Unstable drone on faster than real time simulations

[markostamos]

Drone gets unstable and falls when update rate of gazebo goes above 2x real time.

[Jaeyoung-Lim]

@markostamos I think this issue is lacking context / details to decide whether if this is a real issue.

How are you controlling the vehicle? Could you provide instruction can we reproduce your problem?

[markostamos]

@Jaeyoung-Lim i set a position (x,y,z) using the lee controller. As expected this works and the uav stays in that position but if i try to do the same thing with a faster update rate ( more than 2x the real time) the uav gets unstable and eventually falls.

[tbanus]

I am having a similar issue. I tried it with both Lee Controller and the PID controller I implemented. Both causes issues. My observation is that when the RAM usage reaches %99, the drone falls. However, I haven't tested this. I also suspect the ros time acts differently under these conditions since it messes with the derivative (dt) term of the PID controller. Has anyone reached a solution for this?

[Shono1]

I ran into this problem as well. Initially, I was keeping with the Gazebo default step size of 0.01 and just increasing the update rate. This caused instability. I fixed my issue by decreasing the step size as well. I did this because I noticed that 0.01 was the max step size for the controller I was using. I figured that working at the max step size was probably a bad idea anyway, as if even one frame was dropped the drone could become unstable. After changing the step size down to 0.005 and upping the update rate to 1 000 hz, I was able to get stable performance (indistinguishable from just a sped up 1x speed) at 5x realtime speed.