Accession Number:
ADA055686
Title:
State Noise Covariance Computation in the Kalman Filter.
Descriptive Note:
Master's thesis,
Corporate Author:
AIR FORCE INST OF TECH WRIGHT-PATTERSON AFB OHIO SCHOOL OF ENGINEERING
Personal Author(s):
Report Date:
1977-12-01
Pagination or Media Count:
150.0
Abstract:
This report investigates forms of the state noise covariance matrix in the Kalman Filter. This matrix, denoted Q sub d, incorporates the effects of random errors driving the system dynamics into the filter computations. The Q sub d matrix is derived by integration from the matrix of continuous time driving noise strengths, which normally includes only diagnoal terms. This often leads to use of a diagonal Q sub d matrix with constant terms. However, the derivation shows that Q sub d should have off-diagonal and time varying terms. The study investigates the effects of including such terms in Q sub d. Three alternate forms of Q sub d are derived for a specific inertial navigation system. These, and a standard diagonal form, are tested using a covariance analysis. The results show little difference in performance for the different filters. This is attributed to two primary factors highly accurate external measurements, and the use of integration sub-intervals for covariance propagation.
Descriptors:
Subject Categories:
- Statistics and Probability