Reinforcement learning is an approach to trial and error learning where a creature's actions are guided by a class of signals called rewards. These signals have been used in engineered systems and computer programs to equip these systems with goals, and to use the goals to guide learning.