it would be roaming in? In most areas,
have the tags spread in low density (*several* feet/yard/meters
apart). As you sense different tags coming and going, you know your
relative position. In the areas where you need high precision, bury a
few tags separated by some known distance (a small grid, maybe 3x3, of
tags). If you're picking up all 9 tags in a small grid, you're pretty
well on top of it... if you're only getting a few tags, you're off to
the side (and which side you're off to depends on which tags you can
On Wed, Jul 30, 2008 at 12:03 PM, Sean Breheny <cornell.edu> wrote: shb7
> A final possibility would be to place some markers around the yard
> which could be seen by a camera on the robot. Depending on which ones
> it sees at which locations it could figure out is location.
I'm not sure that it wouldn't work down to 6" or better. There's probably some assumptions I'm making that aren't right? Probably in the area of what resolution you can get out of cheaper GPS modules (I think I answered myself)
Basically the system would be a roll-your-own DGPS - The 'fixed' part would have to be just that - never moving in your yard.
It will give you some co-ordinates that will drift - but over 2-3 acres (?) I'd imagine an identical GPS unit would have almost identical position drift.
In the mobile unit you get the co-ordinates of the fixed unit, take your local co-ordinates, and calculate the offset - there's your position relative to 'home'
>>> On 30 Jul 08 at 12:03, in message
<mail.gmail.com>, "Sean Breheny" e726f69f0807300903x6b051027l22b2fd1d658d46ad
<cornell.edu> wrote: shb7
Actually, that gives me another idea. I am not opposed to
any solution, just trying to find a cheap one :) I have an
invisible fence for the dogs, so maybe I could have one
robot follow the invisible fence around the yard kind of
like a line following robot. Then that one could relay
course and heading to the lawn cutting robot. This would
allow line of site control.
I downloaded a couple of academic papers on beacons sysems
and it is definitly a non-trivial solution to the problem.
Thanks to all for the suggestions.
On Wed, 30 Jul 2008, Eoin Ross wrote:
|DGPS is great for reducing large errors in position, but it will still
have a severeal meter error at best.
It was meant to correct for the 100+ meter errors introduced intentionally
into the GPS system to keep 'the enemy' from using it, but now days it has
been turned off.
A do it yourself DGPS will certainly improve your accuracy but you will
still have several meters of error to deal with.
Programming a set route also is not going to work well, as ground
conditions will change causing at the very least traction slippage which
will quickly add up to large errors as it moves around.
What about ultrasonic triangulation? Lets see, speed of sound is 1100ft/s
so thats 13,000in/s so a tick rate of 13,000 per second gets you down to
an inch of percision. Three transmitters will do it, more if you have an
odd area that can't be covered by just three.
Another idea is a beacon like an IR led on the robot, and cheap USB
cameras looking down. From the position of the LED in the camera's view
you can easily calculate it's position. A system to do just that us used
to track foul balls in tennis. Then you just need enough cameras to cover
the area it will move about in, and a computer to calculate and transmit
the position to your robot.
I'm not an expert in this but I don't think this is quite right.
DGPS can be used to correct for errors caused by the ionosphere, which
as I understand it is the principal source of error in the present
system. As the signals go through the atmosphere, they experience a
varying propagation delay. However, if you measure the signals at two
points on the earth which are very close, then it is pretty guaranteed
that the two paths they took will have been through the same volume of
atmosphere/ionosphere, so whatever path length error is introduced
will be the same for both, being common-mode error not differential
I'm under the impression that a decent, run of the mill GPS receiver
with a full view of the sky typically gets in the 3 to 5 meter
accuracy range presently. I also thought that DGPS could get you
sub-meter, and that differential carrier-phase GPS could get you to
centimeter accuracy (but would cost something like $20k at least). If
you have access to the second GPS signal frequency, that can help even
further (I think that one is still encrypted).
On Wed, Jul 30, 2008 at 5:33 PM, <ian.org> wrote: piclist
Robin D. Bussell
>Another idea is a beacon like an IR led on the robot, and cheap USB
>cameras looking down. From the position of the LED in the camera's
>you can easily calculate it's position. A system to do just that us
>to track foul balls in tennis.
Maybe this chap's wii remote hacking projects will be of use, he's
tracking things very well:
Although you'd need a PC or similar with Bluetooth on the mower I
suppose... does the mower need to know where it is or does an outside
control system need to track the mower?
P.S. the stuff on this page is fantastic! (though nothing of any use for
lawnmower tracking :) )
|On Wed, 30 Jul 2008 17:33:19 -0400 (EDT)
ian.org wrote: piclist
> What about ultrasonic triangulation? Lets see, speed of sound is
> 1100ft/s so thats 13,000in/s so a tick rate of 13,000 per second gets
> you down to an inch of percision. Three transmitters will do it, more
> if you have an odd area that can't be covered by just three.
It will need temperature sensor - sound speed changes quite a lot (about
1% each 5 degrees Centigrade). That's not too much - about 1 meter over
50 m for a 10 degree C change - but still... I'd also worry about
reflections against walls and other obstacles.
Also, take into account wind direction/speed. A 10 knot wind (18km/h) is
about 5 m/s error on the 330m/s sound speed, another 1.5%.
> Another idea is a beacon like an IR led on the robot, and cheap USB
> cameras looking down. From the position of the LED in the camera's
> view you can easily calculate it's position. A system to do just that
> us used to track foul balls in tennis. Then you just need enough
> cameras to cover the area it will move about in, and a computer to
> calculate and transmit the position to your robot.
This might get complicated - USB cables only bridge 2m each... And,
ideally you'd have to put the cams _above_ the grass to be cut, else
resolution errors will cause large positional errors after perspective
correction (and I have seen distorsion of cheap camera lenses reach 10%).
I don't want this to sound as criticism... I'm as interested as the next
man to get my machine do its work solo ;-) (it doesn't - yet)
I haven't studied those laser measuring devices I've seen advertised.
Maybe they have some digital output that could be used to maintain a
constant distance from a wall or so.
|On Wed, 30 Jul 2008, John Coppens wrote:
> This might get complicated - USB cables only bridge 2m each... And,
> ideally you'd have to put the cams _above_ the grass to be cut, else
> resolution errors will cause large positional errors after perspective
> correction (and I have seen distorsion of cheap camera lenses reach 10%).
> I don't want this to sound as criticism... I'm as interested as the next
> man to get my machine do its work solo ;-) (it doesn't - yet)
I wouldn't post ideas if I didn't expect them to get poked with a stick. :)
I read some papers on optical tracking using cameras and one of the
techniques is to move an object around the cameras field of vision and
track it's real location with the screen position. You make up a grid of
mappings, and when you are done you just interpolate between all the
points so you end up with a matrix of screen to position values. The
studie found that if youi are just interpolating between small areas of
the screen, a simple linear model works perfectly, no fancy math involved.
USB cable length is an issue, alas. And wireless cameras are not exactly
cheap to throw half a dozen around to cover the yard if it stretches
around your entire house.
On Wed, 30 Jul 2008 19:53:05 -0400 (EDT)
ian.org wrote: piclist
> You make up a grid of
> mappings, and when you are done you just interpolate between all the
> points so you end up with a matrix of screen to position values.
Yes... The thing is that 640x480 cameras don't give much leeway to
interpolate, less if you look at an angle. That's what I meant in the
Given the relative slowness of the machine, cheap still cameras
could work with better resolution. Leaves the USB problem. There are
Ethernet cams too, but cost is similar to wireless.
> Anybody know of a website or links to a cheap beacon
> for postion location for a mobile robot ? I looked around
> the net and everything was pretty expensive.
Somewhat internally redundant content follows ...
LineS of sight may work for you.
Mower has a modulated tx beacon and eg rotating sensors at
various points that deduce the angle to the beacon. You
require enough sensors such that 3 can always see the mower.
You can also have multiple stationary senders and a single
scanning receiver on the mower. Probably cheaper and easier.
TX's could be coded eg IR sources but with care you may be
able to use simple uncoded omnidirectional sources and some
extra information from knowing the beacon position
If you use RF AND you can "see" through the building then
you can use radio direction finding - probably again with a
rotating sensor. If you can't 'see' through the building
then placing more sensors, with some outside the covered
area if necessary, allows lines of sight / RF that never
intersect the building.
Simple variant. Spinning IR "line of light" mounted on tower
(just a pole would do) high above the house so it 'looks
down' on mower. A detector on the mower and a knowledge of
tx position tells where the mower is on an angle around the
house. A second tx on another tall tower (tree, power on
fence etc) does similarly. Maybe a few of these on boundary
only. Turns back into original system with enough beacons.
You can also implement (as some have) a multi transmitter
phase measurement system to create your won GPS style
system. Not for the faint hearted.
Very cool demo of infrared dot tracking
A few "brainstorming" thoughts...
The big one: Make the rover as dumb as possible. Putting this
intelligence in the mower is silly... Build this in an older PC or
similar, and then have the radio control simply provide direction to the
mower. That is, the PC will determine where the mower is now, and
where it needs to go, and tell the mower to move say left or right or
forward or backward.
So, then you simply have to know where the rover is, within a few
1) Use radio direction finding. Place a "beacon" on the mower - just a
transmitter on a specific frequency, no modulation or anything. In the
yard area place one or more receivers which can detect phasing
differences between two antennas. The simplest way to do this is with
a pair of antennas and an antenna switch to switch between them, hooked
into a FM receiver on the same frequency. With suitable signal
processing, this will give you an angle from the antenna where the rover is.
2) Use high resolution video and some sort of video processing... You
will need to ensure something like 1 inch per pixel or similar and then
write some code to locate the mower. You should be able to make this
easy to locate by painting it fluorescent orange or similar. Of note,
a 1024x768 camera should be able to handle an 85x64 ft yard if you can
mount it in the right spot and one inch per pixel is sufficient
resolution (subject to experimentation). These are commonly available
for less than $40 in the form of a USB webcam. I think this is where I
would start.. The loop would look something like: grab a picture, look
for orange. Figure out where it is in relation to where it needs to
be. Command the mower. Repeat.
You may want to rethink the problem a little when it comes to
GPS. You don't need to know the absolute position of the lawn
mower only that cuts the whole yard and not one segment and
it is in your yard.
On Thu, 31 Jul 2008, Walter Banks wrote:
Some good brainstorming ideas ! Couple of issues I should
mention, I don't think I could get a camera looking down
because of several large trees in the yard. Also, the yard
is half level, the other half has a fall of probably 10 ft
in 60 ft. So, one section is a steep hill. That may be
a problem for a transmitter pole or sensor on top of the
I plan on making the mower as dumb as possible. I will have
my computer perform the processing and relay course and
direction to the mower through my wifi network or something
similiar (zigbee, bluetooth, etc...).
M. Adam Davis
Your yard sounds like it has a number of interesting topological
features! I wonder if you could put a triple axis accelerometer and
triple axis magnetometer in the mower to get the vector of the ground
surface and the magnetic field. If you map out your entire yard you
may find that between those two vectors you can roughly identify the
location of the mower. Place some large metal or magnetic objects
underground in areas where the sloping is not so great. Or put
magnetic loops underground with time-varying fields. Turn your yard
into a giant Wacom tablet!
I like the ultrasonic transducers idea, though. There's a lot of
literature out there about using ultrasonic transducers for measuring
distance. All you need (heh heh heh - there's that trivializing
phrase again!) is to place a dozen transducers around the yard, each
listening to each other and transmitting a code at a particular time.
Since they can be synchronized to each other, then you can gaurantee
that no transitter will be transmitting at the same time as another
The robot can hear at least 4 at any given time, and reflections can
be dealt with by only paying attention to the first sound heard. If a
reflection is messing up one of the transmissions, there's lots of
redundancy, so you can safely ignore that one.
Since they can hear each other, wind speed, direction, and temperature
can all be calculated and accounted for on a per measurment basis.
Or reverse that and have the robot beep with transducers around the
yard picking the sound up - but then you'll need auxiliary beeps to
determine wind speed, direction, and temperature.
Fun stuff to think about...
On 7/31/08, Kevin <dca.net> wrote: kben
On the camera thing... Nothing precludes you from using several cameras
scattered around...perhaps linked back to the base using wireless
transmissions. Same with other location ideas.
- A wire grid would relatively easy to install ("mole plow"
insertion) and probably as easy and accurate to drive as any
- A small UAV
1 flies a search grid looking for the mower.
2 carries a camera which views the mover and targets for
- A hot air ballon on a cable, powerd by a heater which
allows it to always stay aloft ...
Microphones mounted on the house ... .
- An ultrasonic ranger on the mower identifies IR beacons
and calculates their distance.
Utrasonic rangers detect the mowers location using a beacon
on the mower.
- A ...
Sounds like fun.
If somebody wanted me to show them a demo working within a
week I'd consider going for the mower carrying a rotating
optical detector with N modulated beacons. Demos in no time
are easy. Making the final product work is another matter.
'[EE] Beacon system for position location'
Please send us an update when you achieve 99.999% success ( or sooner
An autonomous robot to push the mower. Comes complete with beer/
chips ingesting orifice ( for energy ) , expandable thorax filled
with beer/chips ( for additional traction ).
On Fri, 01 Aug 2008 03:42:12 +1200
Apptech <paradise.net.nz> wrote: apptech
> Sounds like fun.
For an acre, I suggest a couple of sheep. Low maintenance, no computer
needed, no cameras, IR, remote. Proven efficiency, much less noise. Added
advantage is the wool...
Of course, this would not be an [EE] solution.
Some OP OP'ed:
>> I plan on making the mower as dumb as possible. I will have
>> my computer perform the processing and relay course and
>> direction to the mower through my wifi network or something
>> similiar (zigbee, bluetooth, etc...).
So, uh... belated reaction... Is your requirement really "Control
mower's position with 6 inches of precision" and not "Make it mow where
I want, but not where I don't want"? I'm wondering if it might be
easier to do it Roomba-style and put markers around to define where you
*don't* want it to go, then send it on some kind of random walk that
will generally mow about everything.
Timothy J. Weber
2008/8/2 Cedric Chang <nope9.com>:cc
> An autonomous robot to push the mower. Comes complete with beer/
> chips ingesting orifice ( for energy ) , expandable thorax filled
> with beer/chips ( for additional traction ).
Or make it run on grass clippings?
Richard Prosser wrote:
> 2008/8/2 Cedric Chang <nope9.com>:cc
> > An autonomous robot to push the mower. Comes complete with beer/
> > chips ingesting orifice ( for energy ) , expandable thorax filled
> > with beer/chips ( for additional traction ).
> Or make it run on grass clippings?
That's called a "cow"...
More... (looser matching)
- Last day of these posts
- In 2008
, 2009 only
- New search...