Lightwheel needed a way to overcome slow real-world robotics data collection and the sim-to-real gap so customers could train and deploy embodied AI systems without months of physical prototyping and expensive on-robot testing.
Reported impact
Lightwheel shortened robotics development cycles from months to weeks, achieved a 100:1 simulated-to-real data ratio, and demonstrated factory deployment of GR00T N1.5-powered humanoid robots at Geely while supporting partners such as AgiBot, ByteDance, and Figure.
Lightwheel built the Lightwheel Simulation Platform as a simulation-first embodied-AI workflow on NVIDIA Isaac Sim, Isaac Lab, Omniverse libraries, and the GR00T N1.5 vision-language-action foundation model. SimReady assets encoded in USD and MJCF let teams assemble accurate digital twins quickly, while VR headsets, exoskeleton controls, and teleoperation tools capture demonstrations inside simulation instead of on production robots. Lightwheel then combines MimicGen and DexMimicGen with those synthetic scenes to expand training diversity by 100 to 1000 times before fine-tuning GR00T N1.5 for specific robot morphologies such as Unitree H1. During prototyping, RTX 4090 GPUs provide local compute for embodiment adaptation and task optimization, and the same pipeline has already been used to deploy humanoid robots in Geely's live automotive environment.
SimReady assets, scene metadata, and teleoperation demonstrations are assembled in Isaac Sim and Omniverse-based environments, then expanded with MimicGen and DexMimicGen into large synthetic datasets. Those data are used to fine-tune GR00T N1.5, and the resulting policies are deployed onto robots such as Unitree H1 in live industrial settings.
100:1 simulated-to-real data ratio, eliminating expensive real-world data collection while maintaining physically realistic training environments
Lightwheel successfully deployed GR00T N1.5 foundation models in Unitree H1 humanoid robots at Geely's live automotive factory.
Powered by OpenUSD, the Lightwheel Simulation Platform delivers a complete embodied AI pipeline, from SimReady asset generation to cloud-based simulation and collection of teleoperation data in virtual reality.
The platform combines MimicGen, DexMimicGen with Lightwheel's SimReady assets, environments, and Isaac Sim to generalize teleoperated simulation data, scaling the value of synthetic data by 100 to 1000 times.
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
The company-controlled profile says Lightwheel Limited is "in 10080 N Wolfe Rd SW3 200, Cupertino, California, 95014, United States."
Related
NVIDIA Sample Sep 22, 2026 Sluicebox Sluicebox needed to turn fragmented bills of materials, supplier records, PDFs, CAD files, and emissions inventories into timely product carbon assessments, but manual life-cycle-analysis work was slow, expensive, and… 51-80% lower inference costs Lucy · NVIDIA Nemotron 3 Ultra · carbon graph NVIDIA Sample Sep 15, 2026 Lambda Lambda needed to raise token throughput and improve AI-cloud economics, but fixed rack and facility power budgets were leaving GPU capacity stranded and limiting how many nodes could run safely at once. +17% Hybrid Training Throughput Dynamic Power Software · NVIDIA DSX AI Factory Platform · NVIDIA DSX MaxLPS NVIDIA Sample Sep 1, 2026 Onetera Onetera needed to help more than 90 municipalities turn fragmented regulations, zoning maps, permit rules, and legacy documents into fast resident-facing services without imposing heavy IT overhead or expensive manual… 1.7x increase in data throughput GPT OSS 20B · Government Intelligence Agent · LangGraph 