Isaac Lab
Overview
Isaac Lab is the official robot learning framework for Isaac Sim, providing APIs and examples for reinforcement learning, imitation learning, and more. The framework provides the ability to design tasks in different workflows, including a modular design to easily and efficiently create robot learning environments, while leveraging the latest simulation capabilities.
Some of its core features include:
Modular configuration-driven system to easily create and modify environments
Flexible user-designed workflow for optimized performance
Suite of robot learning environments for training and evaluation
Support for different reinforcement learning and imitation learning libraries
Connection to peripheral devices, such as game-pads and keyboards, for collecting demonstrations
Ability to augment simulation with custom actuator models for sim-to-real transfer
Deprecated Frameworks
Isaac Lab will be replacing previously released frameworks for robot learning and reinformcement learning, including IsaacGymEnvs for the Isaac Gym Preview Release, OmniIsaacGymEnvs for Isaac Sim, and Orbit for Isaac Sim.
These frameworks are now deprecated in favor of continuing development in Isaac Lab. We encourage current users of these frameworks to migrate your work over to Isaac Lab. Migration guides are available to support the migration process:
Deploying RL Policies in Isaac Sim
Isaac Sim has provided robot definition files and simulation demos for the Unitree H1, Anymal_c, and Boston Dynamics Spot robot for deploying reinforcement learning policies trained by Isaac Lab inside Isaac Sim.
Further Reading
For more background and full documentation of Isaac Lab, please see the following external references:
Additional Tutorials
The below set of tutorials details usage of reinforcement learning related components in Isaac Sim.