Maker Pro
Maker Pro

Triangulating Location

TBennettcc

Dec 4, 2010
292
Joined
Dec 4, 2010
Messages
292
Gamenut,

I'm not sure where you're getting your numbers from. Light and radio waves are both forms of electromagnetic radiation, the speed of which (in a vacuum) is always 3.0 x 10^8 m/s. I don't understand your reasoning between the correlation of the speed of electromagnetic waves and the amount of transmissions able to be made? The amount of information transmitted and the time it takes to transmit that information is dependent on several factors, including frequency, how the information in encoded, bandwidth, encode and decode times, etc.

In order to make a measurement, the measurement has to have a reference point. They have a 'standard meter' and a 'standard kilogram' for reference at the International Bureau of Weights and Measures, to ensure that when someone says '1 kilogram', everyone is referencing the SAME kilogram. So is the problem with your project. Sure, it's easy to measure the length of time a signal was transmitted. But how do you know WHEN that signal was transmitted? Both the source and the target have to reference the SAME time source, not an easy task, especially when you are talking accuracy of +/- 1 mm.

The US GPS satellites #1) have their own atomic clocks on board, and #2) are constantly in contact with ground stations for error corrections and timing of the signals. Check out WAAS for more information. WAAS accuracy is claimed to be about 25 feet, which is pretty darn good considering those satellites are orbiting at an altitude of over 12,000 miles.

This might also be good reading: http://en.wikipedia.org/wiki/Global_Positioning_System#Correcting_a_GPS_receiver.27s_clock

Hope this sheds some light on the subject.
 

gamenut

Sep 22, 2011
12
Joined
Sep 22, 2011
Messages
12
The problem of using timing to determine when the signal was transmitted is that I can't measure in numbers that are that small. For a 10 meter range, and lets use the speed in a vacuum since it is easier to use for now, that means that it will take about 1/3.0x10^7 seconds, which is very, very small. My idea is to just use a computer as the constant for time, tell each transceiver when to start transmitting and when to stop, and use a ping-pong effect. Basically, there is no information encoded into the transmission. As for frequency, I'm not sure about that. The main point is that when that receiver gets the transmission, it adds 1 to the current transmission counter. In the end, if it received, say, 6x10^8 transmissions, then we can tell it is 2 meters away by dividing by the speed of the wave.
 

(*steve*)

¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd
Moderator
Jan 21, 2010
25,510
Joined
Jan 21, 2010
Messages
25,510
The main point is that when that receiver gets the transmission, it adds 1 to the current transmission counter. In the end, if it received, say, 6x10^8 transmissions, then we can tell it is 2 meters away by dividing by the speed of the wave.

That sounds like some crazy sort of radar? Any way you do it, you have to time in increments of 1x10^-11 seconds to get mm resolution(*). You also need to use a wavelength of a mm or less. Very demanding.

(*) if you can count at that rate, you can also time intervals that small.
 

gamenut

Sep 22, 2011
12
Joined
Sep 22, 2011
Messages
12
Good point. Guess I missed that. One more idea, is it possible to change the range of the transmitter by altering the power applied? Or any way at all? If you reduced the range to say 25 and the receiver didn't pick it up, then the receiver is more than 25 meters away.
 

gamenut

Sep 22, 2011
12
Joined
Sep 22, 2011
Messages
12
A better way to put this is say you have an object that is 5,555 millimeters away. To find the distance, can you change the range of the transmission to see if it received? So using 10,000 millimeters, can't you break it up, for example :

Range :
10,000 millimeters ------ Received
9,000 millimeters -------- Received
8,000 millimeters -------- Received
7,000 millimeters -------- Received
6,000 millimeters -------- Received
5,000 millimeters -------- Not Received
_______________________________

Next Find the hundreds value :
RAnge :
6,000 millimeters -------- Received
5,900 millimeters -------- Received
5,800 millimeters -------- Received
5,700 millimeters -------- Received
5,600 millimeters -------- Received
5,500 millimeters -------- Not Received

And so on.
Would this work, basically can I decrease the range of the transmission like this?
 

poor mystic

Apr 8, 2011
1,074
Joined
Apr 8, 2011
Messages
1,074
I don't know of a method you could use to reduce the range of a transmission without increasing noise. The power received at a receiver falls off with the square of distance from the transmitter.
The only possibilities I believe in use light, like Hawk Eye.
 

davenn

Moderator
Sep 5, 2009
14,254
Joined
Sep 5, 2009
Messages
14,254
A better way to put this is say you have an object that is 5,555 millimeters away. To find the distance, can you change the range of the transmission to see if it received? So using 10,000 millimeters, can't you break it up, for example :

Range :
10,000 millimeters ------ Received
9,000 millimeters -------- Received
8,000 millimeters -------- Received
7,000 millimeters -------- Received
6,000 millimeters -------- Received
5,000 millimeters -------- Not Received
And so on.
Would this work, basically can I decrease the range of the transmission like this?

you cant do that with a radio signal, as poor mystic hinted at the signal strength of the radio field isnt constant. It is affected by many things including the enviroment. You just cant control a radio signal strength accurately enough to do that sort of measurement.

for example depending on conditions my 1W, 30MHz signal can only get across the city, other times its enough to travel 1000's of km's

as an aside.... changing freq doesnt change the accuracy, but it does change the resolution capability of the system ( radar) The higher the frequency, the better the resolution of smaller objects

The accuracies even for commercial radar are not as good as what you are wanting, it is + - many metres in any axis.
Hi accuracy and very expensive GPS location equip, like I work with, for construction surveying is still only down to a bit less than 0.2 metre (200mm). Surveyors have to then do a lot of additional work to bring that down to an accuracy of less than 10mm.

Dave
 

daddles

Jun 10, 2011
443
Joined
Jun 10, 2011
Messages
443
Hi accuracy and very expensive GPS location equip, like I work with, for construction surveying is still only down to a bit less than 0.2 metre (200mm). Surveyors have to then do a lot of additional work to bring that down to an accuracy of less than 10mm.
And it's certainly not real-time...
 

poor mystic

Apr 8, 2011
1,074
Joined
Apr 8, 2011
Messages
1,074
I wrote to my friend who knows more than I, and asked him to review this thread and advise of any suitable software. I received this response:


re:
I think there is some fancy hardware keeping the cameras synchronised,
and a fast network feeding the vision into a purpose-made computer
cluster. Those little quads probably have 512Hz motor controlers
(certainly not less than 300Hz), so the outer stabilisation loop will
be running > 1 KHz. There is no way the vision system will be sampling
position at >2 KHz! (Shannon-Niquist sampling theorem), so the
computers must have an internal state model running at that sort of
speed instead (e.g. some sort of Kalman Filter, probably EKF or
Scented KF). My guess is the vision system is maybe sampling at 100
Hz, because more than that is likely to run into optics issues. They
are probably capable of "close to mm" accuracy, they have youtube
clips of a quad playing a piano, and the rubber finger looks like it's
hitting the middle of the keys to me.

If you only wanted a 1Hz position solution, you could oversample the
vision with cheap web cams and wouldn't have to do the whole
intermediate state model thing. All the computer science has been done
for you and is available in the OpenCV library, however don't
underestimate the difficulty of the task, it's an impressive project.
In my opinion, if you are talking about a small well lit room, and you
don't mind waiting a few seconds for you answer, then a single
computer with a bunch of USB cameras will probably do the job.

Alternatively, check out off-the-shelf LIDAR modules (scanning
rangefinders). That would cost more but possibly be less development
effort. Certainly sub-mm accuracy would be possible with LIDAR though,
those systems have been available and used by architectural surveyors
for over a decade (e.g. scanning the inside of Cathedrals, etc).

 
Last edited:

gamenut

Sep 22, 2011
12
Joined
Sep 22, 2011
Messages
12
Alright, well how about this question. Radar guns use radio waves to find out how fast a car is moving, depending on the time it takes for the wave to return. Considering how accurate a number they get back, and occasionally from a distance of 500 meters, how is that possible?
 

davenn

Moderator
Sep 5, 2009
14,254
Joined
Sep 5, 2009
Messages
14,254
radar and laser guns use doppler shift , they measure a small change in frequency which is relative to the speed of the object. Their accurcey is still no better than 1 km / hr and it has no
ability to fix a location. And they are quite expensive.

Sorry we all sound so negative. We went through this identical topic some months ago with another forum member. other than the horrifically expensive camera system mentioned above.
There is nothing "off the shelf" that I am aware of that can do positional fixing with the sort of accuracies you are wanting

cheers
Dave
 

gamenut

Sep 22, 2011
12
Joined
Sep 22, 2011
Messages
12
Its fine. Better to learn now then later. Now one more question. Say you had that camera system, or 4 basic cameras spread out in the 4 corners of your room. You then used some type of identifier for the "receiver", like a bright pink color, something that is uncommon. Then the camera looks for that color, and gets the position based on the other 3 views of it. How do you think that would work? Similar to the xbox's kinect system, if you have seen that.
 

(*steve*)

¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd
Moderator
Jan 21, 2010
25,510
Joined
Jan 21, 2010
Messages
25,510
Essentially you would be triangulating like a surveyor.

You would take the azimuth and elevation from each camera position and do some fairly simple math.

Your accuracy would depend a lot on the resolution of your cameras, the size of the "pink" ball and the accuracy of position of the camera. The latter is probably the accuracy of your knowledge of the frontal node of the lens used in the camera (and that may not actually be within the camera!).

I'd possibly go for a flashing light as that is somewhat easier to fins in a series of images, and also serves to synchronise the cameras so you know they're all getting direction data at the same time.

mm accuracy? doubt it. Maybe to within 30 cm, The quality of your cameras will be the determining factor.
 
Top