Welcome to the machines

According to a new study carried out within the Swedish Transport Administration’s industry program Sustainable Shipping, which is run by Lighthouse, navigators are positive to receive support from algorithms.

“The machines are our friends, without them no paradise”, sang Swedish artist Kjell Höglund in the 80s. An exaggeration, of course. At least if you think that paradise would consist of the machines doing all the work for us. Fully automated unmanned vessels, for example, will probably never be realized (well, there´s no demand) except on shorter routes across rivers and the like.

But whatever happens in the future, artificial intelligence will of course affect us, so also at sea. The question is how? The maritime industry is undergoing a transformation driven by digitalization and connectivity. The technological realization of Maritime Autonomous Surface Ships (MASS) presents significant challenges for the maritime human factors research community. These challenges relate to system design, human-automation interaction, stakeholder training, use and acceptance of new technology systems, and on a larger scale, how the regulatory framework, including the Collision Regulations (COLREGs) will be impacted within a MASS system. Decision support is the next step in the transformation towards more connected ships, however, such systems for navigation are largely unexplored from the users’ perspective.

In the project Operationalizing COLREGs in SMART ship navigation: Understanding the limitations of algorithm-based decision support systems in traffic situations, researchers from Chalmers and RISE have taken a closer look at how well an AI-based decision support system, developed by Wärtsilä, works. At Chalmers, 19 navigators got acquainted with the system in a simulator study that included three simple traffic scenarios.

“The response from the participants was very positive. It actually surprised me. I did not think they would appreciate getting different suggestions and brainstorming ideas with a computer. Perhaps Wärtsilä’s simple and user-friendly interface had some significance. Then you become more forgiving and tolerate that all proposals are not completely perfect”, says Reto Weber, a researcher at Chalmers, who led the project.

The decision support system studied in this project was developed by Wärtsilä and is called Advanced Intelligent Manoeuvring (AIM), aligning with “low-level automation” or Level 1 (out of a 4-level progression) of the technological realization of Maritime Autonomous Surface Ships (MASS)

“Anyone should not use it. You have to know all the rules. And things like experience, good seamanship, situational awareness and other non-technical skills that affect safe and efficient navigation are difficult to implement in algorithms.”

A navigator is required who is aware of COLREG – Rule 8 of the International Maritime Regulations, which deals with measures to avoid collisions. But what happens to him when the decision support systems are so easy to work with and do most things right? Is there not a risk that the navigator casually trusts the system and loses the feeling of navigation?

– Absolutely. There is a danger with this that we have not been able to investigate. I remember myself when I was at sea and the GPS came in the 90’s and how I was told: “do not trust that, you have to double check”. Nowadays, people completely trust GPS, not least when it comes to driving a car. It makes us lose our sense of place. And there is the same risk with decision support systems. Will navigators lose their knowledge of rules, analytical skills and so on?

It is already clear that the pre-study will be continued. In COLREG 2 a greater focus is placed on how well the algorithm in the decision support system works.

– In the tests we did, all ships had the same settings in the decision support system, but what what about, for example, if we have traffic situations with autonomous ships that all have different settings? That’s what we’re going to test now.

Prior to the continuation project, Wärtsila has sent over a new, really updated version of the system that will be run on three vessels in the simulator at Chalmers.

The pre-study Operationalizing COLREGs in SMART ship navigation: Understanding the limitations of algorithm-based has been authored by: Reto Weber, Katie Aylward, Scott MacKinnon, Monica Lundh – all from Chalmers – and Mikael Hägg, RISE

Read the report