ABB Robots and AI Agents That Control Machines

Early computers and machine tools took their instructions from punched paper tape. A programmer worked out every step, punched it as rows of holes and fed the tape through a reader like the one on the DEC PDP-1. One wrong hole meant a wrong move.

The paper tape punch and reader of a DEC PDP-1 at the Computer History Museum. Photo: ArnoldReinhold / Wikimedia Commons (CC BY-SA 4.0)

The first industrial robots were not much easier to instruct. ASEA's IRB 6, launched in 1974, was taught point by point with a handheld pendant, and each program belonged to one machine. Lab automation went the same way: liquid handlers such as the Eppendorf epMotion or the Andrew+ pipetting robot each came with their own software.

An Eppendorf epMotion 5075 liquid handling robot in a lab. Photo: Magnus Manske / Wikimedia Commons (CC BY-SA 3.0)

On 27 August, Anthropic announced a research preview of the Model Hardware Standard, a shared driver that lets AI agents read and command lab and factory equipment, including robotic arms, liquid handlers and microscopes. Each device describes its state in plain language along with the safety limits it enforces. For now, testing is limited to selected labs and manufacturers.

The Andrew+ pipetting robot holding standard hand pipettes. Photo: Pocar19 / Wikimedia Commons (CC BY-SA 4.0)

ABB robots already share one language, RAPID, and ABB's RobotStudio lets engineers write and test programs on a laptop against a digital twin before anything moves. ABB has also added a generative AI assistant to RobotStudio to help with programming questions.

Early ASEA and ABB industrial robots on display. Photo: Dependability / Wikimedia Commons (CC BY-SA 4.0)

If agents start giving the orders, the limits have to live somewhere the agent can't reach. ABB's OmniCore controller runs SafeMove, which enforces speed and zone limits inside the controller, whatever sent the command. An AI agent could ask a robot to move, and SafeMove would still decide how far and how fast.

An engineer programming an ABB robot from a laptop with RobotStudio. Photo: ABB

I like the idea of telling a robot cell what I want in one sentence. I would still want the limits enforced in hardware and firmware that no language model can talk its way past. Splitting the work that way, AI for planning and fixed code for safety, looks like the sensible path for factories first and homes later.

An ABB OmniCore controller beside a robot cell. Photo: ABB

Innovation

AI solutions for effortless ABB robot control.

Automation

Robotics

ceojohntran@chatwithrobot.net

+84905311611

© 2025. All rights reserved.

qtran1215@gmail.com