Federated Learning & Edge Processing for Safe and Efficient Operation

Background

The methods for data collection and processing in the mobility sector have remained relatively unchanged over the past two decades. However, emerging challenges – best exemplified by electromobility – necessitate a novel approach to building these systems, with enhanced flexibility in storage and data management. Edge processing offers the potential for a new data logging infrastructure by reducing transmission costs and lowering analytics latency, benefitting vehicle manufacturers, fleet owners and drivers.

Goal and scope

The FREEPORT project aims to support electromobility transformation by addressing three key challenges faced by heavy-duty vehicle operators today: efficiency, safety and uptime.

The business value and use cases encompass monitoring electric components such as batteries and motors, developing foundations for using third-party services in edge devices, energy consumption predictions to optimise charging, and improving functional safety through continuous surveillance to alert the operators as needed. We expect to demonstrate edge data collection and processing for at least 20 vehicles, with the goal of connecting 50 heavy-duty electric trucks by the project's conclusion.

Planned approach and implementation

FREEPORT will develop cutting-edge data analytics capabilities on edge: novel real-time streaming anomaly detection algorithms tailored to the automotive sector, a versatile event-based data collection framework, a cybersecurity-aware architecture for real-time safety alerts, and comparative evaluation of state-of-the-art federated learning methods. The potential of edge processing and learning will be showcased using AI Sweden Edge Learning Lab to a broader audience.

To share knowledge regarding development and results, FREEPORT organized a number of workshops and events in several Swedish cities.

5 March - Workshop in Gothenburg (Lindholmen)

AI Sweden, Halmstad University and RISE invited to a workshop to present the project’s results regarding the electromobility transformation and the challenges faced by heavy vehicle operators.

The workshop focused on technical and practical challenges for edge processing in the automotive industry.

Sessions included demonstrations of, among others, synthetic data generation, anomaly detection with graphical neural networks and federated fault detection. Participants gained insights into data stream processing with SA Engine and edge processing for vehicle systems on a Volvo test truck, followed by in-depth discussions. The goal was to present the findings in an accessible way to relevant Swedish stakeholders in sectors such as automotive, healthcare and space.

The event was aimed at researchers, engineers and professionals in automotive and edge technology.

3June 2025 - Workshop in Boliden

The purpose of the demonstration at Boliden was to illustrate the practical application of Industry 4.0 concepts in a real-world industrial setting.

The demonstration highlighted how multiple services and data streams could be seamlessly connected to enable more informed decision-making and operational efficiency.

A key focus was on bridging edge and cloud systems to provide enhanced insights into production processes. The deployment included edge analytics integrated into a production vehicle, showcasing its ability to support personalized services such as monitoring and predicting the State of Charge (SOC), estimating energy consumption based on specific vehicle usage patterns, and detecting anomalies in safety-critical systems.

This comprehensive setup exemplifies how advanced digital solutions can drive innovation and robustness of operation in industrial operations.

18 June 2025 - Workshop in Halmstad