Abstract: In the complex underground coal mine environment, to meet the high-precision positioning requirements of inspection robots and address the issues of insufficient positioning accuracy, high cost, and poor reliability of single sensor in traditional GPS-denied environments, this study proposes a fusion positioning method based on strapdown inertial navigation system (SINS) and radio frequency identification (RFID) technology. This method effectively integrates SINS and RFID technology using the extended Kalman filter (EKF) technique. It calculates continuous pose information via the SINS as predicted values, and utilizes discrete absolute positions provided by RFID technology as observed values to achieve optimal estimation and error feedback correction. To further enhance the positioning continuity of the inspection robot positioning system in areas covered by RFID tag signals, piecewise linear interpolation is introduced to compensate for the motion trajectory of the inspection robot positioning system. Meanwhile, static calibration and low-pass filtering are adopted to eliminate constant drift and random drift of the inertial measurement unit, significantly improving the anti-interference capability and overall robustness of the system. Experimental results show that: when the RFID tag interval ranges from 50 m to 150 m, the proposed positioning method can stably maintain centimeter-level high-precision positioning performance, with a high degree of agreement between its trajectory and the real path, and the maximum positioning error is strictly controlled within 15 cm; when the moving distances are 10 m and 30 m, the system errors are only 9.3 cm and 9.5 cm, respectively, indicating that the cumulative drift of the inertial navigation system is effectively suppressed within the effective RFID correction interval; even in continuous turning sections of 70-80 m where the inertial navigation error is amplified, causing the error to rise to 25.9 cm, the system still maintains acceptable centimeter-level accuracy overall. In terms of real-time performance, the data fusion and calculation processing delay of the entire system is consistently stable within 17 ms, and the time jitter is always below 0.5 ms. In contrast, the delay of the laser simultaneous localization and mapping (SLAM) method based on graph optimization increases from 26 ms to 39 ms under the same conditions, while the delay of the inertial navigation-ultra-wideband (UWB) method increases by 55.5%, proving that the proposed method has low computational complexity and can meet the strict real-time positioning requirements of inspection robots.