Analysis of Disparate Inertial Systems Data

Howard Musoff

Abstract:	A common situation in inertial systems testing arises when there is data from a number of different systems taken at different time and it is required to compare the data from these systems. Also, data from different tests taken at different times (possibly over several years) on the same system must be compared as well. This paper presents and analyzes different methods (such as cluster analysis) for accomplishing this task. For example, a system might be tested under different circumstances such as in centrifuge testing, flight testing, laboratory calibration, etc. We expect that for each type of test the estimates provided by the different Kalman filters designed for each test will not be exactly the same. One of the topics of this paper will be how to handle this data in order to reconcile it between the different tests. Part of the solution to this problem can be as simple as choosing the proper common states (each filter has a number of states that are not common with the other filters). Another complementary approach is the analysis of the disparate estimates using a non-parametric statistics approach where no underlying statistics (which is different for each filter) are assumed. With regard to the analysis of data (including estimates from the past) for a large number of separate systems for example, again a non-parametric approach such as cluster analyses (where the data is grouped into clusters) could be revealing as to which systems can be grouped according to age, manufacturer, etc. without relying on any assumptions about underlying statistics (which are not available).
Published in:	Proceedings of the 52nd Annual Meeting of The Institute of Navigation (1996) June 19 - 21, 1996 Royal Sonesta Hotel Cambridge, MA
Pages:	15 - 19
Cite this article:	Musoff, Howard, "Analysis of Disparate Inertial Systems Data," Proceedings of the 52nd Annual Meeting of The Institute of Navigation (1996), Cambridge, MA, June 1996, pp. 15-19.
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In