An Indoor Multi-sensor Fusion Localization Method Based on UWB

Abstract

Global navigation satellite system (GNSS) localization accuracy is significantly degraded in indoor environments, failing to satisfy the precise localization requirements of robotic systems. Ultra-wideband (UWB) localization technology has been applied to indoor navigation. However, its accuracy degrades substantially in non-line-of-sight (NLOS) environments. To enhance the indoor localization accuracy of intelligent vehicles, this paper proposes a multi-sensor fusion localization method based on UWB, an inertial measurement unit (IMU), and light detection and ranging (LiDAR). The extended Kalman filter (EKF) is employed to construct separate UWB/IMU and UWB/LiDAR subsystems. The localization results of these two subsystems are subsequently optimized and corrected using the weight-adaptive particle filter (WAPF) algorithm to produce optimal localization state estimates. We conducted experiments using an intelligent vehicle in the non-line-of-sight environment of 4.2m×2.4m and the line-of-sight environment of 6.0m×1.0m respectively. In the non-line-of-sight environment, the average localization error of the method proposed in this paper reaches 0.0420m, which is 59.70% smaller than that of the single-sensor system and 18.52% smaller than that of the EKF method. In the line-of-sight environment, the average localization error of the method proposed in this paper reaches 0.0209m, which is 44.71% smaller than that of the single-sensor system and 14.34% smaller than that of the EKF method. Experiments were conducted using an intelligent vehicle in a non-line-of-sight (NLOS) environment measuring 4.2 m × 2.4 m and a line-of-sight (LOS) environment measuring 6.0 m × 1.0 m. In the NLOS environment, the proposed method achieved an average localization error of 0.0420 m, which is 59.70% lower than that of the single-sensor system and 18.52% lower than that of the EKF method. In the LOS environment, the average localization error was reduced to 0.0209 m, representing a 44.71% reduction compared to the single-sensor system and a 14.34% reduction compared to the EKF method. Data from the UWB/IMU and UWB/LiDAR subsystems were processed using the EKF and subsequently fused through the WAPF algorithm, achieving robust localization in indoor environments. The proposed multi-sensor fusion localization method effectively combines the continuous motion perception of the IMU, the environmental feature matching capability of LiDAR, and the long-range absolute position constraints provided by UWB. This integration successfully compensates for the shortcomings of individual sensors and improves the accuracy and stability of localization.