NTRS - NASA Technical Reports Server

Back to Results
approximate effect of parameter pseudonoise intensity on rate of convergence for ekf parameter estimatorsWhen using parameter estimation methods based on extended Kalman filter (EKF) theory, it is common practice to assume that the unknown parameter values behave like a random process, such as a random walk, in order to guarantee their identifiability by the filter. The present work is the result of an ongoing effort to quantitatively describe the effect that the assumption of a fictitious noise (called pseudonoise) driving the unknown parameter values has on the parameter estimate convergence rate in filter-based parameter estimators. The initial approach is to examine a first-order system described by one state variable with one parameter to be estimated. The intent is to derive analytical results for this simple system that might offer insight into the effect of the pseudonoise assumption for more complex systems. Such results would make it possible to predict the estimator error convergence behavior as a function of the assumed pseudonoise intensity, and this leads to the natural application of the results to the design of filter-based parameter estimators. The results obtained show that the analytical description of the convergence behavior is very difficult.
Document ID
Document Type
Conference Paper
Hill, Bryon K.
(NASA Headquarters Washington, DC United States)
Walker, Bruce K.
(Cincinnati Univ. OH, United States)
Date Acquired
August 15, 2013
Publication Date
January 1, 1991
Publication Information
Publication: In: IEEE Conference on Decision and Control, 30th, Brighton, United Kingdom, Dec. 11-13, 1991, Proceedings. Vol. 2 (A93-13001 02-63)
Subject Category
Distribution Limits