Geometry-Graded ESKF UWB/IMU Tracking Algorithm With Base-Station Optimization for Hybrid LoS/NLoS Environments

Abstract

With the advancement of Internet of Things (IoT) technologies, the demand for accurate indoor positioning is increasing. Existing methods perform well in line-of-sight (LoS) conditions but suffer in non-line-of-sight (NLoS) conditions due to signal reflections and obstructions. To address this challenge, this paper presents a UWB/IMU fusion tracking algorithm that integrates Ultra-Wideband (UWB) and Inertial Measurement Unit (IMU) data, with a BS-status identification and geometry-aware update strategy. Firstly, an initial position is estimated using a nonlinear least squares (LS) method enhanced by a Huber loss function. Secondly, a BS status identification algorithm is introduced to identify LoS/NLoS conditions and evaluate BS availability based on received signal power differences and ranging variation rates. Finally, a Geometry-Graded Error State Kalman Filter (GG-ESKF) is developed to fuse IMU data with UWB measurements according to base-station availability. In the dual-BS condition, the original range measurements are reparameterized as a position-domain pseudo-measurement whose covariance is propagated from the status-dependent ranging uncertainty and local dual-BS geometry. Pseudo-measurement validation with fallback to the original range measurement update is introduced to reject unreliable pseudo-measurements. Experimental results demonstrate that the proposed algorithm significantly enhances positioning accuracy in hybrid LoS/NLoS environments, achieving an average error of 0.056 m and a root mean square error (RMSE) of 0.072 m in the laboratory, and an average error of 0.094 m with an RMSE of 0.129 m in the garage, outperforming existing filtering methods.