Robust Edge AI for Safer Trucks, 700 meters underground.

After two years of collaboration, the FREEPORT project, funded by Vinnova, has reached its conclusion. And the ambition has been bold: to move AI out of the lab and into the field — onto the machines, into the mines, and closer to the decisions that matter.

The project brought together RISE, Boliden, Volvo, Stream Analyze, AI Sweden and Halmstad University, along with researchers and engineers exploring one shared goal: to understand how AI, edge computing, and cyber-secure data communication can be enabled in industrial operations.

From use cases to real impact

Throughout the project, the partners explored several industrial AI applications, all aimed at improving efficiency, sustainability, and safety in production environments.

The work focused on three main areas: predictive maintenance, enabling early detection of anomalies before failures occur; energy prediction, understanding how energy is consumed and recovered; and AI on the edge — running machine-learning models directly on vehicles and systems, even in extreme industrial settings.

Each use case demonstrates how moving intelligence closer to the data source can shorten feedback loops and enable new levels of operational awareness.

A demo 700 meters below ground

On June 3, 2025, the team gathered in Boliden, Västerbotten, to put their ideas into action. An electric mining truck began its descent — nearly 700 meters below ground — while AI algorithms onboard monitored how the vehicle consumed electricity on the way down, how it regenerated energy on the way up, and how driving style affected overall efficiency.

Senior AI researcher Sepideh Pashami from RISE explains that the team has been testing AI functionality by following the truck all the way down into the mine. They’ve been able to measure how the battery behaves, how energy is consumed and recovered, and how driving style impacts performance.

What makes this setup unique is that the AI doesn’t run in a distant cloud. It operates directly on the truck, and on a Volvo-installed logger, using Stream Analyze software to process the data on the edge — in real time, underground.

How the demo worked

The demonstration combined live data, on-board analytics, and real-time visualization. Data streams were exported from the edge devices to external systems using MQTT, a lightweight communication protocol widely used in industrial IoT.

This setup enabled seamless integration between Volvo’s systems and Boliden’s operations center, allowing engineers at both sites to monitor the same data live. The data flow — what was shared and how — was securely controlled by Volvo analysts, ensuring safe and traceable information handling.

In the Boliden control room, researchers and engineers from RISE, Volvo, Stream Analyze, Halmstad university and Boliden could see the data visualized in real time and discuss what it revealed about the truck’s performance underground.

As Erik Zetler from Stream Analyze describes it:

“We execute models specified by our users on the edge fleet.  The results of these models are streamed to external systems and integrated with various environments. In this case, we’re running models specified by Volvo on their truck, and the operations center at Boliden is feeding on that same result data stream.”

It was, in essence, a complete AI infrastructure in action — from edge to cloud, from code to metal, and from research to reality.

Insights from the depths

The demonstration validated that AI on the edge works — even in one of the harshest environments imaginable. It showed that advanced analytics and machine-learning models can be built, tested, and deployed on real industrial vehicles without long lead times.

The team learned how to deploy and iterate models directly on the edge, cutting development cycles from months to minutes. They also developed methods for predicting energy consumption and battery charge state dynamically based on real driving patterns — and for integrating real-time analytics across multiple industrial systems securely and reliably.

The project also underlined the importance of standards and interoperability. Discussions during the demo highlighted Sparkplug B for MQTT as a promising way to harmonize data structures across partners — paving the way for scalable industrial applications.

Slawomir Nowaczyk from Halmstad University noted:

“It’s one thing to make a model work on your desktop — it’s another to make it run reliably on a fleet of trucks, deep underground.”

Beyond technical achievement, the results point toward significant long-term benefits. By understanding how electric trucks consume and regenerate energy underground, the team uncovered ways to fine-tune charging cycles and driving patterns — potentially charging can plan and regenerative energy can be considered to avoid overcharging.

Toward an intelligent future

With the project now complete, the FREEPORT partners see vast opportunities ahead. The methods and tools developed for this demo can now be extended to other industrial domains — from construction and logistics to renewable energy and manufacturing.

They also open the door to smarter, greener, and safer operations where data, connectivity, and intelligence work hand in hand. Future explorations will focus on deepening edge integration, refining data standards, and scaling real-time AI analytics across entire production ecosystems.

Beyond the technical achievements, the partners highlight the collaborative spirit that made FREEPORT stand out.

Dennis Forslund, Boliden representative, describes it as:

“A true team effort — blending engineering and analytics with real-world testing and making every step a joy.”

From Volvo’s perspective, the collaboration proved how industry and research can push innovation together:

“Edge computing and AI analytics are clearly part of the future. Seeing it live on our customer trucks was a great experience — collaborative, innovative, and impactful.”

For RISE, FREEPORT marks an important milestone in applied AI. Sepideh Pashami reflects:

“This project has shown what happens when we bridge research and reality. When AI moves closer to the machines, innovation moves faster.”

The FREEPORT project may have come to an end — but its results signal a beginning. Combining electrification, AI, and secure edge computing could enable heavy-industry operators like Boliden to dramatically cut emissions while improving safety and uptime — a model for future smart, sustainable mining.

For more information about the demo or continued work in this field, contact Sepideh Pashami.