Shanghai-based Fourier has unveiled its latest humanoid robot, GR-2, which showcases advancements in hardware design, adaptability and dexterity. The robot was developed using NVIDIA’s Isaac Gym platform for reinforcement learning, which allows for the simulation of complex real-world scenarios and reduces testing time and costs.
Fourier also uses NVIDIA Isaac Lab, an open-source modular framework, which aims to simplify robot learning. By integrating NVIDIA TensorRT and CUDA libraries, Fourier has optimized GR-2’s AI capabilities, enabling complex maneuvers and real-time motion control. The adoption of NVIDIA technologies by Fourier has reduced training times and improved simulation accuracy, allowing the company to explore new possibilities for complex AI functions.
Fourier CEO Alex Gu emphasizes the improvements in real-time motion control and AI decision-making are setting new benchmarks for humanoid robotics, particularly in service, academic research, and medical rehabilitation sectors.
