Abstract: Current underground driverless train positioning technology calculates the train displacement based on optical sensor installed on the wheels. When the underground driverless train skids on the wet track, the positioning error will be generated. In view of the above problem, an inertial navigation and positioning system for underground driverless train was proposed. Inertial navigation module was introduced in the system based on current underground driverless train positioning technology, and double threshold algorithm was used to judge abnormal driving condition of the train combined with photoelectric sensor data and inertial navigation data, and safety factor of driverless train was increased. The inertial navigation module uses LPMS-NAV2 to obtain acceleration and heading angle of target, and the position coordinates of the target is calculated. In view of the problem that acceleration measurement of the target is affected by gravity acceleration, z-axis acceleration compensation method is used to eliminate the error brought by gravity. In view of the problem of cumulative error of positioning of the target, weight feedback constraint algorithm is introduced, and square difference loss function is constructed for weighted constraint of the positioning point, so as to reduce the cumulative error. A position beacon is set at each fork of the underground roadway to perform secondary calibration on the positioning information to further improve the positioning accuracy. The indoor test results show that the average positioning error of the inertial navigation and positioning system for the underground driverless train is 0.52 m.