A neuro-fuzzy method to parametric estimation with unknown-but-bounded-error

The determination of mathematical models consistent with observation and prior knowledge is important to many fields. When is necessary to estimate the unknown parameters models from inexact data, the information available on the various sources of error should be take into account to derive a proper estimator. This estimator is such that the influence of all kind of error should be minimised. In this paper, a neurofuzzy algorithm is proposed to estimate the system model parameters when the output error is considered unknown-but-bounded. A modified Hopfield's network is developed whose equilibrium points are the estimated model parameters. To aid the network convergence to these equilibrium points, a fuzzy controller is also developed. Simulation examples are presented to illustrate the performance of proposed algorithm.