Dipl . - Ing . Paul Meissner

Location awareness is a key enabler for a multitude of indoor applications, such as logistics, intelligent warehouses, or flexible production. The robust provision of accurate position information requires a careful fusion of location-dependent parameters acquired by various sensors, including range measurements from radio signals. The latter are the basis for successful outdoor localization systems like the Global Positioning System (GPS). However, indoor environments are often characterized by dense multipath channels. These originate from the superposition of many physical propagation mechanisms, which are caused by many spatially close interacting objects. Especially for indoor localization systems, this is still the main source of errors. However, deterministic reflections of the radio signal on surfaces such as walls can be modeled geometrically, leading to defined relationships of their path parameters with the position of an agent that is receiving such a signal and thus additional position-related information. The position-related information inherent in these signal paths is the main focus of this thesis. A geometric-stochastic channel model allows for the proper decomposition of signals into useful information and non-resolvable, interfering components. The concept of virtual anchors (VAs) is employed for the geometric modeling of deterministic multipath components (MPCs) using a known floor plan as prior knowledge. It is shown by experiments that these MPCs carry a significant part of the energy of a received radio signal. Statistical performance bounds are derived for multipath-assisted positioning, defining the position-related information of each MPC as a function of signal and channel parameters. Most importantly, a Signal-to-Interference-andNoise Ratio (SINR) is introduced that quantifies the amount of this information for each MPC. Estimation methods are presented for the SINR, which allow to characterize the position-related information of an environment. Based on the position-related information, the results presented in this thesis allow (i) to gain an understanding of the physically relevant propagation phenomena using measured data, (ii) to find statistical models to characterize their influence on the position-related parameters in the signals, and (iii) to show the excellent accuracy and robustness that can be achieved by exploiting multipath. Tracking algorithms are proposed that can be made aware of the positionrelevant propagation phenomena and their respective uncertainty. Using experimental data from various scenarios, this is shown to be the key factor enabling the constructive use of MPCs for positioning, consistently leading to position errors below 5 cm for 90 % of the estimates. The work in this thesis is in contrast to most existing literature on radio-based indoor positioning, in which methods are presented to counteract the performance impairments of multipath. One outcome is a real-time demonstration system that allows for fast and flexible testing of existing and new algorithms and serves as a proof-of-concept for multipath-assisted tracking. The experimental results in this thesis are valuable for the design and evaluation of future indoor wireless applications. This is especially true for the presented results on channel analysis, which give guidelines for the design and parametrization of spatially consistent geometric-stochastic channel models.