ABB and NVIDIA Use Physical AI Simulation to Transform Factory Automation

ABB and NVIDIA are joining forces to advance the use of physical AI simulation in industrial automation, seeking to bridge the long-standing gap between virtual training and real-world deployment of robotics in manufacturing. Their partnership centres on embedding NVIDIA Omniverse libraries into ABB’s RobotStudio software, allowing for highly accurate digital simulations of factory operations.

Manufacturers have typically struggled to transfer AI-driven robotics from testing environments to actual production lines, mainly due to the unpredictability of material properties, lighting, and product variation on the factory floor. As a result, many teams have had to construct physical prototypes, increasing both cost and time to market.

The upcoming solution, RobotStudio HyperReality, is anticipated in the second half of 2026 and is already generating global interest. Through its integration with NVIDIA technology, engineers can design and virtually validate complete automation cells, including robots, sensors, lighting, and parts, before any hardware is installed. Central to this workflow is the export of a fully parameterised digital station as a USD file to the Omniverse environment, where a virtual controller mirrors the firmware of the physical machine.

According to ABB, these advances could cut deployment costs by up to 40 percent and halve time to market. The use of synthetic images for computer vision model training—paired with ABB’s Absolute Accuracy technology—allows for extremely precise movement, reducing positioning errors from 8-15 millimetres to about 0.5 millimetres.

The technology is already being trialled by early adopters. Foxconn, a major electronics manufacturer, is piloting the system for assembling consumer devices—a process complicated by frequent product changes and delicate components. By adopting virtual training and simulation, Foxconn expects higher accuracy and quicker setup times while reducing reliance on costly physical testing.

Workr, a California-based automation provider, also integrates its own platform with ABB robotics trained in the Omniverse environment. The company aims to showcase the onboarding of new parts within minutes, removing the need for specialised programming.

Beyond software, ABB is considering the use of NVIDIA’s Jetson edge computing devices within its Omnicore controllers. This move could enable real-time AI inference across deployed robotic fleets.

Marc Segura, President of ABB Robotics, stated that the combination of RobotStudio with NVIDIA Omniverse “has closed technology’s long-standing ‘sim-to-real’ gap—a huge milestone to deploying physical AI with industrial-grade precision.” Deepu Talla, NVIDIA’s VP of Robotics and Edge AI, emphasised the need for high-fidelity simulation to accelerate the adoption of AI-driven robotics at scale within the industrial sector.

Adopting digital-first simulation workflows is projected to reduce setup and commissioning times by up to 80 percent. As AI hardware and simulation advance, manufacturers are expected to shift towards data-centric engineering and targeted skills development for working with synthetic data.

Reference: artificialintelligence-news.com