Bosch and AIT to research federated learning for industrial machine clusters

Bosch Research and the Austrian Institute of Technology (AIT) have initiated a three-year research project focused on federated learning. The collaboration aims to implement machine learning models that allow industrial machines to share operational knowledge without the need to exchange sensitive raw data.

Unlike centralized cloud-based training, federated learning enables individual machines to refine models locally on their own control units. Only the learned model parameters are sent to a central repository to improve the system's overall performance. This approach is intended to enhance data privacy while allowing machines in a fleet to benefit from the collective experience, such as adjusting to extreme operating temperatures or identifying hardware malfunctions through behavioral anomalies.

AIT will contribute by addressing scientific gaps in current algorithms, while Bosch Research will provide the industrial application context, specifically targeting complex, dynamic systems. The project seeks to develop scalable methods that can be integrated across various Bosch product divisions.