Computer Vision Brings Vessel Autonomy One Step Closer

AI-assisted navigation has made leaps and bounds over the last several years, and has progressed to the point that automated ships can carry out berth-to-berth pilotage and long-distance voyages with little human input. TME recently sat down with a leader in this growing field, Orca AI CEO Yarden Gross, to learn more about how advanced automation is changing shipping - and just how far the change could go. Orca AI's technology provided the foundation for an autonomous transit pilot study aboard the Japanese container feeder Suzaku, which was completed successfully earlier this year.

TME: Can you tell us about the history of Orca AI's founding?

We started the company in 2018, but it goes back much further. I met my co-founder, Dor Raviv, about 15 years ago when we were serving in the Israeli Navy. He was leading a unit that was operating surface drones. Thanks to our service, we bring extensive experience in the maritime domain, both on the operational side and on the technology side. 

After the navy, we started working in the tech sector. In 2018, we saw a new opportunity in shipping, which is the backbone of the global supply chain. Shipping takes a conservative approach to new technology, but we saw clearly that something was changing. The main driver is ship connectivity. A large share of the world's fleet is now connected to the cloud, and more ships are installing VSAT service all the time. This enables shipowners to deploy new technology on board - for example, advanced services to improve fuel economy or safety of navigation. 

And this is where we come into the picture. Our vision is to enable the highest degree of safety for the ship and the fleet, creating a better solution for the crew on board and for fleet managers. 

For that purpose we have built a unique technology powered by computer vision sensors and AI algorithms. Radars are good at detecting targets farther away; our computer vision technology is very good at detecting smaller targets in close proximity, and with thermal cameras, it is equally good in darkness or fog. This technology empowers the crew to make safer and more efficient decisions in real time.

How did the system's development unfold?

In order to develop the technology, we had to collect a lot of data and then train the AI models to understand the context of the ship's surroundings, detect risks and classify them. Our algorithms reflect the process that a human watchstander carries out today - visually looking out the window, identifying targets, calculating CPA and evaluating risk. 

This time-honored method works well when you have three or five targets, but beyond that it becomes a challenge. Our system really shines when there are dozens of targets at once. Orca AI mimics the watchstander's process, but it does it in milliseconds and cross-references information from its cameras with data from the AIS, radar and other inputs. Then it ranks all the risks based on target size, type, proximity and other factors. 

Orca AI enables the ship's officers to gain a deeper, more immediate understanding of their surroundings. In any weather, day or night, in any waters, Orca AI delivers instant alerts and prioritization of risks. The Orca AI system is like an additional look-out on the bridge, and it is complementary to traditional navigation systems, which have some inherent limitations. A radar screen gets very cluttered in congested waters, making it hard to follow, while small boats or buoys can also be missed by radar due to the noise generated by the waves. The ECDIS is limited by the fact that it can only present targets that have an AIS transciever or a radar signature.

Orca AI helps officers trace small ships or boats at a distance, day and night, and to prioritize high-risk targets in congested waters. It reduces the workload for the officer on watch and provides them with additional information. We've matured it into a robust system and have deployed it on a large number of vessels, including the ships of blue-chip operators like Maran Tankers, and it has demonstrated a very high level of performance. 

Can you tell us about any lessons learned from the process? 

The key is to work closely with your customers. If a ship has been operating in a certain manner for decades, it will take time for the crew to adapt to the new technology. Once they adapt to it, they can gain great value from it, but it takes time for that to happen. You need to work closely with them, hear their feedback and understand what you need to improve in order to make the system easy to use for their day-to-day operations. Since Dor and I have served in the navy and have experienced what it is like to operate a ship, that understanding is part of the DNA of our company. 

Can this also help with safety management?

Over the course of developing our solution, we grew to understand that our data is gold for fleet managers. As the voyage progresses, we collect data from the ship and send it to the cloud for analysis. By comparing Orca AI's cloud-based records with the navigation standards set out in the customer's SMS, we can detect procedural violations. We provide that information to the fleet manager so that they can extract insights into the safety of their entire fleet.

This helps the operator understand what is happening with their ships much better, with the objective of improving procedures and crew training. We can spot patterns, like which ships in the fleet are more prone to be involved in an incident, or which time of day sees the most risky behavior (the 0400 watch turnover).

Navigation safety test results with Maran Tankers:

• 75+ voyages
• 1500+ days of sailing
• 60+ gigabytes of data
• 400,000+ nautical miles sailed
• 33% reduction in close encounter events
• 27% increase in minimum distance from other vessels

What does the future look like for Orca AI? 

Lately, we've begun engaging in strategic partnerships with leading shipowners on the development of fully autonomous vessels, like the Suzaku autonomous container ship project in Japan. I think that there is reason to expect massive adoption across the board at many shipping companies. It solves major problems for them and they see that it is something that they can't miss. As this process unfolds, we are seeing growth for our company and our customer base.

Control room for the Suzaku autonomous ship trial (Orca AI)

This content is sponsored by Orca AI.