High accuracy wireless triangulation in a small space?

Question

the question pretty much states the basic idea, but here is my use-case; tabletop war-gaming.

Optimally I would like to be able to tag each of my “play pieces” with some sort of tag, place a few sensors around the perimeter of the play space, and have a device running an application that would be able to tell how far such and such tag/unit is from any other in the play area.

As such:

1. Such a system would need to be able to differentiate different ‘tags’, or if tags aren’t used, some method of doing so
2. Have millimeter or better accuracy, and scope over a variable play area approx. 5 to 10 feet cubed.
3. Be able to ‘refresh’ in a reasonable amount of time. (every few seconds is good enough)

Optionally

4. Some sort of powerless tag if its workable (so I don’t have to worry about switching batteries on hundreds of small figures)
5. As many ‘store bought’ components as possible. (and hence, hopefully reasonably affordable and with a minimum of home made parts)

I’m a software guy, so once something gets to the point where there are raw numbers coming into a computer I'm good. The figures just need to be found/tracked in the play-space accurately, and the computer can crunch the numbers to determine distances. As far as electronics, I’m having a hard time finding (see edit below) or even thinking of how to put this system together. I have looked at rfid tags, and they seem good for storing information, but not so good at triangulating distances. Any help or advice would be appreciated, but for this question;

What hardware/technology/setup can possibly (and hopefully nicely) fulfill the above requirements? Is there such a thing?

EDIT (plausibility)

I did some more searching and found; http://lunantech.blogspot.com/ The videos seem to demonstrate that this is possible? This gentleman seems to have achieved an effect similar to what I'm looking for. (I'm just not sure how, is this applicable?)

Likewise, the Soloshot device tracking a 'tag': https://www.youtube.com/watch?v=ApqQW5Nx1qI

Answer 1

Cameras, as many as you can get, and tons of image processing. Stick fluorescent stickers on your mini figures, or better yet, paint them with distinct fluorescent colors and you can get 1cm accuracy. Any radio below 30GHz will not give you the required accuracy due to the wavelength limit.

Answer 2

I'd consider using at least two HD webcams, looking from outside the table, across and down onto it, at more or less orthogonal directions. Run the cams into something running a suitable vision analysis library, OpenCV would work well, running from your favourite language, it will run under C, Python, Perl, many things. This will allow you to triangulate pieces in 3 dimensions. An alternative could be a single camera looking down, for 2D positioning.

Tagging the pieces? If they are visually distinct, and the lighting affords sufficient contrast, then there's no need to tag. Otherwise, perhaps a small vertical round rod, say wood dowel, so identical look from any direction, with bar-coded bands, for the 2 camera solution. An alternative for the single top camera would be a flat disc on the head, permitting a QR-like pattern, potentially more compact than the rod.

With just black and white, you would have max contrast, but need at least log2(number of pieces) bands. With more colours, you can get more bits per band, so with the spectrum, black and white, 8 colours is 3 bits per band. The Higher the Def of the camera, the smaller the tag's features can be.

The camera does not need to be in a calibrated position, if the table top includes a couple of calibration target.

When you get the recognition system working, you could add a projector, to throw hints or scenes onto the table.

Hey, have I just designed and given away a viable product?

Answer 3

There's lots of ways to approach this all with different pros and cons. 3D imaging will give you everything you've asked for but you'll need excellent lighting (and line of sight), and some pretty impressive hardware and software to go with it. You never mentioned what the refresh rate or size of the pieces needs to be so I'm going to make a number of assumptions along the way.

Doppler-effect triangulation with radio waves is going to be pretty much near impossible. I'm not saying it's impossible but given your DIY criteria just don't go down this path.

Triangulation with ultrasonic is feasible but parts are not tiny (if that's the size you're going for) and there will be a much larger power requirement than with radio.

An IR camera with each piece emitting a unique pulse would be cheap and would be battery energy efficient, but it will be very hard to get 3D.

I did a project like this once for triangulation in a building. We opted for signal strength as it changes significantly over small distances. The problem with this approach is that it fluctuates while the tag is static, and is easily influenced by surroundings. I think the solution here is to use 2 or more receivers at each detection station, then use the difference between the 2 to get a bearing. Compare the bearing with other stations to triangulate a position. If each game piece is transmitting a unique id, then the software requirements would be limited. Here's a quick sketch.

The biggest problem you're going to face other than technical feasibility, is how to power each game piece. RFID and other 'powerless' options are going make something challenging even more so, or impossible. You're either going to need large batteries relative to consumption, a low refresh rate to minimize battery needs (ie one location per second instead of hundreds), or small solar panels (ie solar calculator) to slowly build up enough charge to chirp a position.

Lastly, since you're a software guy, you're going to want to find a hardware guy to work with. If you can't find one then be prepared to do a lot of reading and learning. What you're trying to accomplish is not easy no matter who you are. I hope your succeed though because it would be pretty cool.

Answer 4

Use RFID radar.

Not sure what the resolution is these days.

I think they use two antennas for phase sensing to determine the direction and a third antenna for position.