The current climate crisis calls for the use of all available technology to try to understand, model, predict and hopefully work towards its mitigation. Oceans play a key role in grasping the complex and intertwined processes that govern these phenomena. Oceans -and rivers- play a key role in regulating the planet’s climate, weather and ecology. Recent advances in computer sciences and applied mathematics, such as machine learning, artificial intelligence, scientific computation, among others, have produced a revolution in our capacity for understanding the emergence of patterns and dynamics in complex systems while at the same time the complexity of these problems pose significant challenges to computer science itself. The key factor deciding about the success of failure of the application of these methods is having sufficient and adequate data. Oceanographic vessels have been extensively used to gather this data. However, they have been shown to be insufficient because their high operation cost, the risks involved and their limited availability. Autonomous sailboats present themselves as a viable alternative. In principle, by relying on wind energy they could operate for indefinite periods being only limited by the effects of fouling and the wear and tear of materials. Recent results in the area of machine learning are especially suited to fill this gap. In particular, reinforcement learning (RL), transfer learning (TL) and autonomous learning (AL). The combination of those methods could overcome the need of programming particular controller for every boat as it would be capable of replicating at some degree, the learning process of human skippers and sailors.


General Goal

Design and evaluate a self-learning controller for autonomous sailboats relying on reinforcement learning, transfer learning and autonomous learning. This controller should be independent of the particular vessel and able to perform a multi-level process.

Specific Goals

  1. Create a sailboat simulator capable of representing the different combinations atmospheric and sea conditions, the characteristics and control surfaces of modern sailboats and the interactions between the boat, the wind and sea.
  2. A neural network-based surrogate of the simulator. Simulators have a high computational footprint and, therefore, unsuitable for training, re-iterative methods like those based on RL. However, a neural network can be trained based on simulator data to replicate its behavior at a fraction of the computational cost.
  3. Reinforcement learning-based models trained on the surrogate simulator.
  4. Understanding the potential of application of transfer learning methods to quickly adapt a pre-trained model to different vessels.
  5. Autonomous learning, adaptation and continuous learning for adapting model to meet changes in the vessel.
  6. Extending current state-of-the-art sailboat N-Boat prototype with an upgraded sensor package for obstacle avoidance and incoming wave detection.
  7. An IoT sensor pack and data fusion framework that ensembles a multi-level sailing strategy.



Universidade Federal Fluminense

  • Esteban Clua
  • José Viterbo Filho
  • Flavia Bernardini

Universidade Federal Rio Grande do Norte

  • Luiz Marcos G Goncalves
  • Bruno Marques Ferreira da Silva

Chile (Inria Chile)

  • Nayat Sánchez Pi
  • Luis Martí
  • Andrés Vignaga
  • Hugo Carrillo
  • Taco de Wolff
  • Hernán Lira
  • Jaime Aranda


Inria Project Team SCOOL

  • Philippe Preux
  • Odalric-Ambrym Maillard

Inria Project Team EVA

  • Thomas Watteyne
  • Malisa Vucinic

Uruguay (Universidad de la República)

  • Dr. Gonzalo Tejera, Universidad de la República

publications & events


join us

page not found