BMW announced that it will officially deploy two Aeon humanoid robots developed by Hexagon Robotics at its Leipzig plant in Germany this summer, marking the first mass-production line application of humanoid robots for BMW in Europe. Aeon stands about 1.65 meters tall, weighs 60 kg, has a maximum movement speed of 2.4 meters per second, a short-term load capacity of 15 kg, a continuous load capacity of about 8 kg, and is equipped with 21 sensors. Notably, the robot uses a wheeled chassis instead of bipedal legs, which BMW believes is more efficient and stable in a flat factory environment. In terms of battery life, Aeon lasts about 3 hours per charge, far short of the standard 8-hour shift. Hexagon has therefore designed an automatic battery swap system that can be completed in about 3 minutes. Michael Nikolaides, Head of Production Digitalization at BMW, said the core reason for choosing humanoid robots is their ability to directly adapt to existing production lines — their form factor is similar to human workers, allowing them to fit into existing workstations without the need for large-scale facility modifications for robots. Initial tasks will focus on repetitive work such as parts transportation and pick-and-place operations in battery assembly.
In terms of training, BMW combines “Digital Twin” factory simulations using Nvidia software with remote control, allowing the robots to first optimize their movements in a virtual environment through reinforcement learning before being deployed on real production lines. Arnaud Robert, President of Hexagon Robotics, said that “imitation learning” is expected to compress the training cycle from months to days, with the ultimate goal being that the robot can replicate an action after observing a human perform it — a capability that may be achieved within one to two years. Previously, BMW tested Figure O2 robots at its Spartanburg plant in the U.S., which participated in the production of about 30,000 X3 vehicles, and found that AI-driven humanoid robots significantly outperformed traditional industrial robotic arms when dealing with complex disturbances such as slight part displacements. Bill Ray, VP Analyst at Gartner, pointed out that as robot costs decline significantly, it is more cost-effective to adapt robots to existing workflows than to redesign factories. However, he also cautioned that the market may be overestimating humanoid robots, saying that “many demonstrations are more like capital market performances aimed at inflating valuations”.