BMW Group Plant Landshut is becoming a centre for developing the software behind AI-supported humanoid robotics in component production. The site will handle open software platforms, training data, simulation and motion planning, complementing humanoid robot pilots already running at BMW’s Leipzig and Spartanburg plants. The work is being carried out with partners from industry and academia.
The BMW Group is deepening its work on so-called physical AI, and it has handed BMW Group Plant Landshut a central role in developing the software that makes AI-supported robotics possible on the factory floor. According to the company, the plant will focus on open software platforms, generating and processing training data, simulation and motion planning, and the training of robotic systems.
BMW says the effort complements pilot projects already exploring humanoid robots at its plants in Leipzig, Germany, and Spartanburg in the United States. Landshut, which the company describes as a centre of expertise for in-house component manufacturing, offers a broad mix of products and processes that BMW argues makes it well suited to testing AI-driven robotics under real production conditions.
What the robots need to learn
The central question, BMW says, is which tasks will actually suit humanoid robots in the future — particularly jobs that demand flexibility, fine motor skills and an understanding of the surrounding environment. To that end, Landshut is developing software that lets robots perceive their surroundings, assess a situation and decide on the right action independently.
“At BMW Group Plant Landshut, we are combining software expertise with industrial practice in our component production. This allows us to test new technologies at an early stage and carefully evaluate their benefits for production,” says Dr Wolfgang Bluemlhuber, head of Driving Dynamics and In-house Component Manufacturing.
Open platforms and learning by demonstration
The technical foundation is a reworked software architecture built on open platforms. BMW says these combine traditional deterministic programming with modern AI in the form of Vision-Language-Action (VLA) models, blending visual information, language comprehension and robot actions. The stated goal is a modular robotics ecosystem in which different systems can work together.
Reliable data is key. The project team captures information from real production environments and test setups and feeds it into robot simulations, where movement patterns and motion planning are modelled virtually so processes can be trialled in advance. New methods are first tested away from live production, run as pilots, then scaled up in stages.
In parallel, the humanoid systems pick up new skills through demonstration-based learning. Team members first perform a movement or work sequence, which is recorded by camera systems, motion-capture suits and data gloves. That data is then turned into generalisable behavioural models — called policies — that can be reused across different tasks.
Working with startups and universities
To refine the technology, BMW is collaborating closely with partners at the Landshut site. Chief among them is Athenyx Robotics, a startup created specifically for this field that grew out of the Laboratory for Machine Tools and Production Engineering (WZL) at RWTH Aachen University and sits close to the plant. BMW says the pair are building an “intelligence stack” that ties together perception, learning, simulation and decision-making, with the aim of letting humanoid robots handle complex industrial tasks on their own. The plant is also working with specialists at the University of Applied Sciences Landshut on data generation, simulation and specific use cases.
“For us, it is crucial to develop the technology in very close alignment with real-world requirements. We are not just testing what is technically possible, but also where humanoid robotics can truly deliver reliable benefits in component production,” says Christoph Jagoda, the physical AI project manager at BMW Group Plant Landshut.
What it means
Carmakers are increasingly betting that humanoid robots could take on awkward, low-volume or fiddly jobs that fixed automation cannot easily cover. By concentrating the software work at Landshut, BMW is signalling that it wants to shape the underlying platforms itself rather than simply buying finished machines. For now, the emphasis is firmly on evaluation and pilots, with any wider rollout to follow in stages.
Source: www.press.bmwgroup.com