Continuous Indoor Positioning Under Random UWB Channel Occlusions Using RBEKF for UWB/IMU Data Fusion

Abstract

To meet the demands of continuous indoor positioning in Internet of Things (IoT) applications, this paper proposes a UWB/IMU fusion positioning system for robust operation in complex indoor environments. Firstly, to represent the inevitable non-line-of-sight (NLOS) phenomenon, we model occlusion states using a Markov chain with different degrees of occlusions, and incorporate them as extended states into the system model. Secondly, to mitigate the heavy computational burden of particle filtering (PF), we combine extended Kalman filtering (EKF) with Rao-Blackwellization to derive a RBEKF algorithm for Bayesian filtering. Simulations using field-collected data reveal that the extended state model including the occlusion improves positioning accuracy. Also, the proposed RBEKF algorithm achieves nearly optimal accuracy while significantly reducing the filtering complexity in comparison with the standard PF method. This makes the proposed method a valid option for indoor positioning tasks. Note to Practitioners—This work is motivated by the challenge of maintaining accurate indoor positioning in complex environments containing multiple types of occlusions, while also considering computational efficiency. To address this problem, multiple discrete occlusion states are modeled and incorporated as extended estimation states within the filtering framework, enabling a more accurate representation of the underlying system dynamics. Combined with SVM-based occlusion identification, the proposed method improves the mitigation of NLOS-induced ranging errors. Furthermore, a Rao-Blackwellized formulation is employed to significantly reduce the computational burden while preserving estimation accuracy. The proposed approach has been validated using real-world datasets collected from an indoor positioning platform together with simulation-based analysis. However, the algorithm has not yet been fully deployed and evaluated in a real-time operational positioning system. Future work will focus on implementing the proposed framework in practical indoor localization systems and further assessing its performance under long-term real-world operating conditions.