I would like to utilize the anchors in a unique way and I’m not sure how this will impact the accuracy.
Essentially, I would like to have 3 anchors within a few inches of each other, and use them to track a tag - instead of having the anchors around the perimeter of the tags. I know that when used around the perimeter, they will give an accuracy of ~10cm, however does anyone know what I can expect with the orientation mentioned above?
Any help is appreciated
Hi ,
I take it you have EVBs (EVK/TREK).
We claim accuracy of +/- 10cm. But some customers would achieve better then that eg by applying a different two way ranging algorithm (see APS013 and IC user manual Appendis 12 on page 225).
As you could have read in the EVK user manual and for example the power management application notes (APS023) , one can read that the PHR limit is calibrated to just below the -41.3 dBm / MHz limit, so within regulatory maximum limit.
So if you’re requirement is to have static nodes 10-20cm apart there is no need to have that level of power and so you may well decrease the TX power to meet you’re requirements. See APS023 series and IC user manual (eg section 7.2.31 on page 105)
I suggest you try your set up by experimenting with some different configuration. Tune /calibrate the boards (TX pwr & Antenna), avoid reflections (see APS006) and develop SW to filter out reflections if that would apply.
/L
The DW1000 or DWM1000 accuracy is <10 cm for a 1-1 range. In products like TREK location accuracy is given by R95. In your set up, each individual range should give you good accuracy… however you will get varying 2-D/3-D accuracy depending on the geometry of your set up and algorithms used to solve/get tag’s position. Having anchors far apart will give better location accuracy as the intersection volume of each of the spheres (defined by the tag-anchor range) is minimised.
I am interested in understanding if and how the position accuracy of the tag will improve with the number of anchors used. If I double the number of anchors, does that halve the error in position? Any help on this would be much appreciated!
I have tested this with 4 anchors and 6 anchors (MDEK1001) bud haven’t seen any difference…
lol…
guys, you need to know how the position algorithm is working, and that will tell you if increasing anchors will do anything to the location estimation accuracy.
In TREK, the trilateration algorithm uses 3 (tag to anchor) ranges to try and estimate tag’s position.
In PANS SW (DWM1001/MDEK), algorithm uses 4 anchors (actually gets 4 ranges, and uses sets of 3 at a time, then gets average) to estimate tag’s position.
Main point is: good data in -> good data out
By good data I mean, correct tag-anchor ranges (low error) and correct anchor positioning, good anchor geometry (e.g. having all anchors in one line will obviously not work as well as if you had them spaced out in a grid)
I am not sure why you thought that adding anchors would improve the performance of the system. Please try and understand how the algorithms work and that will tell you what could be done to help them get better location estimates. Note TREK algorithm is fully open/available in the GUI source code. The PANS one is not.
I bought 4 DWM1001-DEV boards - so I only have 3 anchors.
Would 3 anchors mean I’ll have a decreased accuracy or does it not work at all?
No, as the discussion above indicates the TREK system needs 3 anchors and will only ever use 3 anchors to calculate a location so adding more is of minimal benefit assuming you are always in range with line of sight.
The PANS system will take 4 ranges in and then use 3 of these to calculate a location and so give a slight improvement in performance.
Other position algorithms can give greater accuracy from larger numbers of anchors but you would need to implement that yourself.
Yes but I guess what I meant to ask was ‘what if I use 3 anchors for the PAN system instead 4? How would that affect the location determination?’
With 3 good line of sight signals the answer is not much.
The advantage in having 4 measurements and a position algorithm that only uses 3 of them is that you can try to detect bad range measurements and avoid using them. If all of your measurements are good this doesn’t help you but in real world situations where there will be signal blockages or occasional lost/corrupt packets then it can give a significant improvement in reliability and accuracy.