n and Exc lus ion

His main research area is scientific computing with particular emphasis on matrix computations and applications. Currently his interests include GPS integrity monitoring and positioning. His main research area is numerical linear algebra and its applications, and one of his interests is efficient and numerically reliable estimation for linear systems. Dr. Studenny is the Aerospace GPS Engineering Manager at BAE Systems Canada. He and his team are responsible for the development of the next generation GPS receivers for precision approach applications (WAAS/LAAS) at BAES. His research interests include GPS receiver development, GPS receiver processing algorithms, and integrated navigation systems development. ABSTRACT This paper presents two carrier phase based approaches for receiver autonomous fault detection and exclusion integrity monitoring (FDE). Some models based on carrier phase measurements are established. Any typical snapshot FDE algorithm based on a linear model can be applied directly to the models. The first approach uses only measurements of consecutive epochs. For a single difference measurement model, if 4 epochs are used, at least 6 satellites are required to detect a fault and 7 are required to identify the faulty satellite. The second approach combines the measurements of the current epoch and estimates of integer ambiguities and error covariances obtained from the positioning algorithm of Van Graas and Lee. This approach is more suitable for computing the horizontal protection levels. If single difference measurements are used, at least 5 satellites are required for fault detection and 6 are required for identification. 1 INTRODUCTION Receiver autonomous fault detection and exclusion integrity monitoring (FDE) is crucial for some GPS applications. FDE consists of two distinct parts: satellite fault detection/exclusion and bad geometry detec-tion/exclusion, see Kelly [4]. The term "exclusion" includes "identification" or "isolation" (see [5]). We will use "identification" in the context of this paper, even though we use the more general term "exclusion" in the title. Most existing fault detection and identification methods are for code based linear models and are roughly equivalent. In essence, for fault detection and identification, they just remove an equation from the model and check the reduction of the residual, see for example Kelly [4]. It is important to distinguish between models and methods in this area. For any model there may be several methods we can apply in order to carry out a given integrity test. Two subclasses of methods will be of particular interest:-equation removal methods, and satellite removal methods. The former …