Imitation Learning

The following workflows demonstrate end-to-end imitation learning with Isaac Lab Arena, covering teleoperation data collection, data generation, policy post-training, and closed-loop evaluation.

Currently, the following imitation learning workflow examples are provided:

• G1 Loco-Manipulation Box Pick and Place Task

• GR1 Open Microwave Door Task

• GR1 Sequential Pick & Place and Close Door Task

GR00T Container

Some steps in these workflows (policy post-training and evaluation) require the Base + GR00T container, which includes the GR00T model dependencies in addition to the standard Arena Base container. To launch it:

./docker/run_docker.sh -g

Not every step requires this container — the workflow pages will tell you when to use it.

• G1 Loco-Manipulation Box Pick and Place Task
• GR1 Open Microwave Door Task
• GR1 Sequential Pick & Place and Close Door Task