Robust Tightly Coupled UWB-ODO-Inertial Navigation in Complex Indoor Environments

In complex indoor environments, ultra-wideband (UWB) technology can be used to achieve continuous absolute positioning. However, its accuracy is often degraded by non-line-of-sight (NLOS) signal interference. In this paper, we propose a robust UWB-odometer (ODO)-inertial navigation system, named MODR-UOINS, to meet the high-accuracy positioning requirements for wheeled robots. Multi-sensor observations are tightly fused using the multi-state constraint Kalman filter (MSCKF) framework, enabling efficient management of historical states. The proposed method can also mitigate accuracy degradation by fully using the short-term accuracy of ODO/INS odometry, especially when the base stations are sparsely deployed. Meanwhile, a robust multi-epoch outlier detection and rejection (MODR) algorithm based on random sample consensus (RANSAC) is introduced to detect and reject NLOS outliers effectively. Specifically, the high-accuracy short-term relative poses derived from ODO/INS odometry are utilized to evaluate the consistency of multi-epoch UWB ranges. Experiment results demonstrate that the proposed MODR-UOINS achieves a positioning accuracy of 0.1 m in the open LOS environment and 0.15 m in the complex indoor NLOS environment. In addition, MODR-UOINS achieves an accuracy of about 0.17 m even when only two UWB base stations are available, exhibiting superior robustness.