Hello,

I'm running into a problem reproducing drone physics (using the Hector quadrotor model). Trying to tune it frequently produces angular velocities that are NaN.

I've read that Gazebo seems to have a high sensitivity to inputs for models with low mass, so this is a frequent problem.

In reality, something can only rotate so fast. Is there a way to constrain the angular velocity to prevent NaNs, so the simulation remains viable? Or some way to test and ensure applied forces / torques don't trigger this cascade?