AIAA Journal · 1966 · 99 citations · 3 references
ComputingComputer ScienceStatistical LearningTime Series AnalysisUncertainty QuantificationError EstimationKalman FilteringMeasurement EstimationMeasurement UncertaintySensor NoiseState EstimationArtificial NoiseData ProcessingEstimation Of Measurement ErrorLinearizationLinearization TheoryNonlinear EstimationNonlinear State EstimationNonlinearity
It is shown that the usual linearization approach to sequential estimation is seriously inadequate if the measurement function nonlinearity is comparable to the measurement error. The essential difficulty is not that the estimated state, after processing a data point, is seriously affected by nonlinearity, but rather that the calculated covariance matrix of estimation errors becomes inaccurate, thus making future data processing nonoptimal, A powerful iteration scheme is described and its limitations noted. The bias nature of measurement function nonlinearity is also explored. A refinement in forming the measurement residual is introduced which reduces the nonlinearity in the measurement residual to the lowest level that appears practical. The suggestion of adding artificial noise to measurement data when the nonlinearity problem is severe is introduced.
3
Nonlinear estimation techniques
S. R. Neal · IEEE Transactions on Automatic Control · 1968 · 11 citations