Environments to support https://github.com/sholtodouglas/learning_from_play and reinforcement learning for robotic manipulation.

An example playing out the teleop data, which also showcases the functionality of the env as an RL environment (reset goal, visualise goal, reward function etc).

An example where you can use the GUI to control the robot.

A more involved example, which shows how a test rig is used to reset the environment to specific locations and test goals, and plays a pretrained model.

The file to run if you want to teleoperate new trajectories, this works with any VR headset which can interface with steamVR. Instructions are within the file.

A high level overview of the code can be found here.

The environment also supports a number of more basic test environments, and other arms. The full list of these is defined in the gym registry here.