Tracking an object by radio

Factual Questions
1

Hey all,
Im doing a project whereby I have an object I need to keep track of as it moves in a 3 Dimensional area of aprox 1 meter cubed. The tracking needs to be extremely accurate, and it needs to interface with the computer (so as the object moves around, it generates a series of co ordinates (x,y,z) relative to its starting position (0,0,0))

What sort of choices do I have here? At first I thought of using 3 radio transmitters placed around this 1 meter cubed area, then tri-angulating the position of the object based on the phase of the recieving wave (the object would have an antenna to recieve the radio waves). Is this possible, and if so, what sort of equipment would I need?

I was thinking I could also use GPS, or a camera module, but this would skyrocket the price instantly.

Cheers guys,
Matt Morten

2

RADAR? The signal would give you your range, and the azimuth and elevation can be recorded by reading them from the pan/tilt antenna mount. (Of course you would want the pan and tilt – azimuth and elevation – to be read electronically with sensors attached to the mount.)

3

That sounds like a job for a Theremin.

4

GPS is no good; it doesn’t have the resolution you need.

Anything involving reading tilt and pan mechanically will have terrible precision and probably be unworkable.

How much is “extremely accurate”? millimeters, micrometers or nanometers?

In general I’d suggest that for ease of calibration and general build-ability you should plan on some variety of 3-axis range-only solution. Such as …

If your object is a shiny sphere you could use 3 LED receiver/transmitter units mounted on three sides to form a very short-range LIDAR setup. Flash an LED and time the interval for the resulting reflection to be detected adjacent to the trasnmitter. That range-only info from 3 sides could then be computationally combined to give you an (x, y, z) coordinate. By cycling through the 3 LED transmitters at several KHz, or even MHz, you can achieve very fine-grained tracking.

This idea, or a variation using sound or RF instead of LEDs, has the advantage the your object is totally passive. But it requires that the object be a sphere, since we’re measuring the distance to its surface.

if the object is irregularly shaped and you’re only interested in position, not orientation, then you could put a trasmitter (RF or optical) in/near the center of your object and trigger it with an external starting-clock pulse delivered via RF or optical. Again 3 range-only detectors , triggered by the same starting clock pulse would measure the response time of flight, and then your computer could do the trig to resolve that to (x, y, z).
As to specific hardware, I’m not enough of an expert to make recommendations.

5

My experience is with ICBMs. You’re right; they’re a little more than a meter in size. :wink:

6

IANA Expert, but I think you would have better success triangulating if you had 1 transmitter (on the object) and 3 receivers. Each receiver would need to determine the angle of reception and the 3 receivers together could be used to compute the exact position.

7

Cheers guys for all your suggestions.

The object in question is an irregular shape, with pieces of itself possibly entering into the path of one of the transmissions (hence im not sure how effective a LIDAR would be).

I would need it to be accurate to at least 2mm if possible.

Having the transmitter on the object sounds like a good idea, yet it must be relatively small (nothing larger than could be held in your hand, for example).

Ill look into some equipment based on LIDAR - I can imagine it will be kinda tricky to find this sorts of equipment that interfaces with a PC

:dubious:

8

Actually, GPS would be a fairly decent way to go at this.

The absolute accuracy of GPS is pretty crappy, no argument there.

The relative accuracy of GPS can be phenomenal. That is to say, if you have two identical receivers within a small area, then you can calculate the positional difference between the two quite accurately.

Differential GPS makes use of this to provide postional accuracies of less than a meter. This is done over very large areas (broadcast range of a receiver with a known position,) so the accuracy gets degraded by the different atmosphere conditions and other things.

I will grant you that the NMEA data packets won’t be of much use. Most of the ones I know of only go to (at most) 5 digits after the decimal point for the seconds of arc, which limits you when you need centimeter precision.

You’ll probably need to get a development kit from the GPS receiver manufacturer and pull the data out of it in the manufacturer’s proprietary format. The proprietary formats usually offer raw data that goes to more decimal places. They also can usually output the pseudo-range and other data that you would need to compute the positions with your own software. If you do that, then you can get really down to the nitty gritty.

You set one receiver (actually, the antenna) and use it as your zero point. You attach the antenna of the second receiver to the object being tracked. Get your coordinates and figure the difference then convert to meters. No sweat.

Make sure that the cables for the antennas are the same length to within the precision you need on your measurements. The length of the cable influences the apparent distance from the satellites, so that if the cable lengths are (say) different by one meter then you will have a meter offset somewhere when you calculate the differences.

As for cost, you’re talking a couple of hundred bucks tops for a pair of receivers. You’d have to write the software yourself, but then you’d have to do that no matter what.

Triangulation from a radio signal would be a bitch to handle for what you are doing. You need some way to resolve time differences accurately at below 3.3 picoseconds (time for the signal to cross 1 meter) or else some way to manipulate two directional antennas on two axis each in real time. You will also need to use a very high frequency radio signal to get any accuracy. A 320 MHz signal has a wavelength of 93 cm, so if you need centimeter accuracy you’ll need to go up to something around 3 to 4 GHz - this because you can’t “see” really anything that is smaller than the wavelength of the “light” you are using.

The same requirements on timing accuracy are valid for LSL Guy’s “LIDAR” setup as well. Working things over, you’re going to need a clock running at 300GHz to get 1 meter accuracy. If you need cm accuracy, then you need a 30000GHz clock to get it. Expensive shit. You’d do better with a continuous wave signal and some kind of phase comparison - and then you get into doppler effects so that you’d have postion indications getting distorted depending on how fast you object is moving. Not to say you won’t have that kind of thing with any other solution.

Another way to tackle it would be with a sort of inverse sonar. You pack an audio signal generator into your target so that it broadcasts a single tone. Then you use three microphones and a great pile of FFT stuff to get phase information out of three audio signals, and from that you can get ranges that you can use to calculate the position. A tone of about 170 Hz would give you a phase range of about two meters before wrap around would cause ambiguities.

Another way using audio would be to put a clicker in your target. Your target is connected (electrically or by radio) to your test rig so that a click starts three timers. The times recorded between the click and when the three mikes pick up the click give you your ranges which you then use to figure position. Counters with a clock frequency in the megahertz range would give you tremendous accuracy. A clock frequency of around 34KHz would get you 1 centimeter accuracy.

Lots of ideas there. I hope you can use one of them.

Have lots of fun.

9

I think you missed a factor of 1000. Ballpark, light travels 1 foot per nanosecond, or 3 nanoseconds per meter. That makes the RF hardware a LOT easier.

That nitpick aside I agree with all you’ve said about precision required. Using any EM system will require very high clock speeds and he’s probably better off using sound.

I’m familiar with DGPS. I just wasn’t aware that the resolution (ie precision, not accuracy) was available to that fine a level. I thought it quit at about 1/2 meter. Thanks for the insight.

10

Erp. 3.3E-09 seconds =3.3 nanoseconds, not picoseconds. Head. Ass. Extraction tool.

Sorry.

The application I’m talking about isn’t true differential GPS in that it won’t tell you much about exactly where you are - it only tells you how far apart two antennas are. The high accuracy come from the two units being so close together. The thing that limits the accuracy on normal DGPS is the differing atmospheric conditions along the signal paths. Also, DGPS works by figuring out corection data for the pseudo range calculations and passing the correction factors on to the other receivers for them to use. The application I’m talking about doesn’t even attempt to correct the errors - it just assumes that the errors are the same since the two receivers are so close together.

The application I first heard of using this was an attitude detection system for an aircraft. With the two antenna located on top of the plane the system could give you pretty damned accurate information of the up or down angle along the long axis of the plane. More antennas could be used for the other axis.

Another that I’ve read of, though never seen in action, was a forklift that could tell the height of the fork this way. Only good out of doors, of course.

Unless I’ve messed up the math again, 5 decimal places on arc seconds gets you down to 0.3cm just using the NMEA data.

It should be possible to get the needed precison, especially if the receivers can put out raw data for further processing.

11

Sounds to me like you are basically re-inventing the wheel, which in this case the wheel is a motion capture system. 3-D game developers use this sort of thing all the time. Most of the systems I’ve seen have been optical, not RF. This link has some info that you may find useful (click on overview of current input systems):

http://www.siggraph.org/education/materials/HyperGraph/animation/character_animation/motion_capture/motion_optical.htm

Googling “motion capture system” will get you lots of reading material as well.

12

How about aiming two webcams at the space and triangulating the position from the two images?