Of the 510 million km2 defining the surface of the Earth, 70% corresponds to the seas and oceans. This vast expanse of water is teeming with marine animals and underwater plants, but also with other objects resulting from pollution or human activity. Led by ProMare, an organization that promotes marine research and exploration, the Mayflower Autonomous Ship (MAS) project aims to design an autonomous system to help researchers collect valuable data from the oceans and seas.
A project supported by many institutions
This project is the result of a vast collaboration between ProMare, IBM and about forty partners in different sectors, here is a non-exhaustive list:
- For the design and construction of the system: MSubs, Aluship, Marine AI and GUARDIAN by Marine AI.
- For communication: Thales and Iridium
- For services and equipment: NVIDIA, Red Hat and Vodafone
- For research: The universities of Plymouth, Newcastle, Liverpool, Exeter and many laboratories associated with these universities in the field of AI and/or oceanography.
Several experts have been called upon, including Rosie Lickorish, an emerging technology specialist at IBM, and Ollie Thompson, a robotics engineer. The project is led by Brett Phaneuf, founder of ProMare.
The goal of the project is to design a fully autonomous, solar-powered boat that can collect ocean data on issues such as global warming, pollution and its impact on marine life and the movement of particular species. The partners also hope that this boat will be able to carry out long expeditions and cross an entire ocean without any technical problems.
A virtual captain to steer an autonomous boat
In order to build this autonomous boat, the Mayflower 400, artificial intelligence technologies developed by IBM were used to create a “virtual captain”. This system was designed using machine learning and supervised learning. TheIBM Power AC922 powered by IBM Power9 processors and NVIDIA V100 Tensor Core GPUs were used.
This artificial intelligence was trained through thousands of photographs to learn how to identify maritime obstacles such as rocks, other boats sailing on the seas or marine animals venturing on the surface of the water. The “AI Captain” constantly reassesses its environment and automatically adapts its trajectory according to the obstacles it encounters.
This “virtual captain” will try to take on the role of a real captain: 6 cameras associated with computer vision systems will scan the horizon. The IBM Visual Insights computer vision software and the IBM Operational Decision Manager automation software are used in this context.
Researchers will be able to program the model to focus on specific objects, plants or animals and collect specific data based on their movement, sound or number using the 30 on-board sensors and the trained system.
Edge and cloud computing to complement machine learning for scientific experiments
Several devices provide the computing power needed for the ship to operate independently. The ship has about 15 edge devices and uses the cloud whenever the connection allows, to perform updates and download data. To power the onboard computer systems and engines, the trimaran has lithium-ion-phosphate batteries connected to solar panels.
The program also takes into account the weather parameters to prevent potential storms that could affect the boat. However, the Mayflower 400 has not yet been taken out to sea in weather conditions associated with a severe storm. The ship has therefore been trained to cope with 50-metre waves using a simulator.
The Mayflower 400 itself is fifteen meters long, 6.2 meters wide and weighs five tons. It will use all the AI technology on board to carry out scientific experiments under the supervision of the research team, 24 hours a day:
- Oceanography: Detection of marine mammals, analysis of their population distribution.
- Chemistry: Analysis of the chemical composition of water
- Global warming: Measure sea level
- Pollution: Collecting samples to analyze the amount of microplastics in the water
The trimaran is scheduled to depart for the United States around May 15, weather permitting.
Translated from Mayflower Autonomous Ship, un bateau autonome pour mener des expériences scientifiques en haute mer
