Channel-SLAM: multipath assisted positioning

In wireless propagation transmitted signals are reflected, scattered and diffracted by objects. Especially in urban canyons or inside buildings, the signal reaching the receiving antenna consists of multiple replicas of the transmitted signal, which are called multipath components. The positioning accuracy might be drastically reduced due to the distorted received signal by multipath components. Hence, positioning algorithms need to mitigate the impact of multipath components on the received signal to obtain an accurate position estimate. With this thesis, we propose a paradigm shift in how to process the received signal in order to provide accurate position estimation for mobile receivers: rather than mitigating multipath components we propose an algorithm to exploit multipath. We call this algorithm Channel-SLAM. The basic idea of Channel-SLAM is to interpret multipath components as signals emitted from so called virtual transmitters. These virtual transmitters are inherently time synchronized to the physical transmitter and static in their positions. In this thesis, we show that the presence of multipath components allows positioning even if signals of only one physical transmitter are receivable. The concept of virtual transmitters considers multipath propagation occurring due to multiple numbers of reflections, diffractions or scattering as well as the combination of these effects. Specifically, we derive a generic signal model to describe virtual transmitters, where a distinct model detection for reflection, diffraction or scattering is not necessary. To use the information of the multipath components, Channel-SLAM estimates the position of the virtual transmitters without the necessity of any prior information such as a room-layout. The novelty of the algorithm is to estimate the position of the receiver and the virtual transmitters simultaneously, which can be interpreted as simultaneous localization and mapping (SLAM) with radio signals. Instead of mapping the physical environment, Channel-SLAM maps the virtual transmitter positions and interprets them as landmarks. To ease the computational complexity, a hierarchical particle filter based on Rao-Blackwellization is derived that estimates the position of each virtual transmitter using a separated particle filter. The derived hierarchical particle filter allows to use a different amount of particles in each particle filter associated to a virtual transmitter. Additionally, we show that the number of particles can by dynamically adapted during runtime which enables a significant performance gain. In order to quantitatively analyze the performance and location accuracy of Channel-SLAM, the posterior Cramer -Rao lower bound for Channel-SLAM is derived. Based on simulations, the position accuracy of Channel-SLAM is compared to the posterior Cramer-Rao lower bound. Especially for higher signal to noise ratios and beneficial geometric relations, the performances of the position estimations of the virtual transmitters using Channel-SLAM are close to the posterior Cramer -Rao lower bound. We verify the performance of Channel-SLAM using broadband signals in different scenarios. With our experiments we show that Channel-SLAM is able to accurately estimate the mobile receiver position by exploiting multipath propagation even without the prior information on the physical transmitter position. Furthermore, we confirm that Channel-SLAM is able to determine virtual transmitter positions accurately.