UWB/IMU integrated indoor positioning algorithm based on SA-IEKF method

High-precision indoor location services have become an essential requirement. Among various indoor positioning approaches, the inertial measurement unit (IMU)-based inertial localization and ultrawideband (UWB)-based positioning algorithms have attracted extensive research. However, the IMU algorithm has a cumulative error, and the UWB is greatly affected by the environment. To address these issues, this paper presents a UWB/IMU fusion indoor positioning method. It first proposes the adaptive dynamic window-based long short-term memory (LSTM) heading estimation algorithm. This algorithm first recognizes pedestrian motion states via multi-level criteria and an adaptive dynamic sliding window, then constructs a high-precision heading correction model using LSTM networks. Building on this, the scenario-aware iterated extended Kalman filter (SA-IEKF) fusion algorithm is proposed. Experimental results reveal that the proposed SA-IEKF method achieves remarkably lower root-mean-square error in complex indoor scenes. Meanwhile, the proposed method presents a more concentrated error distribution and stronger robustness, showing comprehensive performance advantages under frequent motion switching condition.