2026‐10 - Arrow-air/project-caribou GitHub Wiki

October 6, 2026

Caribou Call — HearHear transcript (Discord); 167 segments failed to transcribe, so parts of the call are missing

Attendees: far1no, errrks.eth, thomasg, alperenag, kbmollysuh

  • Kill Switch Lesson (Quiver, Erick)

    • During an RTL today, Erick thought Quiver was about to hit a branch. He reached for a switch and hit the kill switch instead, which sits next to his arming switch. Quiver dropped from about 7 m.
    • Julius had a similar accident before, dropping from 1–2 m while trying to switch flight mode. He has since unassigned the kill switch from his remote (only 4 buttons) and keeps it in Mission Planner only. Thomas has never bound a kill switch to his remote.
    • Erick will move the kill switch to the other side of the remote. Julius suggested a protective cap, and Erick thinks one could be 3D printed since the switch tops screw off.
    • Julius mentioned ArduPilot's emergency motor stop (throttle down and to the left), which cuts motor output. Nobody had tried it, and Julius needs to read up on exactly what it does.

  • Caribou Status (Julius)

    • The chassis is fixed and can be put back together. It's still partly disassembled. Some Caribou work may happen at the upcoming meetup.
    • The CNC-machined tube connectors are in manufacturing and may arrive next week. They're expensive, but they make life much easier.
    • Julius still has changes to finish on the CF sandwich panels, soon.
    • He created a set of GitHub issues for work needed before November, and assigned KBM the landing gear design (as KBM suggested in the thread). KBM will read the requirements first and pick a tool. Build123d is fine. He's traveling without his SpaceMouse but can do simple shapes.
    • Julius will write the September progress report before Thursday.

  • Crash Report

    • Julius asked if everyone had read the crash report (posted in a channel not everyone may have access to). Alperen had. KBM asked for the flight video, and Julius will post it.
    • Julius's understanding from the analysis: some control loops weren't in sync (one running faster than the other), so the aircraft couldn't handle the delay, leading to oscillation and the crash. It wasn't one particular motor. The control loops were left at the defaults ArduPilot suggests for the prop size (63").
    • Alperen: the long-running roll behavior looks like a bad tune causing a self-feeding overcorrection oscillation. Julius said the parameters had been reviewed with AI help ("Victor" in the transcript), but wasn't sure those changes were in effect for the flight. The analysis ended by adjusting the parameters and running a simulation that showed the oscillation being suppressed. Zeynep reviewed it and had nothing to add.
    • Julius: recheck it, but in the end it has to be flight-tested.

  • Open Items

    • Erick: move the kill switch away from the arming switch; consider a printed cap.
    • Julius: look up ArduPilot's emergency motor stop.
    • Julius: September progress report before Thursday; finish CF sandwich panel changes; post the crash flight video.
    • KBM: landing gear design (GitHub issue).
    • Julius: reassemble Caribou.

  • Open Questions / Risks

    • Crash root cause (control-loop timing vs tune) isn't settled. The fix has only been tested in simulation.