Emerging AI technology can create safer, more efficient maritime operations

SEA AI integrates thermal imaging and optical cameras to spot contacts that escape radar or AIS.
Developed by SandboxAQ, AQNav leverages AI, quantum magnetometers and Earth’s crustal magnetic field to provide unjammable, unspoofable, day/night, all-terrain positioning and navigation.
Developed by SandboxAQ, AQNav leverages AI, quantum magnetometers and Earth’s crustal magnetic field to provide unjammable, unspoofable, day/night, all-terrain positioning and navigation.

What exactly can artificial intelligence (AI) do for mariners? Well, potentially a lot. Like the emergence of electronics more than a century ago, it may not always be the answer, but in general AI can enhance many existing aspects of maritime practices and potentially revolutionize others.

For starters, the capabilities of AI can serve as a critical backstop for crew, reducing or eliminating the inevitable human errors that can cost money or even lives. It can frequently help spot anomalous actions or activities and remind humans to look more closely or even step in to prevent a dangerous development.

Whether it is optimizing power use in changing sea conditions, tweaking navigation to shave off time and costs, or simply reducing human labor needed to operate a vessel, AI has clear potential to contribute to operations on and off the water by improving efficiency.

Implicitly, improving efficiency and eliminating human error adds up to reduced costs. In fact. AI’s smarter and automated analysis means just about anything and everything involved in the maritime world can potentially be improved.

SEA AI’s advanced software and cameras can detect and identify floating hazards.
SEA AI’s advanced software and cameras can detect and identify floating hazards.

Indeed, a recent report by the engineering society IEEE found that “Artificial Intelligence is an essential tool for future technological development in the maritime industry. The maritime transport sectors are emerging with the rapid advancement of AI.”

For a real-world AI example available today, we talked to Tim Moore, East Coast general manager for Furuno USA. “I was in the Coast Guard and got to see what happens when things don’t go right,” Moore said. “I want mariners to be safe,” he added, “and that is the whole point of AI.”

Moore said users of Furuno products can leverage data such as chart depth, NXT solid-state radar and live AIS data to maneuver around land and anything that could cause a problem.

Furuno has worked on adding AI to its product line for about two years. If you let the software run by itself, noted Moore, it will do AI routing. And if you pair it with Furuno doppler radar, you can have it access more data and it will provide the safest route.

He explained that the advantages of AI are substantial. Previously, to navigate from point A to point B, you had to manually develop your own route and you had to look at each section of the route required to get there. “AI does that for you,” Moore said, so wherev­er your boat is, you can tap the app to fill in the route. “It will give you the option of creating the route manually or letting AI do it, you just have to set the end point.”

SEA AI integrates thermal imaging and optical cameras to spot contacts that escape radar or AIS.
SEA AI integrates thermal imaging and optical cameras to spot contacts that escape radar or AIS.

“If I want to leave the Chesapeake to go to the New York area, in about 10 seconds it will create a complete route, including the channels where I want to go,” Moore said. It can perform that task “right out of the box” but is also trainable for specific preferences, such draft requirements or how close or how far from land is preferred.

Additional recent additions to the Furuno AI product include a target analyzer that changes the display color of a given target that is overtaking you or crossing your bow, so it is clearer what to concentrate on.

Most recently, Furuno has added a risk visualizer that “goes to the next level,” according to Moore. Working with radar, it targets landforms, buoys and other vessels without need for user intervention. As a result, it can immediately indicate collision dangers. “If you are unfamiliar with an area, heading toward a buoy, it will put up a visual alarm and an audible alarm; it lets you take evasive action,” explained Moore. Well before serious danger develops, the system also marks objects such as boats and buoys that present a collision danger, with a special icon, he added.

Of course, Furuno isn’t the only company testing the AI waters. Another company harnessing AI is SEA AI, based in Brittany, France. According to the company, SEA AI technology detects floating objects early, using thermal and optical cameras to catch objects that escape conventional systems, such as radar or AIS. Unsignaled craft or other floating obstacles can also be detected with the system.

Solenn Gouerou, head of marketing and communication at the company, said it primarily uses “computer vision,” a branch of AI, in its product. “Our AI detects, identifies and alerts users to the presence of any floating hazards around the vessel,” Gouerou said.

“Our machine vision, based on AI, sees things that radar, AIS and the human eyes cannot; it enhances safety and reduces the risk of collision or piracy with a new level of situational awareness,” Gouerou explained.

The limits of the company’s AI are based on the performance metrics of the software ­— namely detection rate, false positive rate and recognition rate, Gouerou said. In other words, continued improvements in computing and software will only lead to even better accuracy and better rates of detection and identification as those limits are overcome and performance improves further.

Another take on harnessing AI comes from Luca Ferrara, the general manager of navigation at SandboxAQ, a Palo Alto-based tech company focused on AI and advanced computing solutions, as well as large quantitative models (LQMs). He sees one especially important application for the company’s technology: Serving as a backup for GPS, which, especially in conflict zones like the Black Sea and Middle East, has increasingly been subjected to jamming and spoofing. In some cases, ships have been indicated as being located at unlikely spots, such as the Cairo International Airport. But such problems could potentially develop anywhere in the world through third-party use of readily available electronics and hacking.

Ferrara said SandboxAQ can combine advanced sensors, such as magnetometers, attuned to the earth’s magnetic field, along with AI and machine learning, to generate a “clean signal,” allowing comparison with mapping tools and other navigational equipment or processes.

At the moment, the focus is primarily on aviation, where the stakes are especially high, but Ferrara said the technology could also be applied to ensure accurate and safe navigation on the water. Unsurprisingly, current customers include Airbus and the U.S. Air Force, where SandboxAQ technology is being paired with inertial navigation to “reduce uncertainty,” Ferrara explained.

A final perspective comes from Marek Poetzsch, senior product manager, maritime, at Spire Global, a space-to-cloud data and analytics company that harnesses nanosatellites to track maritime and aviation traffic and weather patterns. He said the company tracks AIS data daily from over 265,000 vessels that are collected or acquired from many sources. “The data we receive isn’t always as clean as customers expect, and recently it has included increasing amounts of falsified information, such as vessels spoofing their own positions to evade sanctions,” Poetzch said. “Machine learning can help us identify these anomalies and create data quality indicators,” he explained.

In addition, the company’s Port Congestion product, using similar kinds of analytics, can optimize supply chain planning and provides port insights. For example, its machine learning models can help forecast port congestion metrics and enable efficiency improvements through optimized planning.

Still, noted Poetzsch, AI isn’t magic and isn’t always a better answer.   

Ultimately, he said, buyers of products labeled as “AI” should focus on the value they receive and how products or data can help solve their specific problems.

“Keeping the focus on results is a simple but effective way to discern where AI truly adds value versus where it’s just a marketing label,” he said. “An AI label should not be a deciding factor for a purchasing decision, and equally, its absence should not deter someone from seeing the real value behind products,” Poetzsch said.

Nonetheless, AI can be an incredibly valuable complementary tool in certain cases, providing otherwise unattainable value. “Choosing the right tool for the job is key,” he said.