SENSITIVITY ANALYSIS OF DISCRETE FILTERING AND SMOOTHING ALGORITHMS,
Abstract
Two problems which occur when filtering or smoothing methods are applied to an actual problem are the choice of the prior statistics and the choice of a mathematical model for the system. The model must be complete enough for an adequate description of the system and also sufficiently simple such that the resulting algorithms are computationally feasible. A major tool in both analyzing the effects of these choices and in making the best choice is a sensitivity analysis of the algorithms required in the solution of the filtering and smoothing problems. This investigation presents a complete and definitive sensitivity analysis of the discrete smoothing problem. Algorithms are developed for matrix sensitivity functions for both large and differential errors. Two examples are presented to illustrate the use of the given algorithms: a linearized version for the in-track motion of a satellite traveling in a circular orbit, and a simplified version of an error model of a long term inertial navigation system which might be used for marine application. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 29, 1968
- Accession Number
- AD0677454
Entities
People
- Andrew P. Sage
- Robert E. Griffin
Organizations
- Southern Methodist University