FANUC and NVIDIA partner to bring Physical AI to factory robots

This partnership marks a shift in how industrial robots are developed. Previously, automation relied on deterministic, play-back programming where humans manually guided robots through every motion. The article frames this as moving from a 'human writes the motions' era to a 'design the conditions for AI to learn' era, enabled by training in virtual spaces before deployment on real machines.

A central technical breakthrough described is the use of FANUC's own control algorithms, via ROBOGUIDE, inside NVIDIA's simulation environment. Because the virtual robot behaves exactly like the real one, the gap between simulation and reality shrinks, so policies learned in software can transfer to hardware without fine-tuning. This combination addresses a long-standing challenge in robotics, where simulated results often fail in the real world.

The article also highlights how this could address structural pressures in manufacturing, such as skilled labor shortages and the need for high-mix, low-volume production. The implication is that robots could handle tasks like cable routing or bulk part picking that previously required human sensing and feedback. The ability to pre-validate entire production lines virtually, including PLCs and conveyors, could also reduce the risk and cost of factory automation projects.