Federated Fleet Learning - System topology

Purpose and goal: We need a new paradigm for training AI; one that makes it possible to collaborate globally around data, while solving today´s problems around data security, data privacy and data transferring. With edge learning, vehicles train AI models onboard with their own computing power on locally collected

data. This eliminates the need for data transfer to a central storage and computing infrastructure. Instead the local model improvements can then be transferred to a central infrastructure, or directly exchanged between other vehicles, to be merged into a high-performance global model. Expected results and effects:

* Faster development of applied AI that results in fewer accidents in traffic.

* Safer handling of personal data that enables unfettered development without compromising protection of personal information. * More efficient systems that saves both energy and money. * A generic architecture that can be adapted to other industries beyond automotive. Approach and implementation:

The project follows an agile framework and is divided into smaller milestones. Smaller work packages are refined and evaluated continuously through recurring meetings together with the project manager and steering group.