Teradyne Robotics Unveils Production-Ready Physical AI for UR and MiR Robots

For years, the artificial intelligence boom has lived comfortably in the abstract. Algorithms have rewritten the rules of software architecture, generated flawless digital environments, and automated white-collar workflows at an astonishing clip. But take those same digital brains and try to make them perform a basic human task—like routing a flexible copper cable into a crowded data center server rack or picking a misaligned box off a shifting pallet—and the illusion of AI omnipotence quickly hits a wall. The physical world is chaotic, unpredictable, and notoriously hostile to software built on pure logic.

At the Automate 2026 exposition in Chicago, Teradyne Robotics—the corporate engine behind industrial giants Universal Robots (UR) and Mobile Industrial Robots (MiR)—unveiled a sweeping strategy to conquer this physical friction. Rather than showcasing conceptual, bipedal humanoids that remain years away from practical deployment, Teradyne shifted the goalposts by introducing a comprehensive ecosystem of “Physical AI” applications. Crucially, these systems are not laboratory experiments; they are commercially available tools ready for immediate deployment on factory floors.

The Shift From Deterministic to Reactive Automation

Traditional industrial automation is inherently rigid. For decades, robotic arms and autonomous vehicles operated on deterministic path trajectories. They relied on absolute coordinates, executing the exact same kinematic paths repeatedly. If a part arrived rotated by five degrees, or if a human walked into a logistics lane, the system threw an error code and stalled production.

Teradyne’s new physical AI framework bypasses this architectural limitation. By transitioning to sensor-driven, real-time reactive control, robots can now adapt to their surroundings dynamically. Vision-Language-Action (VLA) neural network architectures process high-dimensional sensory inputs—such as 3D point clouds from RGB-D cameras and real-time torque vectors from multi-axis force sensors—mapping them directly to low-level motor joint velocities.

“With physical AI deployed with UR and MiR robots, we are providing solutions enabling automation to work with the world as it is, not as we wish it to be,” said Jean-Pierre Hathout, President of the Teradyne Robotics Group, during the launch.

PolyScope X: The Next-Gen Software Layer

The technological foundation of this rollout is PolyScope X, Universal Robots’ next-generation software platform. While Teradyne preserved the high-fidelity motion-control algorithms that built its reputation, it completely overhauled the developer environment. The platform modernizes the user stack by incorporating containerized applications, native ROS 2 support, and standard web technologies.

The most notable feature of PolyScope X is the introduction of Logic Programs. These are multi-threaded background routines that run in parallel with the primary robot control loop. This native, PLC-style architecture allows engineers to coordinate complex work cell activities, swap real-time operational data, and integrate third-party peripheral sensors directly through the robot’s core computer. Because these Logic Programs operate independently of standard program pauses, safeguard stops, or power states, they significantly reduce the need for expensive external Programmable Logic Controllers (PLCs).

From Imitation Learning to the Data Center Floor

Teradyne’s hardware applications highlight how physical AI is targeting high-value, labor-starved industrial verticals. A key component of this rollout is the UR AI Trainer, an imitation learning platform developed in partnership with Scale AI.

Instead of writing thousands of lines of code to handle variable parts, a factory operator physically guides a UR cobot arm—such as the compact UR3e for precise bench-top tasks or the versatile UR5e—through a specific trajectory. (For an in-depth operational analysis, see our comprehensive Universal Robots UR5e Review). The system logs exact spatial movements while high-fidelity strain gauges at the tool flange record human force dynamics. The combined multimodal data trains a localized execution policy.

Furthermore, this data integrates natively with cloud-based architectures, allowing manufacturers to pipe proprietary datasets directly into the NVIDIA GR00T open VLA model or validate complex interactions inside the NVIDIA Isaac Sim simulation framework.

Other major practical rollouts showcased include:

• The MiR1200 Pallet Jack: Driven entirely by embedded physical AI vision, this autonomous mobile robot (AMR) navigates dynamic warehouse corridors, identifies shifting pallet openings on the fly, and operates seamlessly alongside existing fleet machinery like the MiR250 and the agile MiR200.
• The Generalist Partnership: Utilizing Generalist’s GEN-1 foundation models, twin robot setups executed intricate, dual-arm dexterous assemblies. The system corrects its own tracking errors in real time based purely on visual observation, eliminating the need for strict CAD-model matching.
• Infrastructure Automation (Cambrian): Aimed directly at the massive infrastructure boom surrounding AI data centers, this application uses dual-arm setups and Cambrian AI vision to automatically plug rigid copper cables into high-density server racks—a task notoriously prone to human error and manual strain.
• Natural Language Deployment (Trener Robotics): Through an AI-native conversational platform called Acteris, operators can deploy machine-tending tasks by typing simple natural language commands, shifting changeover times from hours to under two minutes.

Pragmatism Over Humanoid Hype

Teradyne’s strategic pivot highlights a broader cultural divide in the robotics industry. While venture capital continues to chase the science-fiction promise of general-purpose bipedal humanoids, industrial operators remain deeply skeptical of their high upfront costs, unproven reliability, and complex safety profiles.

Will Healy III, Director of Product and Industry Marketing at Teradyne Robotics, made the distinction clear when positioning the new rollout. “These are real things that real people can deploy in real factories,” Healy emphasized.

By injecting physical AI directly into the standard articulate arms and mobile platforms already trusted by compliance managers and plant floors globally, Teradyne has taken the shortest path to monetization. The physical world may be chaotic, but Teradyne is betting that the transformation of the factory floor will happen through evolutionary intelligence on proven form factors, rather than revolutionary shapes. This focus on immediate commercial scalability mirrors the broader industry metrics explored in our analysis of which robot categories will scale first.