Oh, well, yesterday we were doing some routine checks of the auto mode and the quadcopter just went unstable, flipped over and crashed. This happened in calm weather, at the right altitude, apparently (from the logs) right after reaching the first waypoint and trying to go for the second way point. If you look at the log, the pitch axis seems to have gone unstable, and I am not sure why. It just pitched one way, pretty fast and then it flipped the other way. Had no time to react to move it out of auto, and even if I would have, it's unlikely that I could have done anything to save it (I'm not a super-dooper pilot, and it was quite far).
Initially I thought that perhaps the autopilot did not keep up with the control loops, but it's performance doesn't seem to be low - all loops are "short", and it doesn't seem to have performance issues. I tried to check "everything", but I'm sure I'm missing something. Things I checked: GPS satellites, GPS latitude, GPS longitude, GPS altitude, Barometric altitude, VBatt (appears very low, but it's not, it's a misread), Vcc (also appears very low, but it's also not), I also checked navigation distances to waypoints. All the PID parameters are the default ones and they were doing fine before (have at least 10 minutes or so of flight in auto-mode before without issues). As per instructions I did not enable IMU and MOTORS logging, although by looking at the TLOG it appears that at the moment of the crash one of the four motors was being pushed hard (1800 vs 1500 for the other three) - I'm not sure if it was the "good" motor (that would have saved it), or the "bad" motor (that flipped it).
I am at a loss. If anybody can spare the time to look through the logs, or have any idea on what's going on, please let me know. I'm attaching the datalogs, tlog and picture with the pitch in the last 2 seconds. What is shown is "desired pitch" (from navigation), then "pitch in" from the control loop (follows the desired pitch except for a weird value), then "pitch", probably from the IMU.
Replies
BTW, the pitch graph is "textbook" for an unstable system (that's why I suspected slow autopilot - delays in the control loop can look like that, as are high gains). The problem is that the autopilot seemed to not have been delayed and the gain was at the default value. One more thing: there is an "ERR" entry in the log - it's the battery failsafe; however, the parameter battery failsafe is disabled, so this failsafe was not acted upon (the ERR is many minutes before the crash).