LW Lab
An open-source simulation framework designed to significantly advance robotics and embodied AI research, by Lightwheel and NVIDIA
Built on NVIDIA Isaac Lab , an open source robot learning framework , LW Lab is a simulation framework that aims to accelerate research in manipulation and loco-manipulation through rapid, scalable experimentation, covering the full pipeline from data collection and policy training to comprehensive policy evaluation, all within a unified simulation environment.
Check the news release of LW Lab and more about NVIDIA Opens Portals to the World of Robotics
A Complete Training Pipeline
Key Framework Features
- High-Quality SimReady Scenes and Assets: Includes high-quality OpenUSD-based simulation-ready (SimReady) assets from Lightwheel, enabling fast setup and high-fidelity data collection.
- Unified Benchmarks Skeletons & Home‑Living Scene Benchmark: Convenient, ready‑to‑use templates, including a kitchen‑based manipulation benchmark, to easily create and deploy large‑scale evaluations.
- Data Collection with Teleoperation: Built‑in solution of human‑in‑the‑loop teleoperation and randomization of data generation.
- RL Fine‑Tuning for VLA Models: Easily fine‑tune Vision‑Language‑Action models in simulation with parallelizable compute workflows.
High-Quality SimReady Scenes & Assets
Unified Benchmarks & Home-Living Scenes
Data Collection with Teleoperation
RL Fine-Tuning for VLA Models
This collaboration combines Lightwheel’s high-quality SimReady Asset pipeline, based on Universal Scene Description (OpenUSD), and advanced simulation infrastructure, with NVIDIA’s industry-leading Isaac ecosystem, uniting efforts to push the boundaries of what’s possible in simulation-first robotics development.
Stay Tuned for LW Lab
Open Source Access
LW Lab will be fully open-sourced soon on Lightwheel's GitHub repository for the entire research community. You can explore our other opensource projects built for robotic training in simulation for now.
View on GitHubExplore Now
Explore SimReady Assets
Preview our extensive collection of high-quality environments foundational to the framework.
Visit lightwheel.aiJoin Our Community
Connect with researchers, engineers, and robotics enthusiasts for updates and collaboration.
Join Discord
