Predicting the Speed of a Wave-Propelled Autonomous Surface Vehicle Using Metocean Forecasts

Abstract

Wave-propelled autonomous surface vehicles are becoming increasingly popular in oceanographic research due to their ability to provide persistent observations of the ocean environment. This type of vehicles are propelled purely by environmental forces which greatly enhances endurance. However, unlike vehicles with motorized propulsion, the velocity of wave-propelled autonomous surface vehicles cannot be controlled actively, making mission planning routines depend on predictions. In this work, we compare two methods based on linear regression and Gaussian process regression for predicting the speed of the wave-propelled autonomous surface vehicle AutoNaut. The regression models are trained using onboard measurements gathered during field operations, while the predictions are performed using metocean (wind, wave and current) forecasts. The Gaussian process regression model proves to be the most accurate way of predicting the speed of the vehicle.