Optimized deployment and performance of time-based local positioning systems in urban and industrial environments

Abstract

The contemporary landscape of technological advancements is characterized by a concerted effort to reduce the human oversight for different devices and process. However, achieving a complete autonomous behavior in moving devices requires the employment of accurate, precise, reliable and efficient localization systems. Traditionally, Global Navigation Satellite Systems (GNSS) have constituted the preferred localization system, mainly due to their global range and overall accessibility of use. However, while GNSS are capable of achieving high accuracy and precision for complex environments with difficult orography, their performance can be insufficient for certain contexts and applications. The accuracy of localization systems in deep urban scenarios is significantly compromised by the high obstacle density that defines such environments. Consequently, Non-Line-of-Sight (NLOS) links between satellites and targets, along with multipath interference, are prevalent occurrences that considerably impact the reception and interpretation of localization signals. This challenge is not exclusive to outdoor settings; indoor applications face similar issues. Furthermore, the signals from GNSS experience substantial degradation as they traverse through walls. Consequently, meeting the high accuracy demands of localization applications in these scenarios may prove unfeasible by relying on GNSS. This has prompted the exploration of alternative localization methods. In this context, Local Positioning Systems (LPS), which are locally deployed sensor networks for providing localization service over a bounded region, emerge as a viable solution for achieving precise localization in challenging environments like the two previously discussed scenarios. Although various LPS technologies exist, this dissertation concentrates on time-based LPS due to their capability to deliver high accuracy, robustness, and ease of implementation across a diverse range of applications. Nevertheless, the effectiveness of these systems is closely tied to the specific geometric distribution of sensors in space. With an optimized arrangement, deployed time-based LPS can mitigate the impact of clock inaccuracies, noise, and multipath effects. This is typically accomplished by modeling the localization uncertainties of LPS using the Cramér-Rao-Bound (CRB). This constitutes a LPS distribution evaluation model that can be implemented for optimization algorithms to find adequate results. However, the achieved quality of these solutions depends on the quality of the error modelization as well as the optimization methodology proposed for their resolution. Hence, this thesis focuses on the distribution and performance optimization of these systems. It aims to enhance existing error models by incorporating additional error sources and addressing pertinent aspects related to the practical operation of these systems. Simultaneously, the thesis endeavors to develop effective and efficient optimization algorithms that can complement the devised error models. In this context, Chapter 5 addresses a novel sensor distribution paradigm that incorporates various sensor selection strategies into the optimization process. The consideration of this aspect within the sensor distribution optimization enhances the performance of the resulting architectures over urban environments where sensor selection is employed. Furthermore, Chapter 5 also implements a thorough comparison of various optimization techniques is conducted to address the increased computational complexity arising from simultaneous consideration of sensor selection and sensor distribution optimization. Furthermore, a novel optimization technique is specifically devised for this problem, yielding the most favorable results in the comparison. Based on this innovative evaluation and optimization framework, Chapter 6 conducts a comparison of synchronous architectures for localizing aerial vehicles within urban applications. The results validate the reliability and effectiveness of this procedure, particularly in high-demanding applications. This sensor selection consideration was subsequently extended to indoor industrial environments, and in Chapter 7, a multipath error model and an adaptation of the optimization procedure for indoor layouts are proposed for the localization of moving Unmanned Aerial Vehicles (UAVs) in an industrial plant. In this work, a comparison over synchronous LPS and the asynchronous A-TDOA architecture in the search of the best performing architecture for indoor industrial applications was performed. This comparison is based on the same evaluation framework developed within this thesis. The results highlighted the relevance of clock synchronism errors of time-based LPS in these context and a superiority of the A-TDOA in accordance with this issue, which promoted further insight on the consideration of additional asynchronous architectures for these contexts. Finally, in Chapter 8, the analysis conducted in Chapter 7 is expanded to encompass a thorough comparison of presently recommended asynchronous Two-Way-Ranging (TWR) architectures within an indoor scenario, particularized for Autonomous Ground Vehicles (AGVs). This chapter models the error sources of various TWR architectures and incorporates them into the proposed framework for evaluation and optimization. The outcome is a more comprehensive and practical comparison of these architectures, from which relevant conclusions are drafted for future studies In conclusion, this dissertation offers a comprehensive examination of the utilization of timebased Local Positioning Systems to fulfill high-accuracy demands in practical scenarios across different environments and applications. The developed methodologies and obtained results contribute significantly to current progress in localization literature, establishing a novel and more comprehensive framework for evaluating and optimizing sensor distribution. Furthermore, the insights gained into the modeling of real Time-of-Flight Ranging (TWR) architectures in indoor scenarios create opportunities for the facilitating implementation of these systems in demanding applications. This aspect represents a current research direction outlined by the thesis, actively being pursued.