Accuracy is unacceptable

dwm1001
#1

I purchased MDEK1001 and set it up in my bedroom:
5 Anchors and 1 Tag as shown in the following sho (the 5th Anchor is too many but I was hoping that it will improve the accuracy).


The Tag is Blue and Initial Anchor is Red and other Anchors are Green.

I did everything “by the book/manual” and the distances are not huge (meters) but I am getting accuracy of 10 cm only sometimes and only along one axes (X-axes) most of the time and on the rest axes I get about 30 cm. Those are difference between the numbers I see in the DRTLS Android app and those I actually measure with the tape.
The initiator node is in the left-bottom corner.
What is the reason for these huge errors? What am I doing wrong? Is this due to anchor clock sync? If yes how do I correct it?

Another issue that I am having is that Tag sometimes disappears if I stop moving it.
Any help is appreciated.

0 Likes

#2

Hi securigy,

How accurate do you think your anchors position are ? Have you used a laser distance measurer ? That’s what I usually use to have the best accuracy.

How high are you anchors and is the tag always in line of sight with at least 3 anchors ? According to your drawing that’s the case but i’m wondering if there is any obstacle in the room.

In order to understand what happens when the tag disappear, I would recommend to connect to the tag over serial, enter the uart shell mode and check the output of the les command. There is a possibility that the tag range to 4 anchors but cannot calculate a position estimate if the anchor’s coordinates are not accurate enough. In that case, the tag will “disappear” from the android table.

Thanks
Yves

0 Likes

#3

The laser measurement devices claim high accuracy but affordable ones are not very accurate at all. I measured very carefully with tape so my errors are +/- 1 cm max. mainly due to location of antenna inside plastic cover…

The Tag is always in LOS. But even if it wasn’t - I got impression from all I read that UWB goes through walls, obstacles, etc, except metal obstacles…
I have no tall obstacles in the room - there is a bed with mattress along the bottom wall 30" tall.
All anchors are at 6 feet above the floor (1.83 m) and the Taq is on the bed at 33" (0.84 m)

The Tag does not disappear from the table in the Android app - only from the Grid…

Accuracy that I am getting:
delaX = 0.14-0.23 m
deltaY = 0.29-0.35 m
deltaZ = 0.05-0.07 m

0 Likes

#4

Hi securigy,

Indeed it sounds like the anchors coordinates are accurate enough.

Have you looked at the les command output from a tag ?

Yves

0 Likes

#5

I did not - I needed a second laptop to do it …
But I made some changes:

  • I started measuring Tag at the positions that eliminate shallow angles - that is locating the Tag away from the walls.
    I also changed the orientation of Anchors on the walls in a way that the vector perpendicular to the antenna plane are parallel for all Anchors.

To my amazement I started getting the following accuracy:
X 0.13 - 0.15 m
Y 0.06 - 0.07 m
Z 0.06 - 0.16 m

I still do not know if this this because of Anchor orientation but I strongly suspect shallow angles as a reason the previously high errors.
I also read variety of docs in the Application Notes and it seems that one of the docs claiming that with the right calibration at the production you can achieve 4.5 cm accuracy. So my questions are:

  1. What specific calibrations were already done in the units that ship with MDK1001?
  2. What specific calibrations were already done to shipping units DWM1000 and DWN1001?
  3. What calibrations are required to DWM1000 and DWM1001 in order to achieve accuracy of 4.5 cm or around it?
  4. Will accuracy degrade with the range (10m spaced anchors?) (20m spaced anchors?) - my current anchor spacing is 3-4 m
  5. What accuracy should I expect if I place 4 Anchors spaced 1 ft (as in square) n the moving object and have static Tag outside of it at the distance of few meters?
0 Likes

#6

Hi securigy,

the reason of inaccuracy you have measured is due to the antenna radiation pattern.

Physically it is impossible to create a perfect omni-directional flat antenna, in this case PCB antenna. So depends on the orientation the distance measurement might differs resulting in errors in position estimation.

  1. The antenna calibration has been done between two nodes with antenna pointing face-face. If you measure the distance between Anchor-Tag you should get the accuracy described in the Application Notes.
    I think you are mixing 2 accuracy terms together. Distance accuracy should be as mentioned in the Application Notes. Position accuracy is a combination of errors in: all distance measurements, anchor positions, location engine.

  2. DWM1001 - all modules have been calibrated as mentioned above. DWM1000 - I don’t know.

  3. Again, if you talk about distance measurement then you need to define the antenna orientation. Or you need to use another antenna, see this post: Antenna orientation Calibration

  4. No, it should not.

  5. The accuracy of distance or position? See my response above.

Cheers,
TDK

0 Likes

#7

I am not sure at all if I care about distance accuracy since I do not paln at this point to do any calculations on my own. I do care of position accuracy - I’d like to bring the object as close as possible to particular coordinate in the room. So the error in X, Y, Z direction of the object position vs the given room coordinate matters to me.
So now you can relate to my questions better?

0 Likes

#8

Hi securigy,

regarding your question 5: if I understand correctly then you have 4 anchors placed at the same height in square shape of 1ft and you want to track tags which are outside that area.

A such installation will not yield the best results because of the geometry. Look for GDOP on internet. The best is when the tracking area is inside the area covered by the anchors.

Cheers,
TDK

0 Likes