Reinforcement learning is a machine learning paradigm focused on sequential decision-making, in which an autonomous agent learns optimal behavior by interacting with a dynamic environment to maximize cumulative reward signals.