Professor Tomforde, what is CAPTN X-Ferry all about?

Could you start by introducing yourself?
Prof. Dr. Sven Tomforde: My name is Sven Tomforde. I’m a professor of computer science at Kiel University (Christian-Albrechts-Universität zu Kiel, CAU), where I lead the Intelligent Systems research group.

You’re part of the CAPTN X-Ferry project. What’s it about?
X-Ferry focuses on the acceptance of autonomous ships. The key question is: how can we make passengers and crew feel comfortable using self-driving ferries? Our approach is to provide transparent explanations—so that the ferry itself communicates why it chooses a certain route, performs a maneuver, or behaves in an unusual way.

At the heart of the project are two challenges: identifying situations that require explanation, and tailoring those explanations to different user groups in order to build trust.

When a ship navigates on its own, it feels very different for passengers than when a captain is steering. Just like autonomous cars, future self-driving ferries will need systems that explain their actions.

What is your role in the project?
I serve as project coordinator. X-Ferry is a consortium, not just Kiel University—it brings together several academic institutions and industry partners, including the Vater Group, HH Vision, Addix, Anschütz, Flensburg University of Applied Sciences, and Kiel University of Applied Sciences with its research company, which owns the MS Wavelab research catamaran built as part of the CAPTN project. My job is to manage the project as a whole and make sure all partners collaborate effectively.

On the scientific side, my team works on detecting situations that need explanation, identifying their causes, and developing models that represent these causal links. We then translate this knowledge into data structures, which visualization partners turn into user-friendly displays.

Why is this especially important for ships?
The operating environment is very different compared to other autonomous vehicles. Take the Paris Metro, for instance. It runs through tunnels, and most passengers don’t even notice it’s fully automated. That’s a closed system.

A ferry, on the other hand, operates in open water. Passengers look out the windows and see everything—the traffic, the weather, the constant changes. A captain normally interprets this context and reacts based on his training. As a passenger, I rely on his expertise. Without a captain, it’s a very different experience. People are more likely to trust an autonomous ship if they understand why it makes certain decisions.

How might this look in practice?
We imagine different forms of communication for crew and passengers. In the passenger area, for example, a screen could show the planned route, nearby ships, fixed objects, and the estimated arrival time at the next stop. It might also display a simple reassurance—something like a green signal meaning “everything is fine.”

If the system detects something unusual—say, another vessel behaving unpredictably—it might adjust its route to increase safety. Because that’s a noticeable change, passengers need to be informed. The screen would then display what happened and why.

How do you determine what counts as normal or abnormal behavior?
We are computer scientists, not seafarers. From a technical perspective, we can identify deviations in a data-driven way by collecting and analyzing maneuver data. But from a nautical perspective, we also need to know what actually breaks rules or norms. That’s why we work closely with partners such as Kiel’s ferry and tug company (SFK). Together, we classify what they see as abnormal behavior and what causes it. This forms the basis for our causal models.

We test these ideas not on passenger ferries but on our MS Wavelab research catamaran. With it, we can run routes, simulate conditions, record data, and create scenarios for training our models. Having access to such a research vessel is a real advantage—it allows us to experiment in a real-world environment.

How exactly will the communication with passengers work?
That depends on the ship’s equipment. The current idea is a display showing expected behavior—navigation and context—and highlighting deviations and their causes. But the exact design hasn’t been finalized. The project is still in its early stages, and we’re refining the details step by step.

What results are you aiming for?
The project runs for three years, funded by the Federal Ministry for Economic Affairs and Climate Action as part of the Maritime Research Program. It started in September 2024.

Our goal is to develop a prototype system for Kiel Fjord that can demonstrate core scenarios. The system should be able to identify causes of maneuvers and explain them clearly.

Ideally, by the end of the project, we’ll be able to show this in practice: a ship approaches us, or we simulate an obstacle in the water. Our ferry adjusts its course, understands why it had to do so, and communicates that explanation to the passengers. The maneuvers themselves are being researched elsewhere; our focus is on explainability.

Is CAPTN the only research initiative working on this?
No. There are many international projects on autonomous shipping, including in Germany on inland waters. Each has a different focus, but the common theme is developing autonomous technology. Where we stand out is in working specifically on explainability—especially in passenger operations. That’s still unique to us.

Could the system be applied to other waters, such as Flensburg Fjord?
For now, we’re modeling conditions specific to Kiel—its traffic, normal behavior, and ferry operations. But the underlying technology will be adaptable. That said, our results won’t immediately translate into a finished product. Universities do research; industry partners will need to take the next steps to bring it to market.

I’m convinced that autonomous systems in any new environment will require additional pre-training. So yes, the technology can be transferred, but it will take further work. We’ll revisit that question once the CAPTN X-Ferry project wraps up in three years.

(Translated with the aid of AI.)