NVIDIA unveils brain-and-body stack to train next-gen humanoids

NVIDIA has introduced a comprehensive robotics stack aimed at advancing humanoid robot development by integrating physics simulation, AI reasoning, and infrastructure within its Isaac Lab platform. Central to this update are the open-source, GPU-accelerated Newton Physics Engine and the Isaac GR00T N1.6 robot foundation model. Newton, co-developed with Google DeepMind and Disney Research and managed by the Linux Foundation, enables highly realistic simulations of complex physical interactions—such as walking on uneven terrain or handling fragile objects—facilitating safer and more reliable transfer of robot skills from simulation to real-world environments. Early adopters include leading academic and industry robotics groups. Isaac GR00T N1.6 incorporates NVIDIA’s Cosmos Reason, a vision-language reasoning model designed for physical AI, which enhances humanoid robots’ ability to interpret ambiguous instructions, leverage prior knowledge, and generalize across tasks. This model supports simultaneous movement and object manipulation, tackling advanced challenges like opening heavy doors. Developers can fine-tune GR00T