Uncontrolled climb in Land Mode/Failure to disarm

ArduCopter Copter 4.5
1

I observed a weird flyaway/crash recently with an Octo running copter 4.5.7
I am having trouble identifying why it

  1. It climbed more than 10m while in land mode
  2. It failed to disarm after almost a minute of skidding around inverted on the floor while gradually disassembling itself

It is a new Cube Blue retrofit on an S1000+ airframe with Jeti and RFD900 for control/GCS telem to a MissionPlanner laptop. The aircraft has flown successfully a few times checking the initial tune (no autotune performed yet). This flight testing was taking place in an indoor GPS-denied environment (thus the AltHold flight mode). Onboard BIN log file.

During this flight, there were some altitude/yaw oscillations observed (likely pilot/RC induced) leading up to the aircraft lightly bumping its landing gear on the ground but remaining in flight.

climb in Land Mode
climb in Land Mode842×886 50.7 KB

Directly after the landing bounce the aircraft was switched to Land Mode via the RC controller.
The aircraft continued the ascent speed it had after the bump despite the mode switching to land, so the mode switch was cycled to AltHold and then back to Land a second time. The aircraft continued a steady climb until it hit the ceiling before dropping to the floor. The drop severely damaged its landing gear and it turned inverted on the floor (ejecting some components like the rangefinder and RFD900 radio).
Q1: Why would the aircraft exhibit such a stable climb while in Land mode (which had been previously demonstrated working on this same aircraft)

The excitement was not over yet! Following the crash, the aircraft continued to claw around inverted on the floor for more than a minute while the flight crews struggled to secure the wounded beast and disconnect the battery.
Q2: Why did the crash check and RC disarm commands fail to secure the aircraft following the crash?

I understand the crash check has several requirements to trigger:
:heavy_check_mark:1. the vehicle is armed
:heavy_check_mark:2. the vehicle is not landed (as far as it can tell)
:heavy_check_mark:3. the current flight mode is not ACRO or FLIP
:heavy_check_mark:4. the vehicle is not accelerating by more than 3m/s/s
:heavy_check_mark:5. the actual lean angle has diverged from the desired lean angle (perhaps input by the pilot) by more than 30 degrees?

Any insight into these two mysteries?

2

Looks like a Rangefinder problem to me. Graph Rangefinder distance and some Rangefinder related CTUN parameters.

1 Like
3

Thanks for taking a look.
Does the lack of desired rangefinder altitude mean the autopilot is not using the rangefinder when in Land mode? That green line comes and goes with the mode switches.

S1000_Crash_RNGFND Alt
S1000_Crash_RNGFND Alt1328×906 81.1 KB

This does show that in the last three oscillations leading up to the crash the range finder was reporting less and less accurately, particularly at the beginning of the aircraft’s positive altitude changes.
I still am not sure why the CTUN DAlt was increasing and positive during those land mode segments before the crash.