The reason of crash and reason of drifting from mission are different, based on the flight data collected.
Reason of drift from the planned mission:
As we can see, the UAV drifts far off during flight to the WP5. WP5 is a loiter_turn point. It should rotate at a radius of 3m specified in CIRCLE_RADIUS.
There is no indication of mechanical failure which causes any deviation from the navigation route. Below shows that for the entire auto-flight, the XY position is exactly what the controller demanded. Hence it is most probably a sotware glitch.
According to this,this and this, the LOITER_TURNS function is broken in version 3.1.
"This change will go out with the first AC3.2 release candidate and I'll add it to the list of to-be-tested features."
"NB: The units are meters in AC 3.1 and will be centimeters in AC 3.2. ***Mission Planner will report the units as cm for both versions of the code."
Remark: There is a strange phenomenon that the UAV doesn't actually hit within the radius (4 meter as in the particular flight) of the WPs. The distance of way-point switching can be as far as 10+ meters. The reason should be found in source code analysis (to-do)
WPNAV_ACCEL default 100
Reason of crash
Mechanical Failure:
- We suspect that the front-left propeller had come loose during the latter part of the flight. In normal flight, the propeller is rotating clockwise, which creates a CCW torque to the UAV. However, when the propeller comes loose for some reason, there is not enough CCW torque for the craft. Hence the craft keep yawing CW. The
Evidence:
- The propeller and nut is not on the motor after crash.
- The propeller was found 3 meters away from the site of crash.
- The screws screwing the motor axis became loose, indicating abnormal vibration occurred.
- The motor 3 and 4 rotate much faster than motor 1 and 2, indicating that the craft tend to correct from a continuous CW yawing. However, due to the loosen front-right (No. 3) propeller, the CCW torque is just not enough, and adversely affect the UAV's attitude.
graph showing that the UAV tending to yaw CCW continuously (but failed)
graph showing that there is no yaw stick being applied by pilot, when the UAV keep rotating CW
Further info - battery voltage dropped drastically, indicating that a sudden loss of lift, which in turn requires more power to maintain the UAV in the air.
Remarks: Initially we were suspecting that the auto-tune was the cause of the crash. However, it became less convincing after we realized that the online auto-tune result was in-line with ours. We were not getting abnormal tuning values.
Google Earth Generated FPV for the crashed flight
