Abstract: In order to solve the problem that the existing integrated inertial navigation method loses completely autonomous advantage and increases cost when applied to intelligent sensors in underground roadway, the underground roadways are firstly decomposed into multiple two-dimensional planes with certain vertical height. And multiple established paths are formed in the two-dimensional planes, and the positioning problem of intelligent sensors is converted into the tracking problem of moving tracks on the established paths. Secondly, the inertial measurement unit based on the micro electro-mechanical system (MEMS-based IMU) is used to realize the inertial navigation of the intelligent sensor in the underground roadway, and the moving track is inverted by combining the zero velocity update and the established path calibration. After the zero velocity update of the intelligent sensor at the starting point of the established path, the moving of the whole established path can be divided into two modes, namely linear inertial navigation mode and intersection calibration mode. Linear inertial navigation: when the change values of the heading angle and the roll angle of the sensor do not exceed the threshold value, the speed, the position and the attitude angle of the sensor are calculated through the inertial navigation component, and the real-time relative coordinates and the motion path are calculated. Intersection calibration: when the sensor moves to the intersection, the current real-time value is calibrated according to the known coordinate value of the intersection so as to eliminate the accumulated error of inertial navigation. The results of the test on the ground ring road show that in the context of uncalibrated, the heading deviation of pure inertial navigation is about 30°, and the relative error of moving distance is 5.5%. After the calibration of the established path intersections, the heading deviation is about 2°, and the relative error of the moving distance is 0.8%. After calibration, the moving track is more consistent with the actual road.