LinkTrack UWB on PX4: accuracy outside the anchor polygon and adding anchors

PX4 Autopilot
1

Hello,

Following my general question about indoor positioning for PX4, I’d like to ask specifically about LinkTrack’s anchor layout and positioning accuracy.

Setup and documented constraints

We are evaluating Nooploop LinkTrack P-AS → NMEA-0183 over UART → PX4 for an indoor blimp. We need official firmware and supported configuration/calibration tools only—no custom code or external coordinate converter.

According to the LinkTrack User Manual V2.3 (Chinese):

  • The documented NMEA setup without coordinate transformation requires A0 → A3 to align with true East (p. 73).
  • The six-anchor automatic-calibration layout places A1, A2, A4 and A5 north of that baseline, in clockwise order A0–A4–A1–A2–A5–A3 (pp. 65, 70).

Room layout and concern

Our room is empty and rectangular, but its walls are not aligned with true East/North. Anchors must be wall-mounted.

image
image1448×1086 197 KB

In the attached sketch, following the prescribed layout leaves the orange region inside the room but outside the anchor polygon. This is a geometric example, not measured signal loss or confirmed unusable space. We want reliable positioning throughout the usable room, including that region.

Questions

  1. What accuracy have you measured outside the anchor polygon? Horizontal/vertical errors, dropouts, anchor spacing, mounting heights and distance beyond the boundary would be helpful.

  2. Can a supported configuration improve full-room accuracy? Could additional anchors—including more than six or anchors south of the A0–A3 baseline—or official manual calibration help while retaining true-East alignment and direct NMEA output?

We can purchase additional anchors as needed. Please share tested layouts, positioning models, firmware versions and calibration procedures, noting whether the setup worked with an unmodified PX4 release.

Thank you.

2

One distinction that may be useful here is radio coverage vs positioning geometry.

Being able to communicate with the anchors outside the anchor polygon does not necessarily mean that the position solution will retain the same accuracy there.

I don’t have measured LinkTrack P-AS data for this exact configuration, so I wouldn’t attach an accuracy number to the orange region without testing it.

If you test the setup, I would suggest recording a trajectory through known reference points: inside the anchor polygon → boundary → progressively outside, at several representative heights.

With the corresponding PX4 ULog, it should then be possible to compare position error, solution availability and estimator innovations against distance from the anchor polygon.

Regarding additional anchors, I would also confirm whether the official LinkTrack firmware actually incorporates more than the documented six anchors into its positioning solution. Adding anchors physically would only help if that positioning mode uses them.

If you collect such a dataset, I’d be interested in helping analyze where the positioning or estimator behaviour first begins to diverge.