Reinforcement learning is the paradigm for intelligent agents to make sequential decision-making through interaction with the environment, based on the model of Markov decision process. There are two types of reinforcement learning, i.e., model-based and model-free. The former means that all elements in the model are known so that it can employ planning method, while the latter refers that there are uncertain elements in the model; therefore, it uses the learning approach. This chapter starts with an overview of the development stages of reinforcement learning. Second, we give the definition and discuss the related elements of reinforcement learning. We next focus on the one of model-based method called dynamic programming and the two of model-free methods, namely Monte Carlo learning and temporal difference (TD) learning. Then we study the eligibility trace mechanisms in reinforcement learning. After that, we introduce deep reinforcement learning. The comparison of the aforementioned reinforcement learning methods is made finally.