Bogdan Pricope

• The Global Positioning System (GPS) is a satellite-based location system, which is widely used for location determination, navigation and time synchronization. Because GPS signals are transmitted at relatively low power levels and over great distances, the received GPS signal is relatively weak. Moreover, inside buildings or in urban canyons, the GPS signals are further attenuated by walls, roofs and other objects. Thus, the reception of GPS signals is not reliable in indoor or obstructed environments. An accurate and reliable indoor positioning system is critical to provide navigation support to first responders in emergency scenarios as well as to minimize the rescue time of injured personnel. Moreover, Location Based Services (LBS) drive the development of indoor positioning techniques, mainly due to the need to provide service continuity inside buildings. Location based services are typically used with mobile devices in pedestrian mode for applications such as locationbased advertising, people or pet tracking and friend-finder functionality. All these applications require continuous, seamless and ubiquitous positioning, which GPS is not capable of delivering. In the last ten years several techniques have been proposed to solve the problem of indoor positioning. Some depend on the presence of a known network of base stations such as Wireless Local Area Networks (WLANs) access points, Bluetooth nodes or digital television towers. Other short-range technologies depend on specialized transceivers specifically designed for positioning e.g. GPS pseudolites and repeaters, active Radio Frequency Identification (RFID) or Ultra-Wide Band (UWB). However, no single positioning technology is able to provide high availability and precision. For this reason, the trend is toward hybrid solutions, i.e. the combination of different positioning methods. The seamless cooperation between different technologies is required in order to provide location capability whenever a clear view of the sky is obstructed. While dead-reckoning is not a standalone positioning technology (it needs a known start reference position) it is the most promising option when other infrastructure based solutions, such as GPS, are not available. The principle of operation is based on the ability to accurately measure the velocity vector i.e. both the amplitude (speed) and direction, starting from a known reference point. An accurate starting point can be assumed from a standalone positioning technology such as GPS, and the direction of movement can also be obtained with fidelity using a combination of magnetometers and gyroscopes. In order to determine the walking speed and thus the displacement compared to the reference position, accelerometers together with statistical models of the human walk are used to detect steps and to estimate the step length. However, these step-length estimators require user calibration and fail in practice when the user is moving with an irregular posture like crawling, sliding on a slippery surface, using the fast walking lane in an airport, riding a bicycle or skateboarding. This dissertation proposes a novel speed estimation method for pedestrian dead-reckoning, which is based on the coherence time of wideband terrestrial wireless systems. In the proposed method, the coherence time is defined as the time required for the envelope correlation coefficient of a wideband frequency-selective channel frequency response vector to drop below a certain threshold value. The effect of the several channel propagation conditions, such as signal-to-noise ratio, the direction-of-arrival of the line-of-sight component and Rician factor, is investigated on the performance of the proposed method. Moreover, a proof-of-concept hardware demonstrator is realized and several experimental measurement campaigns are conducted to verify the performance of the proposed algorithm under real-world indoor conditions. Experimental results indicate that speed estimation errors lead to an average positioning uncertainty of 1:1% of the total traveled distance with a standard deviation of 2:4% from the average. Moreover, there is a probability of 97% that the absolute positioning uncertainty will be less 4.5% of the traveled distance. It is demonstrated that the proposed speed estimation method outperforms traditional pedestrian dead-reckoning systems, which are based on step length estimators. In addition, the proposed algorithm does not require user calibration and it is not based on a particular type of movement because it solely measures the absolute speed of the moving antenna.