Storage and computationally efficient permutations of factorized covariance and square-root information matricesA unified method to permute vector-stored upper-triangular diagonal factorized covariance (UD) and vector stored upper-triangular square-root information filter (SRIF) arrays is presented. The method involves cyclical permutation of the rows and columns of the arrays and retriangularization with appropriate square-root-free fast Givens rotations or elementary slow Givens reflections. A minimal amount of computation is performed and only one scratch vector of size N is required, where N is the column dimension of the arrays. To make the method efficient for large SRIF arrays on a virtual memory machine, three additional scratch vectors each of size N are used to avoid expensive paging faults. The method discussed is compared with the methods and routines of Bierman's Estimation Subroutine Library (ESL).
Document ID
19890041158
Acquisition Source
Legacy CDMS
Document Type
Conference Paper
Authors
Muellerschoen, R. J. (California Institute of Technology Jet Propulsion Laboratory, Pasadena, United States)