Reinforcement-learning robotic sailboats: simulator and preliminary results