Gasgoo Munich-The 2026 World Artificial Intelligence Conference and the High-Level Meeting on Global AI Governance kicked off on July 17 across four venues in Shanghai — the Expo, Zhangjiang, and West Bund districts. At this global AI gathering, Lightwheel AI offered a systematic first look at its full-stack infrastructure for physical AI data and evaluation.
As embodied intelligence moves toward large-scale deployment, the key to a robot's continuous evolution is no longer just the model itself. Instead, it hinges on the infrastructure that supports ongoing learning: data, simulation, evaluation, and feedback from deployment.

Image source: Lightwheel AI
The physical AI continuous learning system showcased by Lightwheel AI comprises four components: SimFoundry, EgoSuite, RoboFinals, and RoboStack. Together, they form a Real2Sim2Real closed loop for ongoing learning.
SimFoundry acts as the foundational simulation layer, using a proprietary "solve-measure-generate" pipeline to translate real-world physics into high-fidelity environments. EgoSuite delivers large-scale, first-person human operation data, spanning more than 25,000 environmental nodes and 100,000 task types. RoboFinals offers a physically realistic evaluation benchmark, achieving a high linear correlation between simulation and real-world results. Finally, RoboStack channels operational data and failure cases from actual deployments back into the system, creating a continuous learning flywheel.

Image source: Lightwheel AI
Lightwheel AI is also carving out a voice in shaping international standards. As the only Chinese company on the Newton Simulation Technical Committee, it joins NVIDIA, Google DeepMind, Disney Research, and the Toyota Research Institute in leading the next generation of open-source physical AI simulation standards.
It is also the sole Chinese firm on the EgoVerse Human Data Committee, working with Meta, Scale AI, and others to build the world's largest human data training ecosystem. Beyond that, the company is contributing to evaluation frameworks like Isaac Lab Arena, and its proprietary LeIsaac has been integrated into the official Hugging Face documentation.
Commercially, Lightwheel AI has secured the top spot in global deliveries across three critical areas: human video data, synthetic simulation data, and industrial-grade simulation evaluation. All of the world's top five "world model" teams are now clients. Some premium scenario data has been resold over 10 times, validating the company's shift from project-based delivery to platform-based infrastructure. In the first quarter of 2026 alone, new orders hit 550 million yuan — surpassing the total for all of 2025.









