Physical AI for Industrial Robotics and Automation

FANUC robots can identify and track workpieces moving in three dimensions, following them precisely in real time. This enables operations such as tightening screws on moving parts, a task that traditionally required precise fixturing or part stabilisation. With 1 millisecond high‑speed tracking performance, high-speed 3D tracking ensures consistent accuracy even in dynamic environments.

By issuing instructions in natural language to the AI agent for tasks like photographing order forms, recognising text, and transporting trays and parts, a robot can execute the kitting of workpieces. By combining an AI agent, which understands natural language to control the robot and autonomously execute tasks, with a masterless AI recognition function that identifies objects based on natural language inputs, it becomes possible to modify operation logic, just as if instructing a human.

FANUC robots equipped with AI-driven perception can detect nearby people and adjust their paths in real time without stopping production. Using the torque command input of the “Stream Motion” function, the force generated by the robot arm is limited within a safe operating range. If a human enters the robot’s working area, the robot automatically modifies its trajectory to avoid contact. Once the area is clear, the robot returns to its original path and continues the task. This enables safe, uninterrupted collaboration on the factory floor.

Using Vision-Language-Action (VLA) foundation model, two FANUC robot arms can work together to imitate human tasks, like folding a t-shirt or performing highly dexterous wiring operations. The system detects cable tension in real time and manipulates soft, flexible cables with human-like sensitivity. Leveraging multi-axis articulation, the robots route cables across complex 3D paths, including height, depth, and tilt, while maintaining accurate handling and preventing damage.

FANUC robots can recognise voice commands in multiple languages, automatically generate Python programs using generative AI, and execute the resulting tasks while perceiving their surroundings. Users provide instructions verbally, and the robot interprets the language, generates the appropriate program and carries out the requested action.

These capabilities also support complex tasks such as:

Rolling a die and placing it in the correct location based on the number shownStacking a die of a specific color on top of another

These tasks previously required specialised programming but can now be executed through intuitive voice commands.

FANUC supports advanced simulation and digital‑twin technologies that allow manufacturers to design, test, and validate robotic systems before deployment. Photorealistic virtual factory environments help teams model robot behavior, evaluate cell layouts, and optimise performance in a risk‑free setting. By using NVIDIA Isaac GR00T to train the robot motion and object behavior, the robots autonomously learn to handle previously “unseen” situations. FANUC robots are available in formats suitable for modern simulation platforms, enabling cycle‑time analysis, trajectory validation, and AI training data generation in virtual production workflows. These tools shorten commissioning time and improve overall system accuracy.