Siemens Erlangen Humanoid Robot Completes 8-Hour Factory Shift with 90% Accuracy

Title

Siemens Erlangen Humanoid Robot Completes 8-Hour Factory Shift with 90% Accuracy

Content

Siemens, in partnership with UK robotics startup Humanoid and NVIDIA, deployed the wheeled HMND 01 Alpha humanoid robot into active logistics workflows at its Erlangen electronics factory in Germany, where it autonomously executed tote-handling operations continuously for over eight hours, moving 60 containers per hour while maintaining a pick-and-place accuracy exceeding 90% and integrating seamlessly with Siemens' production infrastructure. Stephan Schlauss, Siemens' Global Head of Manufacturing Motion Control, referred to the Erlangen facility as "customer zero," highlighting Siemens' strategy of piloting innovations internally before commercial release, while Deepu Talla, NVIDIA's Vice President of Robotics and Edge AI, said the deployment "sets a new benchmark for humanoid robots achieving real-world production goals on active factory floors." Unlike controlled laboratory tests, the trial ran for two weeks in January 2026 inside a fully operational factory environment working alongside human operators and other automated systems, with the robot performing the seemingly straightforward but variability-heavy task of retrieving totes from storage, transporting them to conveyor belts, and positioning them for human workers — a class of repetitive, physically intensive jobs that have traditionally resisted full automation because of product diversity and the need for real-time human-robot interaction. The deployment was made possible by Siemens' Xcelerator platform, which provides digital twin environments, AI-driven perception, PLC interfaces, fleet management, and industrial communication networks that let the robot synchronize with production lines, autonomous vehicles, and human collaborators in real time, and by NVIDIA's full-stack robotics layer: NVIDIA Jetson Thor for edge computing, NVIDIA Isaac Sim for high-fidelity simulation, and NVIDIA Isaac Lab for reinforcement learning and policy optimization, with this simulation-first approach letting Humanoid cut prototype development time from an industry norm of 18–24 months to seven months by training and validating behaviors virtually before physical deployment. The wheeled HMND 01 Alpha was previously tested in a proof-of-concept with Schaeffler on metallic bearing ring picking, and Humanoid — founded in 2024 by Artem Sokolov in London with additional offices in Boston and Vancouver and over 200 engineers — is also developing a bipedal variant with 29 degrees of freedom and RGB cameras, depth sensors, and 6D force/torque sensors for versatile manipulation. Siemens and Humanoid have not committed to a specific commercial rollout timeline but are positioning this as a "factory-grade model" for humanoid deployment, a scalable reference architecture for the industry rather than a one-off demonstration, and a path to addressing labor shortages in sectors where fully automated lines remain impractical due to product diversity, safety considerations, or the need for close human-robot collaboration.