Abstract: During the underground construction of gas extraction hole in coal mine, the feeding system of drilling robot is used for pressurization, decompression and feeding in drilling process. In view of the problem that the uncertain disturbance of complex strata affects the working performance of drilling robot and the quality and efficiency of drilling construction, the structural composition and drilling construction technology of drilling robot were analyzed at first. The mathematical model of pressure reducing valve control was established according to the working principle of its feed system, and the mapping relationship between the control input, the electromagnet current, and the control output, the outlet pressure of pressure reducing valve, was obtained. Then, control modeling was conducted for the whole feed system on the basis of defining the driving mode of feed force. Secondly, the feed force tracking control system of the drilling robot was designed. Meanwhile, the Luenburger full-state observer was used to reconstruct the state of the controlled object, and the control structure based on equivalent-input-disturbance (EID) estimation and compensation was established. Besides, the gain matrix was designed for the state feedback controller, state observer and disturbance estimator, to realize the stability of the feed force closed-loop control system with satisfactory tracking and disturbance suppression performance. Finally, the numerical simulation model was built using the Matlab software. Specifically, the simulation research was carried out with the measured drilling data signal of the feed system in the actual drilling construction of a coal mine. In addition, the external disturbance signal was designed with the feed force as the control target and the measured pressure fluctuation of the pressure reducing valve as the basis. Further, comparative simulation was conducted with the proposed EID control method and PID control method respectively. The results show that the proposed method has a smaller peak-to-peak value of the steady-state tracking error than the PID method, as well as a smaller tracking error, which ensures the stable operation of the feed system and has better tracking and disturbance suppression performance. Generally, the research results of this paper provide a control theoretical basis for improving the adaptability of drilling robots to complex coal seam load changes, and ensuring their working performance and safe and efficient construction.