Model Error Compensation Techniques for Linear Filtering
Abstract
The exceptional utility and performance of the sequential, linear, unbiased, minimum variance estimator suffers severely in the presence of dynamic model errors. This problem--perhaps the greatest detriment to the so-called Kalman filter algorithm--is discussed in the light of its divergent effect upon the estimation process. A number of optimal and suboptimal modifying techniques are described which attempt to prevent this divergence. Extensions are developed resulting in adaptive forms and a new algorithm is derived for sequentially estimating the state noise covariance matrix. Performance of the techniques is illustrated by their application to, (1) the terminal phase of an Earth orbit rendezvous mission, and (2) the heliocentric trajectory determination of a solar electric propulsion space vehicle. Numerical results indicate that the model error difficulties can be sufficiently countered, with particularly effective performance being supplemented by the sequential state noise covariance estimator.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 1973
- Accession Number
- AD0770067
Entities
People
- Hamilton Hagar Jr.
Organizations
- University of Texas at Austin