Automatic vehicle parking is an advanced technology that significantly improves road traffic efficiency and reduces congestion compared with manual driving. On the basis of machine learning technologies, in this study, we introduce a new deep-reinforcement-learning-based automatic vehicle parking algorithm in designated free spaces, using the framework of the actor–critic network. The proposed study’s principal features are as follows. (1) The provided system uses a single front camera of the vehicle as the primary sensor, simplifying the necessary vehicle sensing systems. A pretrained operator network receives the acquired picture data and uses it to create an action plan based on the vehicle’s present condition. (2) A reinforcement learning approach based on both critic and actor networks for training autonomous parking schemes is developed. (3) Formulation rules for an autonomous parking task-specific reinforcement learning environment, including creating parking environments, defining states, designing action spaces, and creating reward functions, are provided. It also incorporates a neural network structure suitable for automatic parking with a residual network module as its main feature to meet safety and reliability requirements.