A Statistical Outlier Detection Method of EGNOS Integrity Performance Data

H. Kannemans

Abstract:	For safety of life applications of satellite navigation systems, in e.g. aviation, reliability in the sense of integrity, availability and continuity is essential. In order to test the integrity, availability and continuity, a test methodology is required. This methodology should meet the requirement to be able to perform the analysis based on a limited amount of data collected within an acceptable observation time. This paper presents a test method for analyzing the integrity of receiver output data possibly contaminated with outliers. In the paper the analysis of integrity is discussed in detail. Usually one presents the test results in so-called Stanford plots visualizing the occurrence of Hazardous Misleading Information (HMI). Since the probability of HMI is very low, it is not practical to state that the HMI probability is simply the ratio of the number of HMI results divided by the total number of measured samples. Very often, no HMI condition occurs during the tests; the resulting number of HMI conditions will then be zero and in that case the computed HMI probability on this basis will be zero as well, being obviously incorrect. Therefore, one needs to invent a way for obtaining a realistic estimate of the HMI probability, also called the integrity risk. The method developed for this purpose starts from the determination of two probability density functions: the fitted or measured probability density function of the protection level and the probability density function related to the position error, together forming a two dimensional probability density function. A complicating factor in analyzing the probability density functions is the fact that due to the low probability of events occurring in the tail of the probability distributions this analysis is sensitive to outliers. So a detection method enabling the detection of outliers must be incorporated in the method. From the test case presented in this paper, it is concluded that the estimation of the integrity on the basis of a limited number of test campaign data is possible indeed. It also turns out that the analysis of outliers is absolutely necessary to obtain useful results. The results show that the integrity satisfied the ICAO requirement for the APV-I, APV-II and CAT-I aeronautical services during the actual test case.
Published in:	Proceedings of the 2008 National Technical Meeting of The Institute of Navigation January 28 - 30, 2008 The Catamaran Resort Hotel San Diego, CA
Pages:	420 - 431
Cite this article:	Kannemans, H., "A Statistical Outlier Detection Method of EGNOS Integrity Performance Data," Proceedings of the 2008 National Technical Meeting of The Institute of Navigation, San Diego, CA, January 2008, pp. 420-431.
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In