This paper presents a technique for enhancing the automatic driving of tunnel boring machines using an improved MLP algorithm. The project utilizes deep learning algorithms to perceive and recognize the surrounding environment of the tunnel boring machine as well as its excavation parameters. A training data-generated model is used to predict the current magnitude and latitude, longitude, and altitude information at five positions ahead of the tunnel boring machine during forward excavation. This information is used to calculate the thrust during forward excavation, and the excavation precision is determined through latitude, longitude, and altitude data. The thrust is adjusted accordingly to control the automatic driving of the tunnel boring machine. The paper begins by introducing the structure and working principles of the tunnel boring machine and highlights the challenges associated with manually operating traditional tunnel boring machines. Subsequently, this paper improves upon the fully connected neural network (MLP). Through the analysis of experimental data, it is demonstrated that the enhanced MLP algorithm provides the highest accuracy for tunnel boring machine automatic driving technology. Practical implementation has shown that this method enhances excavation efficiency and precision.