Author: Deborah Kathleen Meduna

Publisher: Stanford University

Published: 2011

Total Pages: 183

ISBN-13:

Terrain Relative Navigation (TRN) provides bounded-error localization relative to an environment by matching range measurements of local terrain against an a priori map. The environment-relative and onboard sensing characteristics of TRN make it a powerful tool for return-to-site missions in GPS-denied environments, with potential applications ranging from underwater and space robotic exploration to pedestrian indoor navigation. For many of these applications, available sensors may be limited by mission power/weight constraints, cost restrictions, and environmental effects (e.g. inability to use a magnetic compass in space). Such limitations not only degrade the accuracy of traditional navigation systems, but further impact the ability to successfully employ TRN. Consequently, despite numerous advances in TRN technology over the past several decades, the application of TRN has been restricted to systems with highly accurate and information-rich sensor systems. In addition, a limited understanding of the effects of map quality and sensor quality on TRN performance has overly restricted the types of missions for which TRN has been considered a viable navigation solution. This thesis develops two new capabilities for TRN methods, resulting in significantly increased TRN applicability. First, a tightly-coupled filtering framework is developed which enables the successful use of TRN on vehicles with both low-accuracy navigation sensors and simple, low-information range sensors. This new filtering framework has similarities to tightly-coupled integration methods for GPS-aided navigation systems. Second, a set of analysis and design tools based on the Posterior Cramer-Rao Lower Bound are developed which allow for reliable TRN performance predictions as a function of both sensor and map quality. These analyses include the development of a new terrain map error model based on the variogram which allows for performance prediction as a function of map resolution. These developed capabilities are validated through field demonstrations on Autonomous Underwater Vehicles (AUVs) operated out of the Monterey Bay Aquarium Research Institute (MBARI), where available sensing has been limited primarily by cost. These trials include a real-time, closed-loop demonstration of the developed tightly-coupled TRN framework, enabling 5m accuracy return-to-site on a sensor-limited AUV where traditional TRN methods failed to provide better than 150m accuracy. The results further demonstrate the accurate prediction capability of the developed performance bounds on fielded systems, verifying their utility as design and planning tools for future TRN missions.