Minimum covariance, minimax and minimum energy linear estimators
The estimators which minimize the error covariance for the filtering, prediction and smoothing of linear plants with Gaussian initial conditions and noises are well-known. We show that these same estimators arise when one seeks to minimize the maximum error assuming that initial conditions and noises are bounded in norm in an appropriate Hilbert space (minimax estimator). They also arise when one seeks the trajectory of least energy necessary to produce the given observations (minimum energy estimate).
KeywordsMinimum Energy Observation Noise Imum Energy Gaussian Random Vector Minimax Estimate
Unable to display preview. Download preview PDF.