Nvidia helps the humanoid robots with learning through Apple Vision
Robotics has played a significant role in Nvidia’s recent stratospheric expansion. When the chip manufacturer unveiled GR00T in March of last year, it was heralded as a revolution in humanoid robots. From the beginning, most of the top brands in the industry were available, including 1X Technologies, Agility Robotics, Apptronik, Boston Dynamics, Figure AI, Fourier Intelligence, Sanctuary AI, and Unitree Robotics.
At CES on Monday, CEO Jensen Huang unveiled Blueprint, another essential element of GR00T. The new modality is based on imitation learning, a crucial element in teaching robots a new skill. It is considerably more important in the realm of humanoids, whose designs are based on a single teaching.
Imitation learning is what it sounds like. Any action taken by a person is replicated by the robot. It is effective for training systems designed to automate repetitive tasks, including those in factories and warehouses, where the first humanoid deployments occur.
Teleoperation is essential to all of this. It may be used to rapidly digitize human movements and remotely instruct robots by mimicking a real-world environment.
With GR00T Blueprint, users can create these activities using Apple’s Vision Pro. Once the operation has been recorded as a digital twin, the robot can replicate it repeatedly.