ABB and NVIDIA bring physical AI simulation to factory robots

Industrial robotics is advancing towards more realistic AI-driven development with ABB Robotics partnering with NVIDIA to integrate NVIDIA Omniverse libraries into its RobotStudio engineering platform. This collaboration aims to enhance simulation accuracy and expedite the deployment of AI-powered automation in factory settings. The new feature, RobotStudio HyperReality, is set to be launched in the second half of 2026 and is currently undergoing testing by manufacturers such as Foxconn and the robotics automation startup Workr.

For readers in the robotics, industrial automation, and AI sectors, this announcement signifies a significant shift towards simulation-driven engineering and synthetic data pipelines that have the potential to revolutionize how factories design and implement robot systems.

Physical AI simulation becomes more tangible

The integration of NVIDIA Omniverse libraries into ABB’s RobotStudio platform introduces high-fidelity, physics-based simulation to over 60,000 robotics engineers who utilize the software. The primary objective is to bridge the longstanding "sim-to-real" gap, where robots trained or programmed in virtual environments exhibit different behavior when deployed in real-world factory environments.

“By combining RobotStudio with the physically accurate simulation capabilities of NVIDIA Omniverse libraries, we have successfully closed the technology gap between simulation and reality. This is a significant milestone in deploying physical AI with industrial-grade precision for real-world applications,” stated Marc Segura, president of ABB Robotics.

The new workflow allows engineers to export a fully parameterized robot station, including robots, sensors, lighting, and kinematics, as a USD file into NVIDIA Omniverse. ABB’s virtual controller then operates using the same firmware as the physical robot, resulting in simulation outcomes that reportedly align with real-world behavior with up to 99% accuracy.

This platform also facilitates synthetic data generation for AI training purposes. Images produced in the simulated environment can be directly input into machine vision models, enabling engineers to train AI systems without the need for extensive real-world datasets.

According to ABB, the combination of photorealistic physics simulation, synthetic data, and its Absolute Accuracy technology, which reduces robot positioning errors from around 8–15 mm to approximately 0.5 mm, can provide a new level of precision for industrial applications.

Enhanced efficiency and cost savings

Besides accuracy enhancements, the companies suggest that this technology could significantly reduce engineering efforts and deployment costs. ABB estimates that the new workflow could decrease engineering time, lower deployment expenses by up to 40%, and expedite time to market by as much as 50%.

Manufacturers now have the ability to virtually design and validate entire automation cells before physical deployment, potentially reducing setup and commissioning times by up to 80% while eliminating the necessity for numerous physical prototypes.

Early adopters are already experimenting with the system. Foxconn is trialing the technology in consumer electronics assembly, where small metal components and frequent product changes often complicate automation processes. Meanwhile, Workr is integrating its WorkrCore platform with ABB robots trained using synthetic data generated through Omniverse.

ABB is also exploring the integration of NVIDIA’s Jetson edge AI platform into its OmniCore robot controller to enable real-time AI inference across its robot lineup, hinting at a future where industrial robots can be virtually trained and swiftly deployed with minimal manual programming.