SYNTHESIS OF COMPUTATIONALLY EFFICIENT SEQUENTIAL LINEAR ESTIMATORS,
Abstract
Optimal sequential linear estimators based on the work of Kalman and others require a recursive computation of the covariance matrix of the estimation error. The matrix multiplication required by the sampled case may become prohibitive for an operational computer designed to solve a practical problem in real time. To obtain a linear estimator with fewer computations the system state vector is partitioned into subsystem state vectors. The equations for approximating the covariance matrix of the estimation error of the subsystem state vectors are developed. These are used to develop an equation for the subsystem gains which are used with the observation residuals to obtain the state vector estimate. Methods for determining and evaluating partitionings on a digital computer are developed. As an example an inertial navigation system with a velocity reference and celestial body tracker was chosen. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1965
- Accession Number
- AD0619869
Entities
People
- Eugene Pentecost
Organizations
- University of California, Los Angeles