Article Abstract
- AEON is BMW’s first humanoid robotic in Europe, headed to Plant Leipzig for battery and part manufacturing — with two models concentrating on full manufacturing by finish of 2026.
- Hexagon’s AEON makes use of 22 sensors, self-swapping batteries, and 4 layers of bodily AI together with imitation studying — the place simply 20 demonstrations are sufficient to coach autonomous operation.
- BMW is making use of laborious classes from its Spartanburg pilot with Determine AI — 30,000 X3s constructed, 90,000 elements moved — to speed up the Leipzig rollout, backed by a brand new Munich-based Heart of Competence for Bodily AI.
BMW is putting humanoid robots on the manufacturing facility flooring at its Leipzig plant — the primary time the corporate has deployed them at a European facility — because the auto trade bets on AI-powered robotics to deal with bodily demanding work and enhance manufacturing effectivity. The machines, provided by Hexagon Robotics, will work on meeting traces and in high-voltage battery manufacturing for electrical automobiles, an space the place workers at present put on cumbersome protecting gear and face repetitive bodily pressure.
The Leipzig pilot started quietly in December 2025. A broader check is deliberate for April 2026, with a full-scale everlasting pilot launching in summer season 2026. Two AEON models will work concurrently throughout two use circumstances, with BMW anticipating each to be in manufacturing by year-end.
“Digitalisation improves the competitiveness of our manufacturing — right here in Europe and worldwide,” stated Milan Nedeljković, BMW’s head of manufacturing and incoming CEO. “The symbiosis of engineering experience and synthetic intelligence opens up totally new prospects in manufacturing.”
Contained in the Announcement: BMW’s Expertise Campus, Munich
BMWBLOG attended the unique reveal occasion at BMW’s Expertise Campus in Munich, the place the corporate introduced collectively media, engineers, and know-how companions to unveil the Leipzig partnership and put AEON by way of its paces in reside demonstrations and hands-on workshops. It was a uncommon look behind the scenes at how BMW is considering the following decade of manufacturing facility work.
Why Leipzig, and Why Hexagon
Leipzig is BMW’s most technologically full German plant, protecting battery manufacturing, part injection molding, press, physique store, and closing meeting — that means testing AEON there covers the total vary of manufacturing environments in a single location.
Hexagon, in the meantime, is a long-standing BMW companion in precision measuring tools and software program. Once they moved into bodily AI with AEON, BMW already knew their strengths — and it reveals. Arnaud Robert, President of Hexagon Robotics, introduced the robotic on the Expertise Campus, making clear that AEON was designed from the bottom up for industrial work. “We’re not within the dancing enterprise — we’re within the working enterprise,” he stated, drawing a deliberate line between AEON and the performance-focused robots that dominate tech demos.
AEON carries 22 built-in sensors — peripheral cameras, time-of-flight, infrared, SLAM cameras, and microphones — giving it full 360-degree real-time consciousness. That sensor suite isn’t only for navigation: it allows high quality inspection duties that conventional fastened robots can’t carry out. AEON strikes on wheels slightly than legs, a alternative Robert stated got here after rigorous testing of a number of locomotion methods. “Wheels grow to be by far essentially the most environment friendly locomotion mechanism when it comes to power use and velocity over distance,” he defined. “A manufacturing facility flooring is often super-even surfaces, so we really profit from the velocity.” AEON reaches 2.5 meters per second and swaps its personal battery in 23 seconds, enabling real around-the-clock operation.
What AEON Does — and How It Learns
Robert demonstrated two reside use circumstances on stage. First, autonomous door panel inspection: AEON mounted a high-resolution scanner able to capturing a million factors per second at 50-micron decision, navigated to a automotive door, and carried out a full high quality scan — checking meeting tolerances and floor defects — with none human enter. Second, a human-robot handoff, demonstrating how AEON adapts when an individual enters its workspace mid-task.
AEON learns throughout 4 layers of bodily AI. On the base: simulation and reinforcement studying, the place hundreds of digital cases run concurrently to find optimum motion methods. “It turned out, by way of simulation and working hundreds of potential methods to do that, essentially the most environment friendly approach is definitely utilizing inertia,” Robert famous — a consequence no engineer would have hard-coded.
Above that: perception-based activity completion, the place telling AEON to “scan the door” is enough — it finds the door, positions itself, executes, and reviews. Third is imitation studying, the place roughly 20 teleoperated demonstrations are sufficient for AEON to generalize into autonomous operation. BMW’s Competence Heart recorded about 2,000 demonstrations for a single bin-picking activity, coaching a mannequin in a single day. “If you happen to would automate that with conventional options, this is able to not be per week — this is able to be much more,” a BMW engineer instructed us throughout the workshop session. Lastly, world fashions permit AEON to deal with unfamiliar objects by reasoning about them in context, adapting grip and motion on the fly.
The Q&A: Candid Solutions on Arduous Issues
After the demos, Robert fielded a pointy set of questions from the ground — and the exchanges throughout the workshops have been extra revealing than the presentation itself.
On LLM hallucination in bodily robots, Robert was direct: “We see very, little or no hallucination” as a result of AEON’s basis fashions are skilled on curated, domain-specific manufacturing facility knowledge slightly than the open web. Two onboard NVIDIA Jetson playing cards deal with all AI processing regionally — one for sensor fusion, one for activity execution — that means the robotic can detect and proper errors in actual time with out cloud dependency.
One attendee put the coaching knowledge problem bluntly: LLMs like ChatGPT are constructed on internet-scale knowledge, however “there’s no Google seek for robots.” Robert acknowledged it and outlined the two-part reply: steady teleoperation knowledge assortment that feeds again into the mannequin for all robots concurrently, plus AEON’s sensor benefit — “we seize not solely the actuator actions, which most individuals do, we really seize the whole knowledge of the surroundings.” BMW added a 3rd layer: their proprietary manufacturing knowledge — CAD fashions, digital twins, documented manufacturing duties — is coaching materials no competitor can entry.
Requested whether or not robots might ever maintain tempo with human employees on BMW’s meeting traces — which run at a takt time of 55–56 seconds per car, with employees bodily climbing into shifting vehicles — BMW was frank: “A robotic leaping right into a shifting automotive and assembling elements — I can not see this within the close to future.” The place robots have already got the sting is consistency. “The scanning is an efficient instance,” Robert famous. “The robotic learns the optimum solution to scan and does it precisely the identical approach each time. People make errors and take a look at once more.”
On cybersecurity — whether or not networked robots that be taught from one another might be hijacked — Robert acknowledged the priority immediately: “We take it fairly significantly, with fairly a little bit of cybersecurity round it and little or no management from the skin.”
Classes from Spartanburg, Utilized in Leipzig
BMW didn’t begin this experiment in Europe. The world’s first humanoid deployment at a BMW plant happened in Spartanburg, South Carolina, with Determine AI. Over ten months, Determine 02 labored every day ten-hour shifts on the BMW X3 physique store — shifting over 90,000 sheet steel elements and logging roughly 1,250 working hours. “The transition from the lab to actual manufacturing was, for me, sooner than we anticipated,” stated Michael Nikolaides, Senior VP for BMW’s Manufacturing Community. Two robots labored in tandem, one working whereas the opposite charged.
The important thing lesson BMW is making use of at Leipzig: contain IT, security, course of administration, and logistics from day one. Infrastructure readiness — security boundaries, Wi-Fi 6 connections, standardized interfaces — was a prerequisite, not an afterthought.
A Competence Heart and a Lengthy-Time period Imaginative and prescient
To scale this work, BMW has established a Heart of Competence for Bodily AI in Manufacturing in Munich. “We intentionally keep away from black field fashions in reside manufacturing,” Nikolaides defined. Each venture is assessed in opposition to 5 standards: industrial manageability, safety, financial viability, integration functionality, and scalability.
BMW’s manufacturing knowledge — its CAD fashions, digital twins, and a long time of documented manufacturing duties — offers it a structural coaching benefit that no exterior lab can replicate. Requested what number of robots BMW might need in 5 years, Robert didn’t hedge: “I hope there will likely be hundreds — we are going to see a distinct world with these robots, additionally outdoors manufacturing.” The Leipzig pilot, with two models and two use circumstances, is the place that depend begins.
