This paper provides an overview of reinforcement learning (RL) and its applications in astronomy. RL, a mechanism that enables artificial intelligence to perform tasks through repeated attempts and feedback, has seen significant advancements in various fields, including computer games, robotics, and data processing. In astronomy, RL can be applied to telescope automation, adaptive optics control, observation scheduling, and data processing pipelines. The paper begins with a theoretical introduction to RL, covering the state, action, and reward concepts, as well as Markov decision processes (MDPs). It then delves into deep RL algorithms, discussing model-free and model-based approaches. Model-free algorithms, such as Q-learning and actor-critic methods, are detailed, including their implementation using deep neural networks (DNNs). Model-based RL, which uses a representative model of the environment to generate training data, is also explored, with a focus on probabilistic ensemble models and hint-assisted RL. Finally, the paper discusses practical aspects of applying RL in astronomy, including the use of deep learning frameworks and environment collections. It highlights the importance of understanding the problem domain and the state representation for effective RL implementation. The paper concludes by emphasizing the potential of RL to enhance various operational aspects of astronomy, from planning and scheduling to data processing.This paper provides an overview of reinforcement learning (RL) and its applications in astronomy. RL, a mechanism that enables artificial intelligence to perform tasks through repeated attempts and feedback, has seen significant advancements in various fields, including computer games, robotics, and data processing. In astronomy, RL can be applied to telescope automation, adaptive optics control, observation scheduling, and data processing pipelines. The paper begins with a theoretical introduction to RL, covering the state, action, and reward concepts, as well as Markov decision processes (MDPs). It then delves into deep RL algorithms, discussing model-free and model-based approaches. Model-free algorithms, such as Q-learning and actor-critic methods, are detailed, including their implementation using deep neural networks (DNNs). Model-based RL, which uses a representative model of the environment to generate training data, is also explored, with a focus on probabilistic ensemble models and hint-assisted RL. Finally, the paper discusses practical aspects of applying RL in astronomy, including the use of deep learning frameworks and environment collections. It highlights the importance of understanding the problem domain and the state representation for effective RL implementation. The paper concludes by emphasizing the potential of RL to enhance various operational aspects of astronomy, from planning and scheduling to data processing.