Indoor localization with ultra wideband radio

Location-awareness has been drawing significant attention in recent years. In particular, the emerging Ultra Wideband (UWB) technology has the potential to revolutionize the systems and applications in many sectors. Due to its fine temporal resolution, UWB radio is especially promising for localization in indoor environment. This thesis focuses on two-step localization and investigates the UWB technology for indoor localization. Indoor environment often creates multipath in radio propagation, and the receiver receives multiple copies of UWB pulses with different delays and amplitudes. When the multipath signals are well-separated from each other, they can be resolved by conventional sliding correlator and do not interfere with the direct path (DP) signal. However, in a dense multipath channel, the subsequent multipath signal could overlap with the DP signal, which distorts the correlator output and introduces error in Time-of-Arrival (TOA) estimation. To measure the TOA accurately, more sophisticated algorithms are required to resolve the overlapping signal components. Most works in the literature adopt iterative approach to find the TOA of each signal components, and the convergence rate depends on the number of signal components. In Chapter 3 of the thesis, a non-iterative Modified Phase-Only Correlator (MPOC) is proposed for high-resolution multiple TOA estimation. The MPOC deconvolves the multipath channel frequency response from the received signal by dividing the latter with the local template spectrum. Due to the inherent spectral nulls of the local template, the MPOC suffers from noise amplification problem. In Chapter 3, a kurtosis-based outlier detection method is also proposed to identify and suppress the amplified noise. Both simulation and experiment verifies the ability of MPOC to resolve overlapping UWB pulses and renders it for highresolution TOA estimation. Besides multipath, the other major source of error in indoor localization is the Non-Line-of-Sight (NLOS) propagation problem, where the DP signal is either obstructed or diffracted. The NLOS problem introduces a positive bias in the TOA estimation. The NLOS measurement should be either discarded or imposed as region constraint. A prerequisite for both treatments is to identify the received signal that undergoes NLOS propagation. Identification of NLOS propagation is therefore critical for accurate target localization. In Chapter 4 of the thesis, a NLOS identification technique based on hybrid Received Signal Strength (RSS) and TOA is proposed. The basic idea of the proposed technique is to detect the divergence in the signal propagation distances estimated from RSS measurement and from TOA measurement. In more practical hyperbolic and elliptical localization systems where the target is not synchronized with the anchor, simple anchor cooperation is needed and the divergence in the range difference is used instead as the detection metric. Performance analysis and result show that the proposed technique works best if the DP signal is attenuated significantly and the bias in time measurement is small. Once the TOA of the signal is estimated, it is converted to meaningful target locus. Conversion of TOA to target locus depends on the ranging protocol, which is mainly determined by the level of synchronization. The concept of elliptical localization is proposed in Chapter 5. In elliptical localization, the target position is found from the intersection of ellipses defined by the sum of ranges from the anchor transmitter to the target and from the target to the anchor receivers. Leveraging on the knowledge of anchor positions, the proposed elliptical localization system requires no explicit synchronization, and timing information is obtained by measurement of the differential TOA between the signal from the anchor transmitter and the signal from the target. Elimination of explicit synchronization is a significant advantage and makes the system suitable for ad-hoc applications where quick set-up is desired. Almost every localization system requires the position of the anchors. While knowledge of anchor position is essential, how to obtain the anchor positions is rarely studied. Chapter 6 of the thesis fills the gap by studying the anchor position calibration problem. The anchors are assumed to be capable of conducting range measurement among them. Depending on the implementation of ranging protocol, the ranges between anchor pairs can be absolute or relative. Absolute range is measured either by highly synchronized anchors or by twoway ranging technique. The major drawbacks are extra hardware complexity and clock drift error correspondingly. When the anchors are synchronized as in hyperbolic localization system, the relative range among the anchors can be estimated by rotational time difference measurement and ranking of the interanchor ranges can then be obtained. With the absolute value or the ranking of the inter-anchor ranges, classical and ordinal multi-dimensional scaling (MDS) can be applied to find the local map of the anchors. Similar to most real-world problems, there is no solution that is best for all. Different applications have different constraint and requirement for the localization system. The optimum or proper implementation and architecture are determined by factors such as cost, set-up time, application scenario, and targeted accuracy. Often, design of a localization system involves performancecost trade-off, which is addressed in this thesis by comparing various solutions and outlining their advantages and drawbacks for various scenarios.DOCTOR OF PHILOSOPHY (EEE