I’m starting a new system design in which I need to get the position of an object inside a
parallelepiped shaped volume (2m x 1m x 1m).
If I place 8 anchors (one anchor at each of the 8 corners), is it possible to get the position of a tag (contained in such parallelepiped shaped volume) with a 0.5cm accuracy?
Which Decawave device do you suggest?
0.5cm i found it very difficult (and only using ranges (ToF or TDoA) to calculate the position). You can obtain something “near <10cm for example” with the position calculation depending on different things like GDOP, LOS, algorithms and even using another sensors maybe. Even the documentation says 10 cm. In my case, my distances errors varies with real distances, 5.010 meters i have a error like 5cm and 3 cm of STD on LOS (calibrated).
I use the Mdek1001 kit to work my firmware.
Sorry cause i didnt help a lot
The lack of details prevents me from being absolute about what can be done, but you can achieve 0.5 cm precision with a system based on a DW1000. It isn’t easy to do this, and none of the freely available code will manage it, either.
The key to achieving this is two things: lots of anchors and excess beacon rate to enable filtering. Having more anchors produces more measurements to average out for any single beacon hit, as long as the location engine can take in more than a set number into the solution. Then, if you can average together multiple beacon results, you get lower and lower position noise. In effect, using more than one beacon is also an anchor count multiplier.
I can give a real example, though much larger than your service volume. We install our UWB system in a basketball arena with 55 anchors in the ceiling. That’s a lot of anchors. The system was operating in TDoA mode (what we call “MultiTime”). We placed a tag on the court center and measured the position noise at about 2 mm standard deviation in the X and Y planes, and about 4 mm in the Z plane. Generally more than 99% of position outputs will fall in the range of +/- 3 std devs, so that gives you a sense of the accuracy.
The doesn’t quite meet your requirements, but this was without averaging. The std dev will decrease by 1/sqrt(N) for N samples averaged, so if you averaged 10 beacons, then you are looking at 1/3 the position noise which puts you clearly within reach of your goals.
A possible off the shelf system would be 12 of our DWETH101 anchors (one every meter on the service volume exterior), 1 or more DWTAG100 tags, and our CUWB 3 software. Operated in MT mode, you can configure the tag to run at very high beacon rates, say 160 Hz, then average 16, and you would get a factor of 4 improvement in position noise and an output data rate of 10 Hz. If you can tolerate slower output data rates, say 1 Hz, then you can average 160 locations and get 1/12th the position noise. At the very least, this is a precise system, and any inaccuracy can be calibrated out.
I should note that there are certain issues that you have to be careful about when trying for sub centimeter accuracy. If the tag changes orientation, rotates in space, antenna group delay and pattern can make more than 0.5 cm difference. Also, you have to find out where the electrical “center” of the antenna is which isn’t always which would be intuitive. Further, signal strength can cause changes in results. To achieve a healthy 0.5 cm working accuracy, you may have to calibrate your working volume be developing a transform from the precisely measured locations to the true accurate location.
There is a way to use phase measurements for VERY precise locations, but this works only if you can run a wire from the moving tag to clock source that also supplies the anchors. This is usually not possible, but your service volume is so small that perhaps your application would allow that. Again, lacking details, I can’t say what works for you or not.
Mike Ciholas, President, Ciholas, Inc
3700 Bell Road, Newburgh, IN 47630 USA email@example.com
+1 812 962 9408