Getting inaccurate distances with multiple anchors

[muhammadalimuzaffar]

Hello everyone I am making a project of human following robot and for that I am using dwm1000 UWB modules. Recently, I am trying to do localization. My experimental setup contains 3 anchors and one tag. I calculated antenna delay by following the instructions given in this library. After calculating antenna delay of each anchor I was getting each distance from all anchors to the tag individually very accurate.

After that I tested "ESP32_UWB_tag2D_3A" code present in the library for localization but after uploading this code in tag I came to knew that I was getting x and y coordinates of tag extremely inaccurate for example:

Actual Distance from tag:
A1 = 1.30
A2 = 1.35
A3 = 1.10

Distance which I was getting on tag serial monitor:
A1 = 1.01 approx
A2 = 0.90 - 1.10 approx
A3 = 1.15 approx

Actually I observe that I get the distance accurate only when one anchor is on but when multiple anchors are on then I get all distances inaccurate. It seems to be signal collision when multiple anchors are on and this might be the reason for inaccurate x and y coordinates of tag as all distances becomes extremely inaccurate when multiple anchors are on. So can anyone help me out to get all distances accurate and x and y coordinates accurate even multiple anchors are on so I can go ahead with localization and following. An response will be appreciated. Thanks.

[jremington]

Those ARE NOT "extremely inaccurate" measurements.

The stated accuracy of DW1000 distance measurements is +/- 10 cm, which is one standard deviation. Statistically speaking, you can expect to routinely see deviations larger than 10 cm.

For higher accuracy, average several measurements. The rms error of the averaged value will be reduced by sqrt(N) for N averaged measurements.

[muhammadalimuzaffar]

Thanks for your response. You said that by averaging some set of distance values of each anchor I can get some sort of accurate distances. Actually in the tag code I already implemented moving average filter. Now as tag will get all three distances it moves those distances into the moving average filter. Before moving average filter I get all 3 distances very fluctuation but after moving average filter I get smooth distance values bout values are still inaccurate with an inaccuracy of +/- 40 - 50cm which is much and this issue is only with multiple anchors. For one anchor I get accurate distance with +/- 10cm inaccuracy.
Now as I am making human following robot when I do trilateration then I get x and y coordinate of tag extremely inaccurate. For example if tag is in 1st quadrant (mean x and y both +ve) then it will show that tag is in 3rd quadrant or sometime in any other quadrant. So what should I do now?

[muhammadalimuzaffar]

@jremington I observed one more thing I thought there should be some signal collision when multiple anchors are on because may be all anchors are responding at the same time to the tag but inside you library I just saw DW1000Ranging.cpp and DW1000Ranging.h file. Inside DW1000Ranging.h file:
#define DEFAULT_REPLY_DELAY_TIME 7000
#define DEFAULT_TIMER_DELAY 80
Here you can see that the reply time for anchors are set to be 7000us or 7ms.

What does that mean that first anchor will respond after 7ms, second anchor will respond after 21ms and third anchor will respond after 35ms and this logic is implemented inside DW1000Ranging.cpp file under transmitPoll function in below line:
for(uint8_t i = 0; i < _networkDevicesNumber; i++) {
//each devices have a different reply delay time.
_networkDevices[i].setReplyTime((2*i+1)*DEFAULT_REPLY_DELAY_TIME); (This line)

Now here I got confused that each anchor is already responding on different time one after 7ms, second 21ms and third at 35ms so how can there be signal collision. So why I am getting these inaccurate distance only when multiple anchors are on?