This paper provides an introduction to deep reinforcement learning (deep RL), a field that combines reinforcement learning (RL) with deep learning. Deep RL has enabled machines to solve complex decision-making tasks that were previously out of reach. The paper covers the fundamentals of deep RL, including machine learning and deep learning concepts, reinforcement learning, value-based methods, policy gradient methods, model-based methods, generalization, online challenges, benchmarking, and applications beyond Markov Decision Processes (MDPs). It also discusses the successes and challenges of deep RL, its relationship with neuroscience, and future directions. The paper emphasizes the importance of generalization and practical applications of deep RL in domains such as healthcare, robotics, finance, and smart grids. It outlines various deep RL algorithms, including Q-learning, deep Q-networks (DQN), double DQN, dueling network architecture, and distributional DQN. The paper also addresses challenges in applying deep RL, such as exploration-exploitation dilemma and the need for efficient learning algorithms. It concludes with a discussion on the future of deep RL and its potential societal impact.This paper provides an introduction to deep reinforcement learning (deep RL), a field that combines reinforcement learning (RL) with deep learning. Deep RL has enabled machines to solve complex decision-making tasks that were previously out of reach. The paper covers the fundamentals of deep RL, including machine learning and deep learning concepts, reinforcement learning, value-based methods, policy gradient methods, model-based methods, generalization, online challenges, benchmarking, and applications beyond Markov Decision Processes (MDPs). It also discusses the successes and challenges of deep RL, its relationship with neuroscience, and future directions. The paper emphasizes the importance of generalization and practical applications of deep RL in domains such as healthcare, robotics, finance, and smart grids. It outlines various deep RL algorithms, including Q-learning, deep Q-networks (DQN), double DQN, dueling network architecture, and distributional DQN. The paper also addresses challenges in applying deep RL, such as exploration-exploitation dilemma and the need for efficient learning algorithms. It concludes with a discussion on the future of deep RL and its potential societal impact.