This section is for the predictable failure modes — the issues that you’ll hit, that we hit, that everyone running this stack hits at some point. Each page is structured the same way: symptom → root cause → fix → verification. Where the cause is non-obvious or the diagnosis took us a while, the page also includes the path of misdirection — the wrong hypotheses we chased before finding the real cause — because that’s frequently more useful than the right answer alone.
The flagship page here is SITL “Waiting for heartbeat” — three layered causes. It’s a full debug saga: one symptom (MAVProxy never gets a heartbeat from ArduPilot), three independent root causes that all had to be fixed in sequence. If you’ve never had this happen to you, you will, and reading the page in advance saves a full day of debugging.
The other pages cover the more contained issues: drone flip on takeoff (motor order, prop direction, PID — the classics), Gazebo realtime factor < 1.0 (when sim slows down and what to do about it), ROS 2 DDS discovery (nodes that can’t see each other), and venv ↔ ROS 2 / colcon conflict (Aleks’s note on the Python-environment mess).
A note on philosophy: we don’t believe in hiding our debugging failures. The errors a team makes are usually more pedagogically useful than the solutions, because the errors are where the assumptions in the docs broke against reality. This whole section is structured around that idea.
Contents
Auto-generated from child entries during build (update-indexes.mjs).