Academic research in the field of autonomous vehicles has reached high\npopularity in recent years related to several topics as sensor technologies,\nV2X communications, safety, security, decision making, control, and even legal\nand standardization rules. Besides classic control design approaches,\nArtificial Intelligence and Machine Learning methods are present in almost all\nof these fields. Another part of research focuses on different layers of Motion\nPlanning, such as strategic decisions, trajectory planning, and control. A wide\nrange of techniques in Machine Learning itself have been developed, and this\narticle describes one of these fields, Deep Reinforcement Learning (DRL). The\npaper provides insight into the hierarchical motion planning problem and\ndescribes the basics of DRL. The main elements of designing such a system are\nthe modeling of the environment, the modeling abstractions, the description of\nthe state and the perception models, the appropriate rewarding, and the\nrealization of the underlying neural network. The paper describes vehicle\nmodels, simulation possibilities and computational requirements. Strategic\ndecisions on different layers and the observation models, e.g., continuous and\ndiscrete state representations, grid-based, and camera-based solutions are\npresented. The paper surveys the state-of-art solutions systematized by the\ndifferent tasks and levels of autonomous driving, such as car-following,\nlane-keeping, trajectory following, merging, or driving in dense traffic.\nFinally, open questions and future challenges are discussed.\n