Linear System State Estimation Using Hopfield Net

The purpose of this paper is to use human knowledge in dynamic optimization theory in combination with neural network to solve dynamic optimization problems. It is shown in this paper that one can perform linear state estimation using Hopfield neural net. This is done by formulating the state estimation problem as a dynamic optimization problem. It is found that the Hopfield net is a suitable computing device for this problem. In addition, the synaptic weights and bias vectors are computed online, using optimal control theory. Case studies are presented and the performance of the neural net based state estimator is compared to the Kalman filter performance.