Reinforcement Learning with PyBullet

[hadarshavit]

Hi,
I'm trying to use pybullet for reinforcement learning. I'm using the NAO robot with 2 cameras.
The environment takes long time to run (it takes about the same time as with a real world robot).
I ran a profiler on the code and most of the time (68%) is spent on stepSimulation.
Update: reducing the camera's FPS improves the performance, but I'm still looking for further improvement.
How can I "tell" the simulation when to do the rendering (i.e. in the end of each step)?
Thanks