16-06-2022 | By Robin Mitchell
Recently, the Mayflower AI boat developed by IBM researchers has finally crossed the Atlantic Ocean demonstrating how AI could play a role in autonomous vessels. What challenges do AI boats face, what features does the Mayflower integrate, and could AI be essential in future boats?
In the field of autonomous transport, it would seem that engineers are primarily focused on automobiles, planes, and drones. But one area that seems to get little attention is ships which are unusual when considering that the vast majority of goods around the world are shipped over oceans.
While some may believe that an AI ship would be far easier to develop than an AI car, the truth is that the two cannot be compared due to fundamental differences between driving and sailing. Firstly, the lack of road signs, speed restrictions, and winding roads means that boats have a large amount of freedom when navigating, and mistakes in direction or speed are tolerated. Secondly, the lack of sudden collision risks (i.e., crossing pedestrians and overtaking traffic) means that an AI system sailing a boat is less reliant on real-time decision-making.
However, that isn’t to say that designing an AI boat is easy. For one, sailing can be complex, depending on the method of propulsion. The use of sails requires careful control over the position and orientation to maximise thrust, while engine-driven boats can quickly encounter engine failure in the harsh salty environment. Furthermore, an autonomous car that breakdown will have access to roadside assistance, but a vessel that fails in the ocean can be a thousand miles from civilisation. Worse, the weather conditions found at sea can be treacherous with 20-meter-high waves, winds exceeding 100mph, and all kinds of turbulent currents that make it nearly impossible to maintain a direction of travel.
Finally, when a ship makes port, it has to position itself carefully due to the close proximity of other ships in the port and those docked. Additionally, movement in water is not the same as on land as water currents and momentum can rock ships in all directions. Thus, an AI is not driving a boat like it would a car but instead observing the environmental conditions to determine how to make the boat move. Essentially, the AI must have the same intelligence and sailing knowledge as an experienced sailor.
For the past few years, IBM has been developing an AI boat designed to autonomously navigate the ocean, and its first planned crossing between Plymouth (UK) and Massachusetts (US) earned it the name Mayflower (after the famous crossing of the boat carrying the Pilgrims to the US 400 years ago). After several years of delays due to technical difficulties, the boat was finally launched on the 29th of April with an expected journey time of three weeks.
However, during the journey, a technical fault resulted in the boat making port in Canada instead of the US. Despite this setback, the vehicle could still navigate its way safely, demonstrating how AI can be used to sail ships. The 15-meter-long boat utilises solar panels as a primary energy source with a diesel engine backup and has a maximum speed of 20km/h (10 knots). Six cameras provide the ship with a large degree of visual awareness, while 30 onboard sensors provide information on temperature, wind speed, humidity, air pressure, forces, water-depth, GPS, weather, and RADAR.
The onboard AI system was trained using IBM’s computer vision technology to recognise ships, debris, bridges, and other potential hazards that ships can face, and the use of over one million images ensures that the AI can identify a multitude of variations.
Even though the Mayflower failed to reach its intended target, its voyage still demonstrated the capability of AI in sailing. While there may be some out there who are concerned that AI will replace traditional sailors, it is more likely that AI systems will be used as co-sailors to the captain of a ship by helping to identify hazards and make decisions under challenging conditions.
Additionally, the use of entirely autonomous drone ships could provide low-cost shipping solutions to nations with a large number of islands (such as the Philippines). As these ships have a small cargo capacity of 1000KG, they could be ideal for getting essential medical supplies and resources to those in difficult-to-reach places where drones would be unsuitable due to a lack of airport infrastructure. Furthermore, using autonomous boats is significantly safer than using flying drones that need to avoid collisions with other aircraft as well as buildings and overhead cables.