Skip to content
World clockEU--:--UK--:--USA--:--CN--:--PLDEFRIT中文EN

portal about AI and technologyevents · analysis · interviews · technical background

Search
LIVE
›

Embodied AI unveils Air, a wheeled humanoid that learns factory tasks from workers

Swiss startup Embodied AI unveiled Air, a wheeled humanoid that learns factory tasks by recording human workers, at Humanoid Europe 2026 in Munich on 6 November, nine months after development began, and said it has signed contracts with major manufacturers.

AI & modelsNewsGrace OkonkwoPublished: 6 November 20263 min readSources 28
Embodied AI unveils Air, a wheeled humanoid that learns factory tasks from workers

What Embodied AI showed in Munich

Air moves on wheels rather than legs, which Embodied AI says gives it more stability on flat factory floors. Its arms are tendon-driven: the motors, gearboxes and electronics sit in the torso, and tendons pull lightweight arms that yield when they meet a person or an obstruction. The arm design grew out of co-founder and CEO Francesco Stella's doctoral research at EPFL, the Swiss Federal Institute of Technology in Lausanne.

Embodied AI says Air typically moves from human demonstration to autonomous operation in about a month, depending on the task. The safety case rests on a published specification: the company says it designed Air under ISO/TS 15066:2016, the technical specification for collaborative industrial robot safety, with CE marking planned for 2027. Embodied AI said it has signed contracts with major manufacturers globally and is preparing for larger-scale production.

Learning by watching workers, not by teleoperation

Air starts under human control. It records what the operator does, and Embodied AI's software learns the movements until the robot can run the task on its own. Human operators stay available for unfamiliar situations, and skills learned by one robot can be transferred to others, according to Embodied AI.

The target is repetitive work that needs precise hand movements and changes often, which Embodied AI says is hard to automate with conventional industrial robots. Air's arms are built for that light handling rather than heavy industrial work.

The software is not new. Embodied AI says its AI is already in paid factory deployments running a separate two-armed industrial robot system.

98% reliability in a customer evaluation

In an evaluation run by a major consumer electronics manufacturer, Embodied AI reported 98% reliability on a cable-handling task, against the customer's 90% target. The company says four other startups in the same evaluation failed to meet the manufacturer's requirements. The evaluation used the two-armed industrial system, not Air.

"Hallucination of AI models is acceptable in the digital space because humans review, reject, or correct the output before acting on it," co-founder and CEO Francesco Stella said. "In robotics, however, incorrect actions have immediate and irreversible physical consequences."

What happens next

CE marking for Air is planned for 2027, and Embodied AI says it is preparing for larger-scale production. The company says Air is already being tested at major manufacturers.

Comments 0

Sources

28
  1. 01Switzerland's Embodied AI Unveils Humanoid Robot Designed to Learn Factory Tasks From WorkersEN
  2. 02Churchill Capital XI shareholders approve $2.5B Agility Robotics humanoid SPACEN
  3. 03Xiaomi’s humanoid robots work a full 8-hour shift in its EV factoryEN
  4. 04I watched a humanoid robot clean my apartment. It felt like a taste of the future.EN
  5. 05Agility Robotics' Nasdaq debut invites scrutiny of humanoid robot marketEN
  6. 06Powered by GPT-6, Unitree humanoid robots master painting, chess without training, says CEOEN
  7. 07The Download: a humanoid cleans house and Samsung builds for the AI boomEN
  8. 08Tesla to wind down Cybertruck production as Elon Musk focuses on robotics and AIEN
  9. 09Agility Robotics begins trading on NasdaqEN
  10. 10Serve Robotics launches new business to become the AWS for robotsEN
  11. 11How Ambi Robotics helps its customers prepare for peak seasonEN
  12. 12Chinese regulators block SoftBank’s $5.4B acquisition of ABB RoboticsEN
  13. 13 Two dead, five injured in Taiwan PCB factory explosion EN
  14. 14Estonia approves 24/7 robotic pharmacies, paving the way for a European firstEN
  15. 15Robotics Startup Radar 2026EN
  16. 16How GlobalLogic helped Haystack Robotics build a UV disinfection robotEN
  17. 17LiberAI Unveils Liber-0, a World Action Model That Grounds Human Experience in Robot ActionEN
  18. 18Generate-then-Rank: Learning Symbolic Goal Interpretation for Embodied AgentsEN
  19. 19China blocks SoftBank’s US$5.4 billion takeover of ABB’s robotics divisionEN
  20. 20 Robot cube explores the International Space Station | Space photo of the day for Nov. 3, 2026 EN
  21. 21New ISO interoperability standard covers collaboration among mobile robots from multiple vendorsEN
  22. 22Maximo milestone: friendly robot has completed 300,000 tasksEN
  23. 23Tutor Intelligence raises Series B to scale model training, robot deploymentsEN
  24. 24Coming in 2027: Apple’s $1,000 home robot for FaceTime and switching on lightsEN
  25. 25Gentherm designs ThermAffyx to secure, heat patients during robotic surgeriesEN
  26. 26Top 10 robotics developments of October 2026EN
  27. 2728 Zero-Day Findings in Less Than Two Days: What AI-Speed Discovery Means for RoboticsEN
  28. 28Baidu’s new AI simulator gives robotaxis a crash course in rare road hazardsEN

All figures and quotations in this text come from the sources listed below.

Content prepared by the editorial team with AI assistance.

Grace Okonkwo

Grace Okonkwo

AI, models and technology

Grace Okonkwo covers AI, models and technology for FLASH24, working from primary sources such as model cards, API documentation and benchmark papers rather than vendor summaries. She checks training data provenance, evaluation conditions and reported scores against the underlying datasets before any figure reaches print. She interviews researchers and engineers directly, tracks release calendars from major labs, and compares successive model versions on the same tests. Her own self-hosting, home-network and documentation-reading habits feed straight into that desk, since she tests tools on her own hardware first. She does not publish benchmark claims without a reproducible method.

Newsroom →

Comments

0
  1. No comments yet — be the first.

Write a comment

Comments are public. We do not publish abuse, spam or advertising.