This dissertation makes four contributions in support of secure and robust position, navigation, and timing (PNT) for automated systems. The first two relate to PNT security while the latter two address robust positioning for automated ground vehicles.

The first contribution is a fundamental theory for provably-secure clock synchronization between two agents in a distributed automated system. All one-way synchronization protocols, such as those based on the Global Positioning System (GPS) and other Global Navigation Satellite Systems (GNSS), are shown to be vulnerable to man-in-the-middle delay attacks. This contribution is the first to identify the necessary and sufficient conditions for provably secure clock synchronization.

The second contribution, also related to PNT security, is a three-year study of the world-wide GPS interference landscape based on data from a dual-frequency GNSS receiver operating continuously on the International Space Station (ISS). This work is the first publicly-reported space-based survey of GNSS interference, and unveils previously-unreported GNSS interference activity.

The third contribution is a novel ground vehicle positioning technique that is robust to GNSS signal blockage, poor lighting conditions, and adverse weather events such as heavy rain and dense fog. The technique relies on sensors that are commonly available on automated vehicles and are insensitive to lighting and inclement weather: automotive radar, low-cost inertial measurement units (IMUs), and GNSS. Remarkably, it is shown that, given a prior radar map, the proposed technique operating on data from off-the-shelf all-weather automotive sensors can maintain sub-50-cm horizontal position accuracy during 60 min of GNSS-denied driving in downtown Austin, TX.

This dissertation’s final contribution is an analysis and demonstration of the feasibility of crowd-sourced digital mapping for automated vehicles. Localization techniques, such as the one described in the previous contribution, rely on such digital maps for accuracy and robustness. A key enabler for large-scale up-to-date maps is enlisting the help of the very consumer vehicles that need the map to build and update it. A method for fusing multi-session vision data into a unified digital map is developed. The asymptotic limit of such a map’s globally-referenced position accuracy is explored for the case in which the mapping agents rely on low-cost GNSS receivers performing standard code-phase-based navigation. Experimental validation along a semi-urban route shows that low-cost consumer vehicles incrementally tighten the accuracy of the jointly-optimized digital map over time enough to support sub-lane-level positioning in a global frame of reference.