Abstract: According to the working characteristics of underground road-header, the walking position and posture deviation model of a road-header is established. Taking the caterpillar moving line speed and turning angle speed as control inputs, the control law of path tracking scheduling is designed and simplified by using principle of Lyapunov and backstepping method. The BP neural network is used to update the dynamic optimization of the key factors in the control law to compensate the tracking deviation of the body position and posture which is from the designed track in real time. The simulation results show that the proposed working control model of road-header based on BP network is simple and easy to implement. The position and posture deviation of the road-header can converge to zero within a limited tracking step and the angular speed adjustment process is stable. This proves that the tracking effect under the control of this model is good and has great potential for application.