A presentation on data-efficient reinforcement learning methods for autonomous robots, covering probabilistic models, control constraints, and meta learning, by Marc Deisenroth at MLAI 2019.