PERFORMANCE ANALYSIS OF SUBOPTIMAL NON-ADAPTIVE AND OPTIMAL ADAPTIVE DISCRETE KALMAN FILTERS.
TEXAS UNIV AUSTIN ELECTRONICS RESEARCH CENTER
Pagination or Media Count:
The performance of discrete suboptimal Kalman filters and discrete optimal adaptive filters can be analyzed by an error covariance matrix which is defined on the specific filtering structure. Iterative algorithms for calculating the absolute performance of suboptimal filters are achievable for the case of known modeling errors. For the case of unknown statistical and dynamical system parameter values and optimal Bayesian estimate of the actual system performance provides a means of analyzing the system. Similarly, the parameter value uncertainties in the adaptive filter formulation necessitate an adaptive approach to estimating, in an optimal mean-square-error manner, a measure of actual system performance. The technique leads to an iterative, real-time measure. This proposed performance measure is conveniently formed by weighting conditional error covariance matrices that are obtained from the adaptive filter. Lastly, an adaptive Kalman filtering procedure is presented which minimizes the risk in system identification. Author