In order to solve the localization problem of underground coal mine shuttle car, a fusion localization method based on the Extended Kalman Filter algorithm with the tight combination of Inertial Measurement Unit and Ultra-Wideband is proposed. Initially, an approach involving sliding window de-extremum and mean-taking is devised to enhance UWB ranging accuracy by mitigating Non-Line-Of-Sight and multipath effects. Processed UWB ranging data serves as EKF's observed value, while the distance between IMU and anchor serves as the predicted value, facilitating a tight combination within the EKF framework. This method achieves precise shuttle car localization within coal mines. During experimentation, three UWB anchor layout schemes are devised and compared to optimize positioning accuracy. Results indicate that equal-height arrangement of four UWB anchors yields the highest accuracy, with an average absolute error of 0.1264 m. To simulate the coal mine environment accurately, experiments are conducted in a long corridor with glass walls to intensify multipath effects. Additionally, personnel movement during experiments increases the NLOS effect, validating the proposed fusion method's robustness in complex environments. Simulation and experimental findings demonstrate that the fusion positioning method outperforms standalone UWB or IMU/UWB loose combination methods. It closely approximates real trajectories, exhibits high accuracy, and robustness in NLOS environments, meeting the positioning needs of underground shuttle car.