Getting Started

To get started, please follow the installation instructions to set up SOFA and SofaPython3.

After that, either look through the implemented Scenes, or have a look at the workflow below to directly get started with implementing your own reinforcement learning environment.

Environment and Scene Development Workflow

1. Create a new directory in sofa_env/sofa_env/scenes for your scene

2. Add a scene_description.py file that contains a createScene function. Look into sofa_env/sofa_env/scenes/controllable_object_example/scene_description.py for an example.

3. Add your SOFA components to the scene. sofa_env.sofa_templates contains a few standard components.

4. Iteratively test your createScene function by passing it to the SOFA binary. $SOFA_ROOT/bin/runSofa <path_to_your_scene_description.py>

5. Let createScene return the root_node, your camera (if you want to render images), and additional interactive components from your scene as a dict.

6. Create your SofaEnv class that describes interactions with the scene. Have a look at Interacting with the Simulation and sofa_env/sofa_env/scenes/controllable_object_example/controllable_env.py for an example.

7. Implement your environment’s _do_action, step, and reset functions.

8. Iteratively test your environment directly through python. python3 <path_to_your_env.py>