ABB robots will make their debut at the 2026 World Artificial Intelligence Conference, showcasing innovative achievements in industrial-level physical AI.

ABB robots and NVIDIA jointly released a white paper to explain the disruptive value of physical AI in precision manufacturing.
The software stack of ABB robots’ physical AI toolchain made its debut, achieving ultra-high precision to bridge the “simulation-to-reality” technological gap.
Two major on-site demonstration workstations made their grand debut, vividly showcasing the latest achievements of implementing physical AI based on “Autonomous Versatile Robot (AVR™)”.
Shanghai, China, July 18, 2026. ABB robots showcased at the China 2026 World Artificial Intelligence Conference (WAIC), fully demonstrating its leading position in the field of physical AI. During the event, ABB robots and NVIDIA jointly released an industry white paper, deeply analyzing how industrial-level physical AI brings disruptive changes to precision manufacturing, and through the on-site demonstration workstations, allowing the audience to experience the cutting-edge applications of physical AI up close.
“Physical AI is fundamentally reshaping the capabilities, application scenarios and value creation methods of robots.” said Ma Sike, the global president of ABB Robotics. “Currently, robots, artificial intelligence, software and data are deeply integrating in the real world, giving rise to a brand-new industrial ecosystem. We are at the forefront of this wave. Leveraging the ‘Autonomous Versatile Robot (AVR™)’ technology, we integrate machine vision, perception and decision-making capabilities, endowing robots with the ability to understand the environment, adapt to changes and continuously learn. At the same time, we ensure their industrial-level reliability and accuracy. With the new physical AI toolchain, we can train robots at a faster speed and with higher accuracy, bridging the gap between simulation and reality with unprecedentedly high precision.”
“China is not only the world’s largest market for robots, but also a leading frontier for the deployment and ecological innovation of physical AI applications.” Han Chen, Senior Vice President of ABB Group and President of ABB Robotics in China, said, “We are actively responding to the urgent demands of Chinese customers for higher precision, greater flexibility and simplified deployment. At the same time, we are closely collaborating with ecosystem partners to jointly advance the research and development of physical AI technology and its application scenarios, ensuring that Chinese manufacturers can benefit first from these cutting-edge innovations that originated from local research and verification.”
White paper officially released: Charting the blueprint for the development of industrial physical AI intelligence
The white paper jointly released by ABB robots and NVIDIA this time focuses on elaborating the disruptive potential of industrial-level physical AI in enhancing the speed and flexibility of precision manufacturing. The white paper presents a set of methodologies to facilitate the rapid deployment of physical AI, emphasizing the need to integrate robot vision, digital simulation, and actual operations to form a standardized engineering implementation process that can be replicated and implemented.
This white paper was jointly compiled by Asin Technology, Deloitte and SKAI Intelligence. The white paper clearly outlines a core engineering implementation path: Before conducting any actual physical robot debugging, the identification of robot visual risks should be advanced to the digital engineering stage through digital twins, task-oriented synthetic data, and AI verification.
The release of this white paper marks another significant milestone following the strategic partnership announced by ABB Robotics and NVIDIA in March 2026. The two parties have joined forces to deeply integrate ABB’s software programming, design and simulation suite RobotStudio with NVIDIA’s OmniverseTM library’s high-precision physical-level simulation technology. This collaboration aims to bridge the long-standing “simulation-to-reality” technological gap in the industry and help manufacturing enterprises truly deploy physical AI in real robot application scenarios.
The complete white paper titled “Industrial-level Physical AI Empowering High-Precision Manufacturing” can be downloaded by clicking on this link.
Full power of physical AI showcased: Two innovative workstations make their debut at the event.
At this year’s WAIC conference, ABB robots further disclosed more details of the underlying software stack of their physical AI toolchain. This integrated training process enables interconnected robots to continuously learn from simulation data, synthetic data, and real data, and apply these learning outcomes with industrial-grade accuracy to real production. This forms a closed-loop system that continuously optimizes and iterates itself.
Through two on-site demonstration workstations, the ABB robots enabled the audience to directly experience the extraordinary capabilities of their “Autonomous Multi-functional Robot (AVRTM)”:
Sim2Real Digital Twin Real-World Closed-loop Workstation
This station has meticulously replicated the product model, robot, visual system and station behavior in a virtual environment, creating a highly realistic digital twin. Through benchmark tests of the robot’s motion accuracy under different materials, geometries and pose conditions in both virtual simulation and reality, and by generating synthetic data to cover the difficult-to-obtain long-tail edge cases, this platform successfully completed the “simulation-to-reality” closed-loop workflow. In this workflow, the fidelity of the simulation can be verified through a structured evaluation process, thereby providing an extremely reliable pre-baseline for the physical deployment on-site.
2. AI-guided clip assembly workstation
This workstation uses the massive synthetic image data generated by the digital twin to pre-train the visual AI model. Due to the extensive coverage of these data across various variables such as materials, lighting, and pose, the generalization ability of the AI model has been greatly enhanced, thereby significantly reducing or even eliminating the reliance on physical prototype production and physical experimentation. The trained visual model seamlessly integrates with the robot motion control. In the actual production environment, only a few corrections are needed to precisely execute complex assembly tasks.