Bitcraze Forums

Sometimes when Crazyflie takes off, it starts to drift to one side, instead of hovering in one spot. It seems to be resolved after a power cycle or two, but it's still worrisome that it happens. For reference, we are using the flow deck, and this issue has been observed even happens when the battery is full. Subsequent take offs that are successful are in the same spot in a low wind environment, with all propellers being free from obstructions (ex. hair). What would cause this?

Hi,

Is it an offset at take off or is it a continuous drift? How are you controlling the Crazyflie?

It takes off, then moves in one direction (say, forward) until we stop the program (we test in a confined space, so there is not much room to let it roam around). We control it using the MotionCommander interface & flow deck; this happens in the absence of a control input. I.e., we tell it to take off, and it should only move on subsequent commands. We print out which commands we send to the copter, so we know this drift isn't caused by an accidental input by the user to move forward - it does that motion on its own.

How is the floor pattern and the lightning conditions?
If the floor has low-contrast pattern and the Crazyflie shadow is dropped (like with a spot-light), it happens that the flow will track the shadow and that can yields to drifts.

The room is bright (artifical light), and the floor is a patterned carpet. The strange part to me is that this occurs within 10-15 minutes of it working normally - under the same lighting conditions, same carpet, and it can have both styles of behaviour (though it usually does work; perhaps 18-19 times out of 20 it's good, if not more). The only fix for this behaviour is (or seems to be) turning it off for a minute or two, and turning it on again.

This is curious. Are you able to see anything wrong with the attitude or velocity estimate in the client when this happens?

The only thing that can take that long time to settle is heat, but it would be very surprising that heat causes this kind of problems. When this happens can you try to remove the battery and re-insert it to see if it fixes the problem (I am thinking of potential initialization problems with capacitor that may still be charged, I have seen that on a prototype deck recently).