Abstract: With the increase of coal mining depth, roof accidents have become one of the main hidden dangers threatening mine safety. Traditional roof pressure monitoring technology generally relies on ground central stations for data processing, which has problems such as large data transmission delay, high bandwidth peak occupancy, and weak real-time warning capabilities, making it difficult to meet the urgent needs of dynamic monitoring in complex underground environments. Therefore, we propose a roof pressure monitoring technology based on edge computing. By sinking the computing power to the edge nodes of the network, the local real-time processing and intelligent analysis of pressure data are realized, and the response speed and reliability of the monitoring system are significantly improved. Firstly, for the roof pressure monitoring scenario, an edge computing node with multi-source heterogeneous data fusion capability is designed. The node integrates high-precision pressure sensors, hydraulic support piston expansion and contraction, and environmental parameter acquisition modules, and combines with adaptive filtering algorithms and lightweight anomaly detection models, to complete data preprocessing and preliminary feature extraction, effectively reducing the amount of redundant data uploaded. Secondly, a three-level distributed monitoring architecture of “perception layer-edge layer-cloud platform” has been constructed. The edge layer achieves dynamic trend prediction and risk grading warning through collaborative computing, while the cloud platform is responsible for global data storage and model optimization iteration, balancing real-time and long-term data analysis needs. In addition, a dynamic scheduling strategy for edge resources based on deep reinforcement learning is proposed to optimize computing task allocation and energy management, ensuring the long-term stable operation of edge nodes in complex underground environments. The experimental results show that compared with the traditional cloud computing mode, the edge computing scheme has the shortest local transmission path and local data computing capability. The wireless transmission power of the pressure sensor is reduced from 10 dbm to 4 dbm, significantly reducing the energy consumption of the equipment. At the same time, the success rate of wireless data transmission is increased to 100%, and the warning time of dangerous states is increased from 8.2 min to 3.5 min, which improves the real-time warning capability of roof pressure monitoring.