Multiband Channel Estimation for Precise Localization in Wireless Networks: Algorithms, Simulations and Experiments

Over the last two decades, we have witnessed a tremendous evolution of wireless communication systems. For example, the data rates in mobile wireless systems have increased from a few tens of kilobits per second to 10 gigabits per second between the first and last, i.e., fifth generation (5G). The main enablers for this growth are signal processing and radio frequency (RF) hardware innovations, which led to more efficient modulation and coding schemes and high-performance RF transceivers. Following these trends, future wireless systems such as 6G and WiFi-7 aim for even higher data rates, requiring higher frequency ranges, wider bandwidths, and massive antenna arrays. These developments pave the way toward joint communication and sensing RF systems with very high range, Doppler, and angular resolutions. In particular, favorable signal and RF transceiver properties such as large bandwidth will enable precise RF localization in rich scattering environments such as indoor or urban canyons where multipath effects severely impair the performance of traditional localization systems like GNSS (Global Navigation Satellite Systems). At the same time, the wide range of emerging applications in areas of autonomous navigation, assisted living, and Internet-of-Things require precise localization, often to cm-level degree accuracy. Therefore, it is evident that new localization approaches and signal processing algorithms that can exploit signal and transceiver properties of emerging wireless systems are needed to solve the problem of precise localization in multipath environments and lead the way to novel applications.

The goal of this thesis is to design signal processing algorithms and protocols that will enable precise ranging in multipath environments while using practical single-antenna RF transceivers. In the first part of this thesis, we introduce a multiband channel model to describe multipath channel measurements collected over multiple separate frequency bands using narrowband and wideband RF transceivers. This model shows that multiband channel measurements have multiple shift-invariance property and that by increasing the frequency aperture of the multiband measurements, we can improve the resolution of multipath time-delay estimation. We use this property of the measurements to develop high-resolution time-delay estimation algorithms based on subspace estimation. To illustrate the performance of these algorithms, we perform extensive numerical experiments which demonstrate that the proposed algorithms are statistically efficient and that multiband time-delay estimation enables precise ranging in multipath environments.

However, the aforementioned results also show that the proposed algorithms are sensitive to errors introduced by hardware impairments of RF transceivers and imperfect calibration. In the second part of the thesis, we focus on the problem of joint RF transceiver calibration and high-resolution time-delay estimation. For example, in practical scenarios, the frequency response of RF transceivers might not be known nor calibrated, and performing time-delay estimation without calibrating these effects will lead to biased estimates. We show that the problem of joint RF transceiver calibration and time-delay estimation can be formulated as a particular case of covariance matching, which after reformulation, can be solved using a simple group Lasso algorithm. Likewise, due to imperfections of oscillators used in RF transceivers, the mobile and anchor nodes are usually not frequency synchronized. This frequency offset severely deteriorates the performance of multiband ranging methods. To solve this issue, we design a two-way protocol for collecting multiband channel measurements and a weighted least squares-based algorithm that enable joint clock synchronization and ranging.

Finally, in the last part of the thesis, we validate our modeling assumptions and illustrate the performance of the multiband time-delay estimation algorithms by considering practical scenarios of localization in future WiFi-7 networks. For these experiments, we use real indoor multipath channel measurements collected in a hospital and a university building environment. The results of the experiments show that using multiband channel measurements with a total bandwidth of 320 MHz, the absolute ranging error is smaller than 4 cm in 80% of the cases. Likewise, using the same scenario setup and three anchors to localize the mobile node, it is observed that the positioning error is below 24 cm in 95% of the cases. These results show that by using the advanced signal processing techniques to design estimation algorithms and channel measurement protocols that can exploit the properties and degrees of freedom offered by future wireless systems and RF transceivers, decimeter-level accurate positioning is achievable.

The signal processing models presented in this thesis are common to the wide area of array signal processing applications, such as radar and ultrasound imaging. Therefore, the results presented in this thesis impact these application areas as well.