How to use PyBullet in shared memory mode in conjunction with Python Multiprocessing?

[narittt]

Hi all,

I am trying to create a simulation with two parts. One is the actual graphical pybullet simulation, and another is an intensive/slow math computation. I want to use multiprocessing to split it into two separate processes. I have created a SimGame object that I want to be shared across processes, and am using multiprocessing BaseManager for that. The SimGame object takes a pybullet client as an argument (ie. SimGame(p, ..) where self.bullet_client = p). The function for my process looks like this:

  def main_game(SimGame):
      proceed = 1

      while proceed:
          SimGame.step()
          p.stepSimulation()
          time.sleep(time_step)

However, since you cannot pickle modules in multiprocessing, I am a little confused about how to use PyBullet to instantiate the SimGame object.

I think the approach would be to use p.connect(GUI_Server) in the main file, and then have p.connect(SHARED_MEMORY) under SimGame. Something like this

SimGame(object):
    def __init__(self):
        p.connect(p.SHARED_MEMORY)

that would then be called from another file

from sim_game import SimGame

def main():
    p.connect(p.GUI_SERVER)
    p.setAdditionalSearchPath(pd.getDataPath())
    class simManager(BaseManager):
        pass

    simManager.register("simulator_game", SimGame)

    manager = simManager()
    manager.start()
    sim = manager.simulator_game("circle")

When I do this, the GUI, but is just stuck and I get a spinning mouse cursor. I'd appreciate any clarification about the right way to do this.

TL;DR: Is it possible to connect to p.GUI_SERVER from one file, and then connect to the same server using p.SHARED_MEMORY from another? I tried to do this, but my simulation freezes. See above for details.

Thanks!