Help Needed: Inconsistency with HTC VIVE Tracker 3.0 Positioning Accuracy

[Matt_2024]

Hi everyone,

I'm working on a project using two HTC VIVE Tracker 3.0 units along with two HTC VIVE Base Station 2.0s, all mounted on tripods and stationary. The trackers are connected via USB-C cables (I'm not using the wireless dongles), and I've disabled USB power management. I’m using Unreal Engine to ingest the tracker data, and then I dump it with custom software.

According to the paper "HTC Vive: Analysis and Accuracy Improvement - Borges, Symington, et. al." the expected standard deviation for position should be about 0.417 mm and 0.00586 degrees for orientation. Unfortunately, I'm not achieving this level of accuracy.

During my tests, every 100 seconds I induce full occlusion by covering both trackers for 10 seconds, which presumably forces reliance on their IMUs. However, when the trackers regain line of sight with the base stations, their position data does not return to the previous positions, but instead shows deviations of about [-1 cm, +1 cm] from a reference position.

In the figure below, I'm showing the (x, y, z) position of each tracker. When you see the step functions, that is when the trackers are occluded.

 

If we zoom in on the X position for each of the trackers, we can see how the positions do not settle back to their original locations. The tracker on the left is not as bad as the one on the right.

To see this difference, consider taking the first position as a reference, and then measuring the deviation from it. As an example, for the figure above, the first window is ~2183.8 mm, and the second window is ~2179.1 mm, so the delta would be -4.7 mm which is shown at window index 1.0 below (in red).

Has anyone here dealt with similar issues? Any tips or insights on potential calibration errors or other factors that might be affecting the accuracy?

Thanks in advance for any help!

[C.T.]

some suggestion I have for now is that to delete everything under this folder
C:\Program Files (x86)\Steam\config\lighthouse\

then try to run the test with only 1 base station. see if it makes any difference. thanks