AI GitHub
Isaac Lab for GPU-accelerated robot learning
Configure simulated robotics tasks for reinforcement or imitation learning.
Key Takeaways
- Isaac Lab builds on Isaac Sim and defines scenes, observations, actions and rewards for learning experiments.
- Use the supported simulator and accelerator environment.
- Run one documented task with a few environments.
- A policy trained in simulation still needs separate testing of sensing, dynamics and safety on hardware.
How it works
Isaac Lab builds on Isaac Sim and defines scenes, observations, actions and rewards for learning experiments. Parallel environments let training collect experience in batches.
What you need
Use the supported simulator and accelerator environment. Make observations, controls and reward values visible for a single simulated task before training many environments.
What to try first
Run one documented task with a few environments. Inspect observation tensors and reward components before increasing the number of simulations.
Run a reset test with the same seed and compare observations. Unexpected changes can come from domain randomization or scene initialization rather than a policy.
Where to look for failures
Simulator and driver versions matter. A policy trained in simulation still needs separate testing of sensing, dynamics and safety on hardware.
Related Franklin pages
Separate the physics simulator, environment interface and learning algorithm. Test resets and termination before increasing the number of workers or training steps.
- Model articulated systems, contacts and actuators in a physics simulation. MuJoCo for simulating robot dynamics
- Define and interact with environments through a shared reset and step interface. Gymnasium for reinforcement-learning environments
Franklin’s robotics coverage gives examples of where simulation, training and hardware meet. Franklin’s robotics coverage
Project documentation
Try it for your workflow
Explore the project’s documentation for setup and usage, or save this page to a collection for later.
Read the project documentation ↗