Model-Based Prediction for Energy-Efficient Route Optimization in Autonomous Sailboats

As autonomous sailboats use wind for propulsion, they can operate for long durations due to their low energy consumption. Energy efficiency is still important for mission endurance, as sensors, computation, communication, and the actuators for sail and rudder draw energy constantly. One difficulty is that sailboats cannot sail directly in every direction. Close to the wind there is a no-go zone that can only be crossed by tacking. But even outside the no-go zone, the energy needed to cover a given distance depends strongly on the boat’s heading relative to the wind. To address this, we use an energy polar to model the boat’s energy consumption. We fit the polar to real sensor data from our sailing platform and use it to describe the energy per meter as a function of wind angle and wind speed. This model serves as the cost function in a Hybrid A* planner, guided by a heuristic derived from the convex hull of the energy polar, which provides a lower bound on the remaining energy cost. This enables the planner to find energy-efficient paths that exploit favorable wind angles and avoid costly headings.