Teradyne Robotics used its Automate booth to show how physical AI is moving from experimental applications into deployable manufacturing workflows, with demonstrations focused on robot training, autonomous assembly and AI-enabled electronics production.

Rather than leaning on forward-looking concepts, Will Healy III, Director of Product and Industry Marketing, Teradyne Robotics, positioned the booth tour around applications that were relevant to today’s production environments. “These are real things that real people can deploy in real factories,” said Healy.

The AI Trainer

The tour started at Universal Robots’ AI Trainer, developed in collaboration with Scale AI. The demonstration used a pair of collaborative robots that operators guided through assembly tasks while the system captured high-frequency force, motion and positional data. Those datasets are used to train vision-language-action (VLA) models, enabling robots to reproduce the tasks autonomously without traditional programming.

READ MORE: A New Agent in Town: Inside the Industrial AI Workforce

“The software from Scale AI is collecting thousands of data points from the robots every second and putting that all into a database to build an AI model,” Healy explained. “As you build that training data, you can build a model to actually have the robots do the tasks [independently].”

The Generalist

This approach was further illustrated in a second demo developed with robotics and AI startup Generalist, where two UR12e cobots performed an assembly operation using a model trained via imitation learning. The application reflects hundreds of hours of operator-guided input, replacing traditional teach-pendant programming with data-driven model training. If the robot makes a mistake, it iterates until the task is completed, noted Healy.

The Cable Inserter

A third demonstration, developed with Cambrian, a developer of AI-powered 3D vision systems, highlighted AI-driven automation in data center infrastructure manufacturing. Two UR7e cobots, guided by Cambrian’s vision system, assembled high-density server racks by routing and inserting copper cabling. This operation is typically performed manually due to the complexity of handling deformable materials.

Wires exhibit what is known as “memory,” meaning they retain their shape and flex unpredictably, creating variability that conventional automation often struggles to accommodate, Healy explained. The vision system identifies connector orientation and dynamically adjusts robot trajectories in real time prior to insertion.

READ MORE: 5 Things Engineers Need to Know About FactoryTalk Orchestration Software

The data center use case also played into Teradyne Robotics’ position that many applications often associated with future humanoid systems, such as flexible assembly and labor-intensive tasks, can already be addressed using collaborative robots paired with AI models.

“There’s a lot of discussion around humanoids solving labor and flexibility challenges,” Healy said. “Whether you want to apply a humanoid in the future or want to apply a humanoid today, you can start with a with AI Trainer...and then go to a dual-arm assembly application like this.”

Building Autonomy

For Healy, building more capable autonomous systems starts with applying collaborative robots, AI vision and imitation learning in current production environments. If the AI Trainer is the starting point, applications such as the ones from Generalist and Cambrian represent the next logical progression in an automation strategy, demonstrating how manufacturers can begin deploying physical AI while laying the groundwork for increasingly autonomous systems.