[Betatesting] Power tower inspection exercise

[javizqh]

This exercise goal is to detect problems in the tower using deep learning.

Share here any issues, not solutions, related to this exercises. Do not comment anything else other than that.

This issue is not meant to be assigned to a single person, as its purpose is that multiple people test it.

[Arkadeep455]

This top window on right side (Web Gui) is always like this for each and every exercise. I followed every installation step properly. Also, the right middle window 'Gazebo Sim' is too laggy, maybe because I am using my CPU to render it but my it's i7 11th gen and last but not the list once I switched from object detection exercise my webcam is not turning off.

I have attatched a screenshot regarding the issue. Rest everything is working fine.

[javizqh]

Regarding the webgui: you need to execute code and send the images to it using the provided functions

[Rudhratej09]

Hi @javizqh ,

While testing the Power Tower Inspection exercise, I observed that when using HAL.set_cmd_mix(vx, 0, z, 0) with a constant height (e.g., z = 4.0) and continuous forward motion, the drone gradually loses altitude over time, even though the height command remains unchanged. The command is being sent continuously in a loop.

Demo video:
https://youtu.be/qVk6S1xWOyI

(Please notice the Z-axis reading in the terminal.)

I’m unsure whether this behavior is expected due to controller tuning, or if mixed control is not intended for sustained forward inspection passes.

I’ve also been exploring the exercise architecture further, as I’m particularly interested in the proposed deep learning extension for Power Tower Inspection (Project no 3), especially regarding how detection outputs might be structured beyond console messages.

Looking forward to continuing testing.

[JitinSaxenaa]

Hi @jmplaza @javizqh

While testing the Power Tower Inspection exercise, I noticed an issue related to the simulation:

1 - When the exercise is paused, the drone continues moving instead of freezing in place. It keeps drifting until it eventually crashes into the ground, and even after touching the ground it continues sliding.

2 - Additionally, when the exercise is reset during such a state, the drone does not return to the initial takeoff position. Instead, it falls at the same position where it was at the time of reset.

3 - Also the altitude keeps on decreasing with time even though the commanded height remains constant.

2026-03-03.01-14-36.mp4