ArduPilot EU Dev Call 2025-02-19

Attendees (max): 9

UTC0702

Andy: The existing polygon function wasn’t accurate enough. I think it’s because it uses Cartesian and not lat/lon.

  • Of course, a new implementation is costing more flash.

Andrew: It used to be lat/lon.

  • But we now convert the coordinates to Vector2f (e.g. read_polygon_from_storage).
  • Hence fences become less accurate, the further out they go.
  • Did you mean to support all fence types?

A: Yes
Andrew: We do have the Vector2f in memory, you can call it directly and save loading/processing anew.
A: What kind of error are we looking at if we implement warning in lat/lon and breach in float?
Andrew: It’s at the worst 1m for a planet-wide fence.
A: Alright, I’ll try with the Vector2fs then.
Michelle: We could be storing the origin along the Vector2fs, and not use lat/lon, if that’s a problem.
Andrew: We do like using lat/lon as it’s the primarly format for input and display.
Peter: At this point we’re +1900B for this PR.

Peter: We should be able to assign each fence with a breach action.
George: Yes, and also it’s more intuitive/transparent to have multiple fences for each warn/rtl/etc action. And they should be created in the GCS side: The user could draw multiple fences, or click a button to automatically copy and offset.


UTC0728

Andrew: LGTM. Randy should approve.


Andrew: Nice!

  • Can you check what would happen if there’s a (0,0) waypoint right after the DLS waypoint?

UTC0735

A: RV_MASK is wrong indeed, but to change it is some effort.

  • You’d expect bit 0 to map to servo 1, but it doesn’t in my experience.

Andrew: I don’t think there’s a way to un-reverse a DShot ESC without rebooting it. That’s a gotcha.

  • Is there an example of bad behaviour? Does the flight controller have an IOMCU?
  • By the way, we should rename all the _BLH_ names. Perhaps DSHOT_*?. SERVO_* is also inaccurate, but also another problem.

UTC0745

Discussion on an issue found on some M10 GNSS RXs where they would get stuck in their configuration.


UTC0801

Andrew: It’s fine to remove, as long as we accept the argument and do nothing without it.


Continuing live-debugging of the problem from UTC0745.