Robotics Company Galbot integrates NVIDIA’s physical AI platform, Jetson Thor, into the G1 Premium Humanoid Robot, with the aim of improving performance and enabling greater autonomy.
The G1 is designed for deployment across the retail, healthcare and logistics sectors and was featured at the World Robot Conference in Beijing earlier this month.
https://www.youtube.com/watch?v=sgmhntlzaee
According to Nvidia, the Jetson Thor model has more than seven times the AI computing capacity of the previous generation Jetson Orion models, and more than three times the energy efficiency.
By taking advantage of this upgrade, Galbot is now able to enable robots to perform complex planning and motion tasks with new levels of accuracy and efficiency.
Galbot credits the Sim2Real approach, combining Nvidia’s Isaac Sim and Isaac Lab frameworks with open source robot simulation, to advance.
In addition to its foundation, the company has created its own vision language action models, such as GraspVla, TrackVla and GroceryVla. This allows robots to do everything from zero-shot grappling and autonomous navigation to dense retailing with human-like performance.
“The G1 Premium, currently operating at Nvidia Jetson Thor, is already experiencing significant advances in speed and improving real-time reasoning capabilities,” said Professor Wang, founder and CTO of Galbot. “This early adoption allows us to push our own VLA models to new levels of real-world capabilities.”
G1 Premium Robots have already bolstered 10 smart pharmacies in Beijing, and will scale to 100 sites nationwide by the end of the year, Galbot said.
The company is also working with Bosch and other manufacturers to bring embodied AI-powered robots onto the production line, replacing rigid robot arms with more adaptive machines.
