Reinforcement Learning (RL): trial-and-error learning; RL algorithms solve optimal control problems using data gathered online and without requiring a dynamic model (although offline, model-based methods can be used as well).