An integrity framework for image-based navigation systems.

摘要

The value of Global Navigation Satellite Systems (GNSS) in a multitude of both military and civilian navigation and timing applications has been indisputably shown over the last twenty plus years. Increased dependence on GNSS has driven the need for risk management in using these systems for critical operations. One example risk reduction is through use of an integrity monitoring process, a process which predicts the impact of a worst case or minimal detectable bias error on the navigation position error. The navigation community has spent nearly two decades in development of GNSS integrity algorithms and continues to refine methodologies and investigate new approaches. As a result, the concept of integrity has been formalized for GPS to encompass fundamental criteria, including the ability to detect measurement errors and notify the user when the system should not be used because of potentially corrupt measurements.;More recently, researchers have begun investigating the potential use of image-based navigation systems for applications where GNSS signals are not available or should not be used. Independence from external radio-frequency transmissions and potentially low-cost implementations of camera and image processing suites are among the key advantages of these newer systems. However, while larger issues of integrity are seemingly well-understood for many GNSS applications, to the best of our knowledge, no one has fully addressed rigorous integrity quantification with respect to image-based navigation systems.;This dissertation first examines and addresses the fundamental differences existing between the measurement models established for GPS and those of proposed image-based navigation systems. For example, GPS pseudorange measurements used in a position determination are provided as a single value per satellite. However, image measurements are inherently angle-based measurements and assumed to be provided as a pair of pixel coordinates for a mapped target. Under these conditions, consideration must be given not only to the units of the transformations between the states and measurements, but also to the fact that multiple rows of the observation matrix relate to particular error states in the image-based case. This is due the added dimensionality of the pixel pair measurements.;Based on this examination, an algorithm is then developed to instantiate a framework for image-based integrity analogous to that of GPS RAIM, as to be understood in common terms by the navigation community. The algorithm is applied to the case where the navigation system is estimating position only and then extended to the case where estimation of both position and attitude is required. Detailed analysis is presented demonstrating the impact of angular error on a single pixel pair measurement, comparing results from both estimation scenarios, and showing that, from an integrity perspective, there is a significant benefit of having known attitude information. Additional studies are done to demonstrate the impact of pixel pair measurement relative geometries on system integrity in this framework, showing potential improvement in image-based integrity through screening and adding measurements to the navigation system solution, when additional measurements are available.