A robot implemented with dynamic graph manager doesn't come with a simulator by default.

There are multiple ways and at different levels to implement a simulator:

Implement a simulator in C++ and start it as dynamic graph manager from the console
Create a device in python and use another simulator from python

7.1/ Simulator implementation in C++

For an example, see https://git-amd.tuebingen.mpg.de/amd-clmc/dynamic_graph_manager/blob/master/demos/single_motor/single_motor_device_simulator.cpp.

7.2/ Device in python

To test a dynamic graph implementation, one way is to create a Device object in python. This device object is created using the same / a similar robot-yaml file as for the dynamic graph manger's real robot. At every integration step of the simulation, the sensor values must be stored on the sensor signals of the device and the dynamic graph is executed. Last, the control commands must be forwarded to a simulator, the simulator is stepped and the next cycle begins.

The following code shows how to create a device and robot. It also scatches how to do the simulation loop.

from dynamic_graph_manager.device import Device
from dynamic_graph_manager.device.robot import Robot
# Create the device and robot object.
device = Device("your_robot_name")
device.initialize("path_to_the_robot_yaml_file.yaml")
robot = Robot(name=local_device.name, device=local_device)
# This creates the controllers and plugs the sensor to the control signals etc.
setup_controller(robot)
# Setup your simulator. This might be pybullet etc.
sim = create_simulator()
# Do the simulation loop. Here, we assume to simulate 4000 steps
for it in range(4000):
  # Fill all the sensor signals based on the simulator values.
  robot.device.<join_positions>.value = sim.get_joint_positions()
  ...
  # Execute the dynamic graph. This computes the control output.
  robot.device.executeGraph()
  # Pass the computed control signals to the simulator.
  sim.control(robot.device.<ctrl_commands>.value)
  # Step the simulator by one dt.
  sim.step()