AI in Robotics Moves to Real-World Applications with the NVIDIA Ecosystem

[Erkan Teskancan]

At the GTC 2026 event, NVIDIA demonstrated its leadership in the transition of artificial intelligence from theoretical stages to real industrial applications. By combining simulation, training data, and hardware with partners like FANUC, Universal Robots, and Infineon, NVIDIA is enabling more efficient and flexible robot control in factories.

### NVIDIA and FANUC Collaboration
FANUC is integrating its Roboguide simulation tool with Nvidia Isaac Sim and Omniverse platforms. This allows for the design of production lines, virtual validation of robot behaviors, and subsequent implementation on the factory floor. As a result, the discrepancy between simulation and real-world data is reduced, commissioning time is shortened, and the need for rework decreases.

### AI-Powered Control in FANUC Robots
FANUC offers a flexible platform that doesn't tie users to closed systems, providing real-time AI inference with Jetson edge modules and supporting ROS 2 and Python. Robot programming is also evolving, moving towards systems that can understand inputs like voice commands and automatically generate code, rather than relying on predefined movements. This reduces setup time and facilitates modifications.

### Universal Robots and Training Data
Universal Robots, with its UR AI Trainer system developed with Scale AI, enables one robot to mimic the movements of another robot that is physically guided by an operator in real-time. The system collects motion, force, and visual data on production-quality hardware, improving model training for real-world conditions.

### Simulation with Nvidia Isaac Sim and Omniverse
Nvidia's Isaac Sim and Omniverse platforms provide synthetic data for AI while simulating complex tasks such as dual-arm assembly. When combined with real data, this creates a rapid cycle of training, validation, and reuse.

### Infineon and Hardware Assurance
Infineon is developing reference architectures for humanoid robots with platforms like Nvidia Jetson Thor. Actuators and sensors are simulated using digital twins, allowing motion control and perception systems to be tested before hardware completion. For security, TPM modules, secure boot, and post-quantum encryption support are offered. Additionally, compliance with functional safety and cybersecurity standards within Nvidia's Halos AI security framework is targeted.

### Other Innovations
Techman demonstrated a training system that captures human movements and translates them into humanoid robot actions. NVIDIA, beyond just providing processors, is becoming the fundamental building block of the entire robotics ecosystem with its simulation, data, hardware, and application infrastructure. These developments signify a transformation from fixed automation systems to scalable systems that can be trained and adapted to real-world conditions.