Universal Robots has launched the UR AI Trainer, a leader-follower system developed with Scale AI that captures synchronized motion, force, and vision data on production-grade cobots rather than research platforms. Unveiled at Nvidia GTC 2026, the platform is intended to shorten the path from AI model development to deployment on factory hardware.
The system allows operators to guide a “leader” robot while a synchronized “follower” reproduces the movement in real time, generating the structured multimodal datasets needed to train Vision-Language-Action models. Universal Robots is positioning that workflow as a way to address one of the more persistent problems in industrial robotics: training data collected on lab machines often transfers poorly to production environments, especially where contact, force control, or delicate handling matter.
Anders Beck, vice president of AI Robotics Products at Universal Robots, said customers now need “a way to collect high-fidelity, synchronized robot and vision data to train AI models on the same robots they intend to deploy”. That emphasis on shared hardware is central to the pitch. Universal Robots says the AI Trainer draws on its direct torque control and force feedback capabilities, allowing developers to capture interaction data on the same class of cobots already used across large industrial installations.
The launch also deepens the company’s wider physical AI push. The AI Trainer runs on UR’s AI Accelerator platform, while Scale AI handles data capture and preparation, and Generalist AI supplied the embodied foundation model used in a GTC demonstration where two UR robots completed a smartphone packaging task. Universal Robots also showed a simulated version of the same workflow in Nvidia Omniverse and Isaac Sim, reflecting a broader industry move towards combining real-world capture with synthetic data generation.
Industrial robotics is moving away from tightly scripted automation towards systems that can generalise across variable tasks, but the economics still depend on reliable data pipelines and equipment that can survive factory duty cycles. Universal Robots is expected to release a large-scale industrial task dataset collected on UR robots later this year. More information is available on UR’s AI Trainer page.
