Atlas position control performance inconsistant across machines

edit

Hi all,

I recently observed a significant difference in performance running the Atlas position control plugin on two different machines. On one machine, the robot stands stably at the zero position. On the other, the robot has a qualitatively different response and falls over immediately.

The setup on both machines is identical (and both are using drcsim 1.3.1 and gazebo 1.3.1). The only discernible difference between the machines are the speed at which the simulation runs---something that should not affect a controller running on simulation time.

Any ideas as to what might be causing this and how it can be fixed? I'd be happy to provide the PD gains I used in this test.

Thanks!

--sk