Coordinated by Ronan Fablet, professor at IMT Atlantique, the Oceanix "chaire" (research consortium) uses artificial intelligence methods to study ocean dynamics. The key is to gain a better understanding of oceanographic conditions, with numerous applications for weather forecasting, ship routing and maritime surveillance. But more than that, it is a paradigm shift for ocean sciences... and perhaps even for many other disciplines.
With its two sites in Brest and Nantes, IMT Atlantique has a front seat from which to observe, study and understand the marine world. A world whose importance is vital for humanity: the sea plays a major role in climate, fisheries and mining resources, trade, energy, health and even geostrategic balance.
It is in this context that the Oceanix "chaire" (research consortium), supported by IMT Atlantique, is situated. Created at the end of 2019 for a period of 5 years and with an endowment of approximately 2 million euros, it is led by Ronan Fablet, professor in the Mathematical and Electrical Engineering (MEE) department. It brings together a number of partners: academic institutions (ENSTA, Ecole navale, University of Brest), sectoral institutions (Ifremer, CNRS, CNES, ESA, etc.), and industrialists such as Naval Group and CLS, as well as several SMEs. For its part, Microsoft is providing computing resources.
Ronan Fablet
The Oceanix chaire is positioned at the meeting point of two disciplines: ocean sciences and artificial intelligence. Its ambition is to use AI methods (data science, deep learning, etc.) to better analyse and predict ocean dynamics. Until now, we were content to use high-performance computing," explains Ronan Fablet. But with AI it is possible to go much further in the exploitation of observation data, to reconstruct certain missing data and to develop new types of modeling.
A gain of around 40-50%
One of the objectives of the Chaire is also to make better use of the available data. "Systems based on human operators can only process about 1% of the AIS (*) data provided by ships. Our AI-based algorithms can process the entire data stream and offer new tools for raising alarms for operators," he points out. Here again, significant progress is possible.
This original approach makes it possible, for example, to 'reconstruct' the evolution of currents with an increased accuracy of 40 to 50% compared to 'classic' solutions. This will also enable the simulation and anticipation of climatic events - and in particular extreme phenomena such as heat waves. Similarly, it will be possible to greatly improve the routing of ships, which is also linked to variations in marine currents. The Chaire team also plans to demonstrate an AI-based method applied to currents, in collaboration with Ifremer.
At the same time, the observation tools themselves are improving in performance. New satellite sensors are providing more and better data. Drifting buoys and other instruments make it possible to observe the ocean depths. In other words, the work of the Chaire opens up a wide range of possibilities, whether it be for monitoring maritime traffic (detection of 'illegal' degassing, for example), for monitoring fishing or for weather forecasting, not to mention research into marine energy.
A real scientific breakthrough in the making
An initial test on the detection of 'abnormal' ship behaviour (e.g. change of route or sudden stop) using systemic analysis based on learning has produced promising results. Another avenue being explored by the Chaire is the possibility of using modeling and prediction tools on board ships in the future. In ocean sciences, the situation will change radically," says Ronan Fablet. We will have more data and we will benefit from the contributions of artificial intelligence. We will thus be able to work on biochemistry, marine biology, etc. All this opens up enormous prospects. A real breakthrough is coming.
Better still: the same modeling tools could be used in a host of disciplines: meteorology, materials science, chemistry, acoustics, etc. The Chaire has already established collaborations on fluid mechanics and turbulence. Those at the head of the Chaire see even further ahead: they hope to one day identify new models for physical or chemical processes, directly from the data. "In the ocean, with the emergence of new satellite and in situ observation methods, we could identify the right models, but also find the equations that explain phenomena that are currently poorly represented," explains Ronan Fablet.
The Oceanix Chaire is also involved in training. It has set up a course combining artificial intelligence and marine sciences. Around 60 participants of all profiles (students, but also working professionals, PhD students and post-doctoral fellows) follow this course, mostly remotely. A double degree between IMT Atlantique and the University of Brest, combining ocean sciences with AI and data sciences, is also in preparation. Already, IMT Atlantique, the University of Brest and ENSTA Bretagne are collaborating on a joint training programme in marine data science as part of the 'Marine and Coastal Science' master's degree at the university.
Other initiatives concern secondary education. Following a meeting with a high school class in Carhaix, the chaire's managers are planning to put on a show (video or play) on issues related to digital technology and the ocean. Popularisation initiatives on the climate are also planned.
(*) The AIS (Automatic Identification System) is a system for exchanging data between ships. Mandatory since 2004, it aims to enhance safety and improve fleet and navigation management.
by Pierre-Hervé VAILLANT
