Controlled gliding and perching through deep-reinforcement-learning
Guido Novati, L. Mahadevan, Petros Koumoutsakos · Physical Review Fluids · 2019
Gliding is an energetically efficient mode of transportation. It is shown that reinforcement learning identifies gliding strategies with minimum energy expenditure and fastest time of arrival, with better performance than model-based optimal control, while being robust to perturbations.