Observer based iterative neural network model inversion

Recently model based techniques have become wide spread in solving measurement, control, identification, etc. problems. For measurement data evaluation and for controller design also the so called inverse models are of considerable interest. In this paper a technique to perform neural network inversion is introduced. For discrete time inputs the proposed method provides good performance if the iterative inversion is fast enough compared to system variations, i.e. the iteration is convergent within the sampling period applied. The proposed method can be considered also as a simple nonlinear state observer, which reconstructs the selected inputs of the neural network from its outputs.