Quickstart

The heart of this simulation is the SimFinger class. It takes care of initialising PyBullet and loading the robot into the environment.

from trifinger_simulation import SimFinger

sim = SimFinger(
    finger_type="fingerpro",
    enable_visualization=True,
)

The above example populates the simulation with a “FingerPro” robot, see Finger Types for the different robots that are supported.

By default, the fingers are pointing straight down initially, which usually means that they are intersecting with the ground. It is thus important to first set them to a feasible position before starting to run the simulation:

# set feasible initial position for FingerPro
sim.reset_finger_positions_and_velocities([0, -0.7, -1.5])

Now you can start to run the simulation by sending actions to the robot. Below is a very basic example that uses position commands to hold the robot at the initial position and prints the actual observed position:

import time

while True:
    # to have it run in real-time
    time.sleep(sim.time_step_s)

    # create a position action
    action = sim.Action(position=[0, -0.7, -1.5])

    # send it to the robot (this executes one simulation time step)
    t = sim.append_desired_action(action)

    # get the observation of the current time step
    observation = sim.get_observation(t)

    print(observation.position)

When using position commands, a PD controller is executed internally to compute the desired torques. You can also directly use torque commands:

action = sim.Action(torque=[0., -0.2, -0.2])

Regarding the meaning of the time index t in the above example, please see Time Relation of Actions and Observations.

Adding Objects to the Simulation

We have a few objects defined in the collision_objects module, which can easily be added to the simulation. For example to add a coloured cube add the following code after initialising SimFinger:

from trifinger_simulation import collision_objects

cube = collision_objects.ColoredCubeV2(position=(0, 0, 0.4))

Modifying the Simulation Environment

If you want to modify any property of the simulation (add objects, change parameters, etc.) beyond the interface that is provided by the classes of this package, you can simply access pybullet directly. For example to change the default camera pose in the PyBullet GUI:

import pybullet

pybullet.resetDebugVisualizerCamera(
    cameraDistance=2.0,
    cameraYaw=0.0,
    cameraPitch=-30.0,
    cameraTargetPosition=(0, 0, 0.2),
    physicsClientId=sim._pybullet_client_id,
)

Note, however, that all these modifications should be done after initialising SimFinger. Ideally also pass the “physics client id” to all PyBullet calls like shown above.

More Examples

For more examples, see the scripts in the demos/ folder of the trifinger_simulation package. Good starting points are

• demo_plain_torque_control.py: A very minimalistic demo, similar to the example above.

• demo_control.py: A slightly more complicated demo.

• demo_inverse_kinematics.py: Uses inverse kinematics to control the finger tip position.

• demo_cameras.py: Shows how to use the cameras in simulation (note that image rendering is pretty slow, though).