Hello,
We have deployed a new drone based on a Cube Orange+ with ArduCopter 4.5.1. A GPS is used as the altitude source. We have completed several flights without issue. During a recent flight, we experienced a major altitude loss that forced us to complete the mission in Stabilize mode to avoid a crash. It is worth noting that this drone is capable of landing on cables and has wheels that allow it to move along those cables once disarmed.
Here is the sequence of events:
- 8:38 AM: Takeoff from the ground.
- 8:40 AM: Landing on the cable at approximately 20 meters altitude.
- 8:40 to 8:44 AM: The disarmed drone moves slowly along the cable.
- Around 8:45 AM: EKF warning (EKF height variance), abrupt altitude variation while the drone is stationary.
- 8:53 AM: Takeoff in Loiter mode, altitude loss again, EKF lane switch, switch to Stabilize mode.
After analyzing the log, we are having difficulty understanding the root cause of the issue.
We observe that the altitude reported by the GPS is indeed correct, but the position estimated by the EKF still drifted:
The altitude variation appears to correlate with a drift in the estimation of the accelerometer biases, and also consequently with the vertical position innovation:
Since the drone was disarmed, vibrations were practically nonexistent:
What is also concerning is that all three EKF/IMU cores experienced this drift, though not necessarily in the same direction:
With no vibration, I find it difficult to understand how the accelerometer biases could have diverged so much. Perhaps I’m on the wrong track and this drift in bias estimation is caused by something else?
The drone has been grounded until we find an explanation—your help would be greatly appreciated to assist with the diagnosis.
Thank you.
Dataflash log: download
