TOA based localization and tracking in indoor multipath environment
Date of Issue2019-01-30
School of Electrical and Electronic Engineering
This thesis addresses the issue of localization and tracking using time-of-arrival (TOA) data for both line-of-sight (LOS) and nonline-of-sight (NLOS) paths measured at mul- tiple reference devices (RDs) in indoor multipath environments. This thesis proposes a novel virtual RD (VRD)based indoor TOA localization algorithm with both LOS and multipath components that can be used when an accurate floor plan is available. By introducing the concept of VRD, multipaths can be considered as virtual LOS paths that originate from mobile device (MD) to VRDs. Due to unknown measurement- to-path correspondence, many possible positions satisfy the localization and tracking equation. A grid-based data association algorithm is proposed to estimate the correct measurement-to-path correspondence. Using the estimated data association result, the MD can be localized with a two-step weighted least squares method. The ex- perimental and simulation results show that the proposed VRD based localization algorithm significantly outperforms conventional LOS based localization algorithms. When an accurate floor plan is not available, this thesis proposes a novel indoor tracking algorithm for joint estimation of the MD and the map. By modeling the floor plan as a collection of map features, the multiple-RD single-cluster probability hypothesis density (MSC-PHD) filter can be used for joint estimation of the MD and map features. Conventional MSC-PHD filters are developed for outdoor radar-based scenarios that only consider backscatter paths. For application in indoor localization and tracking, the LOS path and all higher-order reflections that carry information on the MD and map features must be formulated. This thesis proposes two new MSC- PHD filters by incorporating LOS path and higher order reflection paths, which are re- ferred to as a LOS incorporated MSC-PHD (LMSC-PHD) filter and a multi-reflection incorporated MSC-PHD (MRMSC-PHD) filter, respectively. In addition, to mitigate high computation load of the proposed MSC-PHD filters, a computational tractable implementation that combines a new greedy measurement partitioning scheme and a particle-Gaussian mixture filter is presented. Furthermore, a novel mapping error metric is proposed to evaluate the accuracy of estimated map. The experimental and simulation results show that our proposed LMSC-PHD filter and MRMSC-PHD filter outperforms existing MSC-PHD filters by a significant margin in terms of both localization and mapping accuracy.
DRNTU::Engineering::Electrical and electronic engineering