An Indoor UWB Localization Method Based on Adaptive Channel Bias Estimation

Measurement bias due to non-line-of-sight (NLOS) reflections and indistinguishable multipath fading can be collectively referred to as channel bias, which degrades the performance of the iterative localization algorithms and is the main reason limiting the performance of range/angle-of-arrival (AoA)-based ultra-wideband (UWB) localization systems. To build an adaptive and generalized indoor UWB localization system, channel bias and the corresponding means of compensation and elimination need to be investigated. In this article, we provide an indoor UWB localization method based on adaptive channel bias estimation to compensate for distance and angle measurement errors. By means of a Dirichlet process Gaussian mixture model (DPGMM) based on variational inference, the channel bias model for each anchor can be built without actual position calibration, and furthermore, a maximum a posteriori (MAP) estimator for the tag location is designed. Test results in open indoor environments and complex indoor environments show that our method performs comparably in static line-of-sight (LOS) cases and better in NLOS and dynamic cases compared to the conventional single-anchor localization method.