PX4 Dev Call: Jul 22, 2026 (Team sync, and Community Q&A)

system · July 20, 2026, 10:58am

PX4 Sync / Q&A - 2026-07-22T15:00:00Z

This week: 59 PRs merged | 283 awaiting review | 0 new bugs

Announcements

Proposal: AI Policy for contributions

TL;DR; link here

github.com/PX4/PX4-Autopilot

docs(contribute): add AI coding assistant policy (#27995)

main ← mrpollo/ai-coding-assistants

opened 09:46PM - 20 Jul 26 UTC

mrpollo

+122 -8

## Summary Adds a project policy for AI-assisted contributions, modeled on the …Linux kernel's [coding-assistants](https://github.com/torvalds/linux/blob/master/Documentation/process/coding-assistants.rst) document, and wires it into the contributor docs and the in-repo agent tooling. ## Problem AI-assisted contributions are arriving in volume and PX4 has no stated rules for them. Contributors don't know what is expected, and maintainers have no reference to point to when authorship, licensing, testing claims, or disclosure are unclear. The project position is that AI tools are welcome, but the human in the loop is the author and the responsible entity, and that needs to be written down. ## Solution New docs page [`contribute/ai_assistants.md`](https://docs.px4.io/main/en/contribute/ai_assistants) defining the policy: - The human submitter is the author, accountable for every line; an AI tool never appears as author, co-author, or in a `Signed-off-by` tag. Sign-off is the author's certification of changes they have personally reviewed. - Licensing follows the BSD 3-clause requirements and the [Linux Foundation Generative AI Policy](https://www.linuxfoundation.org/legal/generative-ai). - Disclosure is required: commits with AI-generated or AI-assisted content carry an `Assisted-by: NAME:MODEL` trailer (kernel format). - Maintainer expectations: test claims must be real, no unsolicited bulk AI changes, issues and security reports human-verified, AI-assisted review clearly labeled. Non-compliant contributions may be closed without detailed review. The policy is summarized in CONTRIBUTING.md and enforced by default in the in-repo agent tooling: new `AGENTS.md` with the agent-facing short form (imported by `CLAUDE.md`), and the commit/pr/review skills updated to add the disclosure trailer, protect the sign-off rule, and never fabricate test claims. This commit was written with AI assistance and discloses it per the policy it introduces.

Absute Detected

https://github.com/px4/px4-autopilot/pulls?q=is%3Apr+author%3ABartok9+

https://github.com/mavlink/qgroundcontrol/pulls/Bartok9

Events

Dev Call Items

None

Release

Updates:

Needs more flight testing, we have had bad weather in Chicago, we need more time.

Issues:

Decisions:

There has been a lot of fixes going into main and backports need to happen before the next beta. We will update everyone as soon as it happens.

PRs Merged This Week

59 PRs merged this week →

PRs Awaiting Review

283 PRs awaiting review →

New Bug Reports

No new bugs

Other New Issues

8 new issues →

Q&A

Action Items

JBotwina · July 20, 2026, 9:14pm

This Gimbal fix can be reviewed.

github.com/PX4/PX4-Autopilot

fix(simulation): fix Gazebo gimbal ROI yaw frame and tracking (#27994)

main ← JBotwina:gimbal-roi-yaw-frame-fix-26178

opened 08:36PM - 20 Jul 26 UTC

JBotwina

+21 -3

## Summary fixes #26178 Convert Earth-frame gimbal attitude setpoints into… vehicle-relative Gazebo joint targets, and recompute geographic ROI bearings as the vehicle moves. ## Problem Gazebo gimbal joints consume vehicle-relative angles, but PX4 could send an Earth-frame ROI quaternion directly to those joints, applying the vehicle heading twice. In addition, the MAVLink v2 output recalculated a geographic ROI bearing only when a new ROI command arrived, so the gimbal setpoint froze while the vehicle translated. ## Solution The Gazebo bridge now respects the yaw-frame flags (including the legacy yaw-lock flag) and rotates Earth-frame attitudes into the vehicle frame with the inverse vehicle attitude before commanding the joints. The MAVLink v2 output now refreshes position-based setpoints on every update cycle, matching the v1 output. Validated in SITL with `gz_x500_gimbal` and `MIS_MNT_YAW_CTL=1`: vehicle yaw and Earth-frame gimbal yaw remain independent, and the gimbal keeps tracking a fixed ROI as the vehicle moves. <img width="584" height="340" alt="image" src="https://github.com/user-attachments/assets/ea8643de-026d-45c6-86a7-eb8ccd0a43c9" /> <img width="499" height="341" alt="image" src="https://github.com/user-attachments/assets/8b14a4b2-e961-4a87-8c97-2a444e23a194" /> <img width="513" height="373" alt="image" src="https://github.com/user-attachments/assets/648d370c-ff74-4e47-83fe-ba148e306541" />

JBotwina · July 21, 2026, 6:33pm

This issue is was present in v1.16 but is fixed in v1.17 (commit reference in comment)

github.com/PX4/PX4-Autopilot

[Bug] auto entry into HOLD mode during auto-takeoff testing.

opened 08:21PM - 02 Jun 26 UTC

acsousa

kind:bug scope:control

### Describe the bug ### Px4 entered into HOLD mode due to a bug in the RTL tim…e failsafe on fixed-wing platform. v1.16.1 **Description:** Px4 entered into HOLD (loiter) mode at T+129s into log (this was commanded) and tried to ascend up to the set loiter altitude. nav_state changes from 0 (manual) to 4 (HOLD). Params set to "do nothing" but auto entry into HOLD mode occurred, see relevant params below: Battery failsafe parameters: - `BAT_LOW_THR`: 15% - `BAT_CRIT_THR`: 5% - `BAT_EMERGEN_THR`: 3% - `COM_LOW_BAT_ACT`: 0 (Warning Only) RTL failsafe parameters: - `COM_FLTT_LOW_ACT`: 0 (no action) - `COM_FAIL_ACT_T`: 5 (seconds) - `RTL_TIME_FACTOR`: 1.1 - `RTL_TIME_MARGIN`: 100 (seconds) At T+129 the Battery Low Remaining Failsafe is triggered because the `RTL_time_estimate` (50.9s) drops below `safe_time_estimate` (156s). `COM_FLTT_LOW_ACT` is set to 0 (no action) and `COM_FAIL_ACT_T` is set to a delay of 5 seconds, so nothing _should_ happen. But the delay triggers an IF statement in **failsafe/framework.cpp**, setting `_current_delay = _current_start_delay`. Further below, because `_current_delay > 0`, mode gets hard set to HOLD mode at line 484. So, even through aircraft was on the ground for auto-takeoff testing, the mode changed to HOLD mode and the aircraft attempted to takeoff (to achieve HOLD altitude setpoint). **Temporary resolution:** Set `COM_FLTT_LOW_ACT` to 1 (warning). Documentation is not clear on this (it allows value of 0 (None)). **Steps to reproduce on benchtop:** 1. Have GCS running and connected to vehicle wirelessly Have vehicle fed by Benchtop Power supply 2. Start at ~22.0V (full battery) 3. Arm 4. Go to Manual Mode 5. Reduce Voltage until Battery Warning in announced (~21.0V) 6. Blow into Pitot (Nothing will happen) 7. Drop voltage to ~10-9% (~20.3V) 8. Blow into Pitot Tube 9. Aircraft will go into "Hold" until you stop blowing into pitot tube 10. Able to repeat as long as air is blown into pitot (Px4 thinks there is flight) ### Flight Log / Additional Information **Log Link:** https://review.px4.io/plot_app?log=7906f58a-397e-42a0-9339-d91f9df36927 **Firmware Version:** v1.16.1 (RC) ([94cb2012](https://github.com/PX4/Firmware/commit/94cb2012792b2ae89f0b147cfee53ee31ae550be)) **Hardware:** PX4_FMU_V6C

bgptiste · July 22, 2026, 3:04pm

Hello everyone!

This isn’t directly related to PX4, but it concerns a proposed improvement to the MAVLink GNSS integrity message:

RFC - Improve GNSS integrity information reporting by bgptiste · Pull Request #30 · mavlink/rfcs · GitHub.

If anyone is interested in taking a look at the RFC and sharing feedback, I’d really appreciate it. Thanks!

Topic		Replies	Views
PX4 Dev Call: March 02, 2022 PX4 Autopilot	0	656	February 23, 2022
PX4 Dev Call: February 26, 2020 PX4 Autopilot	0	516	February 19, 2020
PX4 Maintainers Call: November 28, 2023 PX4 Autopilot	0	441	November 28, 2023
PX4 Sync / Q&A: Aug 14, 2024 PX4 Autopilot	0	114	August 12, 2024
PX4 Dev Call: Nov 19, 2025 (Team sync, and Community Q&A) PX4 Autopilot	1	74	November 19, 2025

PX4 Dev Call: Jul 22, 2026 (Team sync, and Community Q&A)

PX4 Sync / Q&A - 2026-07-22T15:00:00Z

Announcements

Events

Dev Call Items

Release

PRs Merged This Week

PRs Awaiting Review

New Bug Reports

Other New Issues

Q&A

Action Items

Related topics