DWM1001 - Data accuracy problem

Wireless Connectivity Ultra-Wideband
1

The anchors are all powered with USB 1A Wall chargers.

The environment has very few objects that can cause physical interference except structural beams at each 20’ and these beams are less than 10 inches wide. This should not be a problem.

I am using the Dev Kit and the data is collected with a Tag connected to a laptop. I was walking on that area to get the data in Tera Term.

The red dots are position that had 0 confidence in data (x, y, z, Confidence), the orange dots had 50 and the green ones had 51 and above. The purple dots are the anchor positions. The units are meters.

I am wondering why the data is not accurate in the red square areas in the pdf?

Data accuracy.pdf (66.5 KB)

Thanks!

2

Did you have line of sight from the tag to the anchors at all times or where the structural beams in the direct path?

3

Hi JmLamb,

Can you confirm you are using dwm1001 ?

Are there some cases for which the tag is ranging to 3 or 4 anchors but cannot calculate a position ? If yes I would suspect the anchors coordinates are not accurate enough and not representative of the real network topology

Regards
Yves

4

There was not always a direct path as the was physical interference by moment.

But if this was caused because of the physical interference of the structural beams, I would have had small localized areas with lack of data. (The beams are 10 in wide and the map units ar meters)

Here the lack of accurary is spread out, it does not seem to represent straight lines of signt being blocked. With tag position accuracy being 10 cm, we should see the line of sights being blocked in the file…!

5

Hi Yves, yes there are a lot. And yes I am using DWM1001.

I have made another post about that with an excel file with data.

The anchor coordinates were measured manually as the auto-positionning function was not working.

I am pretty sure my anchor coordinates are within +/- 1 meter precision if not +/-1 foot.

This means I should have less than 1/82 meter = 1,2% innaccuracy on my anchor coordinates compared to network width.

Could this inprecision cause this really?

I am pretty sure the coordinates were within the foot precision as everything was measured in inches. So the innaccuracy should be less than that…!

6

Hi Jm,

I would say 1 meter precision on anchor coordinates is low and definitely the trilateration would perform poorly.

Try to be sure about the anchors coordinates maybe 10cm on both x/y axis.

We usually use laser pointer to measure distance accurately.

Thanks
Regards
Yves

7

Hi Yves,

Thanks for the information. I will buy a laser pointer to get more accurate data (I’ll aim at 2-3 cm precision on both axis).

Is there anything else I could do to improve the accuracy of the readings?

(I know these are basic questions regarding the system but I do not find information in the pdfs that come with the dev kit package and the dwm1001 pdfs.)

Thanks Yves,

Greatly appreciated!

-Jm

8

Hi Jm,

I would expect the trilateration to behave much better once your coordinates are accurate.

If not come back and we can think at other solution, but for now that’s the main thing to look at.

Yves

9

Hi Yves,

Thank you for the support an quick response, greatly appreciated.

Have a nice one!

-Jm

10

Hi JmLamb,

I was also facing same problem earlier then I figured out that it was false measurements of anchor placements. On top of that you need to consider the variance matrix for finalizing the position of the anchors based on the number of anchors you are using and the algorithm you are using for position calculation.

Regards,
Shivek

11

Its also important to have the tag you’re walking around with be vertical, and that it not be occluded by things like a laptop screen.

I designed and 3D printed a tag holder that holds a raspberry pi (with a small touch screen), a USB charger battery, and a tag.

Google Photos

Google Photos

We mount the assembly either on top of a bicycle helmet, or attached to a dog’s harness, so we can get clear and consistent readings.

I also use a laser tape measure to get my anchors positioned accurate to a few cm.

  • Jon
1 Like
13

That’s really cute!

I like it!

14

When you talk about the variance matrix, are you refering to adding a Kalman Filter and applying it on the collected data?

Thanks.

-Jm

15

No, we can simulate the variance matrix for the fixed anchor positions and see where in the test area we can have the large variance in measurements. In my case if I am keeping mine 4 anchors on the edges of a square it will have large variance in the vicinity, but not when we keep them in different planes due to fact that I am using only 4 anchors as well as the algorithm is more efficient when the vectors are not in a plane.