Universal Robots has partnered with Scale AI to introduce the UR AI Trainer, a sophisticated imitation learning system unveiled at NVIDIA GTC 2026. The platform allows operators to guide one robot while a second robot mirrors its movements, capturing synchronized motion, force, and vision data to train physical AI models in real-time. The innovation is designed to bridge the gap between lab-scale experimentation and real-world deployment, accelerating the development of AI-driven robotics applications in industrial settings.
This system addresses a critical challenge in robotics AI: acquiring high-quality, real-world training data. Traditionally, most datasets were limited to research environments or vision-only data collection, which restricted AI performance in contact-intensive and dynamic tasks. By incorporating force feedback and direct torque control, the UR AI Trainer enables the collection of rich, multimodal data, allowing AI models to learn complex interactions that were previously difficult to capture.
The collaboration between Universal Robots and Scale AI leverages the former’s global robotic platform and the latter’s robust data infrastructure. This integration provides a scalable approach to developing AI models that are accurate, adaptable, and ready for industrial deployment. It also accelerates model iteration, reduces development costs, and shortens the timeline from research to factory-ready AI systems.
Industry analysts view this as a significant step toward practical, large-scale adoption of physical AI in manufacturing and logistics. By enabling robots to learn from direct operator interaction, UR AI Trainer has the potential to reshape the way industrial AI is trained, moving away from purely simulated environments to real-world, production-grade applications.
