BMW Group Plant Landshut has been named as the group’s central hub for AI-supported robotics software development. The site will build open software platforms, training data pipelines, simulation and motion planning capabilities for humanoid robotic systems, extending pilots already under way at Plants Leipzig and Spartanburg.

The plant’s role as a component production centre provides product and process diversity suited to testing AI robotics in real conditions. Software under development at the site is focused on tasks requiring flexibility, fine motor skills and contextual understanding of working environments.

The core architecture combines deterministic programming with vision-language-action models integrating visual information, language comprehension and robot actions. Training draws on data from real production environments and test setups fed into simulations to model movement patterns.

Humanoid systems also acquire new capabilities through demonstration-based learning, with operator movement patterns captured via motion capture suits and data gloves converted into generalisable behavioural models. These can then be transferred across different robotic systems and use cases.

BMW Group is partnering with Athenyx Robotics, a startup from RWTH Aachen University, to develop an intelligence stack covering perception, simulation and decision-making. The University of Applied Sciences Landshut also contributes to data generation and simulation.

Why this matters:

• Separating software development from operational pilots is deliberate architecture.Landshut builds the stack; Leipzig and Spartanburg test it in live production—a structure that lets BMW iterate on core robotics software without disrupting assembly lines while accumulating real-world deployment data simultaneously.
• Demonstration-based learning is the scalability unlock.Recording human movement via motion capture and converting it into generalisable behavioural models is significantly faster than manual programming. Crucially, the resulting skills can transfer across different robotic platforms and tasks rather than being locked to a single application.