A computational comparison of several nonlinear filters

Eight different filtering schemes, one linear and seven nonlinear, are applied to two nonlinear scalar dynamic systems which are forced by additive noise. The nonlinear measurement made of the state is corrupted by further additive noise. The various filters, derived using a continuous formulation, are simulated on a digital computer and are compared on the basis of their estimation-error history. In addition to the comparison, one system/filter combination is used to demonstrate the effects of varying the variance of the noise, of mismatched parameters, and of sampled measurements.