Self-positioning of a shearer is the key technology for mining with a man-less working face. In an underground coal mine all radio navigation; satellite positioning or celestial navigation methods have their limitations. We analyzed an inertial navigation system intended to guide the movement a shearer and designed a self-positioning device for the shearer. Simulation tests were also performed on the system. We analyzed the errors observed in these tests to show that the main reason for the low precision of the self-positioning system is accumulated error in the inertial sensor. A Kalman filtering algorithm used in combination with the shearer motion model effectively reduces the measurement errors of the self-positioning system by compensating for gyroscopic drift. Finally, we built an error compensation model to reduce accumulated errors using continuous correction to provide self-positioning of the shearer within a certain range of accuracy.