UC Irvine

This dissertation focuses on the development of robust and reliable navigation algorithms for mobile autonomous systems. This work is applicable to inertial navigation systems (INS) aided by external measurements from the Global Positioning System (GPS) and/or ultrawideband (UWB) sensors in a cooperative manner. To improve the robustness and continuityof precise navigation systems, this work looks at various factors, such as identifying and excluding erroneous sensor measurements, effective use of machine learning tools for navigation systems, and innovative navigation frameworks and infrastructures to enhance external signal access. On the integrity monitoring front, we develop an information-theoretic approach for fault detection and exclusion (FDE). This work also includes a statistical inference method to optimally estimate the probabilistic fault detection threshold. For robust navigation framework and infrastructure design, we propose innovative solutions based on collaborative navigation approaches such as UWB-based foot-to-foot ranging for dual-mounted INS for pedestrian localization, cooperative navigation for multi-agent systems, and self-localizing on-demand portable wireless beacons for coverage enhancement of RF beacon-based indoor localization systems. To induce robustness in these designs, we incorporate multiple measures. In UWB-based foot-to-foot ranging, we analyze the UWB ranigng measurement acuraccy and how proper relative placement of the UWB sensors can lead to measurement accuracy enhancement. To improve ranging accuracy of the UWB measurements, we also study how to remove the bias in different measurement models, i.e., line-of-sight (LoS) and non-line-of-sight (NLoS), using an artificial neural network (ANN) approach. Our contributions span across the architecture design of ANNs, identifying the informative features of the UWB signal to train the ANNs, and employing an OptiTrack motion capture camera system to collect a diverse set of training data in various relative poses between the sensors. Our work also explores the use of ANNs to improve the computational complexity of loosely coupled cooperative navigation solutions. Our work uses a supervised machine learning approach to learn the solution of computationally expensive optimization processes of loosely coupled cooperative navigation solutions from off-line data. The result is a set of lightweight ANNs that can predict the solutions online in a computationally efficient manner. The innovation in our work is proposing ANN architectures that output feasible solutions, i.e., solutions that are guaranteed to satisfy the constraints of the optimization processes. Our last contribution towards robust navigation solutions is to extend the use of our loosely cooperative navigation method to design a framework to deploy portable on-demand beacons to extend the coverage/signal access of the RF beacon-based localization systems. Our solution addresses the challenge of how to localize these deployed beacons in an on-line and decentralized manner. The proposed solution is a robust deployment method in the sense that if a portable beacon is moved for any reason, it can automatically re-localize itself in the decentralized manner. Simulations and experimental studies demonstrate the results of this thesis work.