Project Summary
Shipping accounts for 2.9% of global CO2 emissions, with a potential increase of 44% by 2050. To address this issue, shipping industry and the European Commission promote an increasing use of wind-assisted propulsion solutions, such as rotor sails and inflatable sails. However, there is limited research on real-time adaptive control strategies for sail propulsion. Our project leverages data science and AI to optimize sail adaptation and autonomous navigation for maritime transport.
The first objective is to develop feedback control methods for sail and rudder operations based on environmental data (wind direction, sail deformation, ship dynamics). The intrinsic complexity of flow dynamics will be addressed by guiding AI with system analysis in the infinite-dimensional framework. By integrating AI algorithms, such as reinforcement learning and neural networks, with real-time data from sensors and conservation laws from physics, we aim to dynamically adjust the sails to maximize propulsion efficiency and minimize energy consumption. Methods exploiting Navier-Stokes equations and CFD will be used to optimize sail trim in varying wind conditions, ensuring optimal lift and drag forces for propulsion.
The second objective is to enhance navigation and path planning for autonomous surface vehicles. Using real-time data from GPS, LiDAR, cameras, and ocean models, we will develop intelligent guidance, navigation, and control systems. Path planning methods will be developed to find optimal, energy-efficient routes that adapt to wind, waves, and ocean currents. AI-based control strategies, such as model-free reinforcement learning, will help the vessel adapt to changing conditions and avoid obstacles in complex marine environments.
AutoSail aims to improve the energy efficiency of sailing ships and reduce emissions by developping safe, high-performance control and AI methods for wind-assisted propulsion (WAP), experimental validation, and autonomous sailing. It addresses modeling & FSI, control for infinite-dimensional systems, and AI-enabled autonomy. It is a 3 years project financed by ANR and NSERC.
Technical Overview
WP1 — Physical Modeling, Machine Learning & Experimental Data
WP1 establishes the core modeling and data foundations required for accurate prediction and control of wind-assisted propulsion. It combines high-fidelity CFD and fluid–structure interaction modeling with machine-learning surrogate models and grey-box approaches to capture nonlinear aerodynamics, sail deformation, and interaction with rigging.
Extensive experimental campaigns on the MiniJI, Mouette 19 and DragonFlite 95 platforms provide the datasets needed to validate simulations and train adaptive, real-time estimators. The final outcome is a suite of physical, ML-based, and hybrid models—along with comprehensive datasets—supporting both offline numerical twins and online control design.
WP2 — Control of Fluid–Structure Interaction for WAP
WP2 develops advanced control strategies that explicitly integrate the coupled dynamics of wind flow, sail deformation, and vessel motion. Methods include boundary control of nonlinear PDEs, hybrid controllers switching across flow regimes, adaptive control under actuator saturation, and safety-critical techniques using Lyapunov and barrier functions.
Reinforcement learning and AI-based components are combined with formal control guarantees to ensure reliability and robustness. The resulting controllers are evaluated on fully instrumented sailing platforms, paving the way for high-performance, autonomous WAP systems.
WP3 — Autonomous Sailing & Route Optimization
WP3 focuses on autonomous navigation for WAP-enabled ASVs, developing dynamic routing strategies that adapt to wind, waves and ocean currents in real time. It integrates nonlinear ship dynamics with environmental forecasts using model predictive control, reinforcement learning, reduced-order modeling, and data assimilation techniques.
Hybrid AI–model-based decision systems ensure physically consistent and robust planning under uncertainty, while large-scale simulations and field experiments on the MiniJI and DragonFlite 95 platforms validate the approach. The final objective is an integrated navigation framework enabling efficient, adaptive and autonomous WAP-based sailing.