New drone unstable loiter mode crash

Hello,

We recently build new drone intended to lift camera package up to 5 kg, and try to test flight it for the first time yesterday. It had no payload and therefor had more power than necessary for its weight. We also forgot to calibrate hover throttle before. We launch in stabilized mode and everything seemed ok. Then we switch to hold mode and drone shot up in the air quickly, then started oscillating and becoming unstable. I think we probably kill the output eventually to prevent it from going out of control, but I hoping to get some feedback about one event in particular:

  • When we switch into hold mode, it immediately pitch upwards. Is there a reason this would happen?

From the log, it look like the drone tuning might not be good since the command for pitch and roll is not being followed very good, even up to the point of switching into loiter/hold mode.

We fix everything (damage not too bad) but want to try and figure out the reason for such a crash in the first place. Could it just be bad tuning?

I see that the vehicle positions are lagging the commands quite a bit from the log.

Please let me know any tips or observations that might help get flying without this issue.

Thanks.

Log file is here: Log of Crash

@Francisco_Dankorov Sorry for your crash.

I can see from the log that no tuning was done on the rate loops (all default gains) and given that the drone is overpowered, it is quite clear that the rate loop gain is way too high. You can also see that when switching to hold, the rate loops immediately started to diverge.
To avoid any further crash, I would suggest you to follow the steps of the Multicopter PID Tuning Guide and to make sure the lower loops are stable enough in manual mode (make some roll/pitch quick motions, fly a bit around, …) before you switch to an assisted mode. Since you already know that the gain of the rate loops is too high, you can already start by scaling down all the gains: this can be done by reducing all three P-I-D gains or the K gain (setting the K gain to 0.5 should do the job).

Hello and thank you for your quick reply. We had look at the guide before, but apparently not in good detail. We were thinking default tuning should have let it fly decent in all modes. We will follow your advice and reduce gains to start, then try tuning in manual mode.

Thanks.

Is there any kind of test stand or setup that can help tune the drone when it is potentially unstable? I am think of something like 2(or 3)-axis platform that is fixed to the floor. i.e. the roll and pitch, and maybe yaw, can move on pivots or bearings, but the drone can’t lift. It can only stabilize its attitude. Would something like this simply not work since altitude and land detection is taking place?

What about a bit round frame/cage that the drone flies within? Do you know of anyone do this?

@bresch We set K to 0.5 and tried again - we were able to fly this time at least in Manual/Stabilized mode.

We see still oscillation in the roll angle and pitch angle plots, and it seem like the roll rate and pitch rate plots are lagging or delayed. Maybe vibration is not great either, based on log file.

See initial tuning log here.

Does this oscillation in position and lag in rate mean that filtering(and thus delay) is too high?

We still feel the drone is oscillating too much so we tune gain down even more. With K = 0.3, the rate loops seem to not track well at all…seems we are moving farther away from good tuning. It still fly, but still oscillating.

See second log here.

We start to think maybe the Roll and Pitch control gains are also too high, so we tune down that about 30%…this time, we keep tipping over at the start before even fly:

See here.

We see integral is not zero when drone is sitting waiting to arm, so we wonder if this integral is causing the initial tilt when taking off.

Any recommendations or advice for trying other tuning steps?