A neural network computation algorithm for discrete-time linear system state estimation

A neurocomputing approach is developed to solve the problem of state estimation for a discrete-time, linear dynamic system. Dynamic optimization techniques are used to develop the online adaptation laws for modifying the weights and biases of a deterministic Hopfield neural network, which in turn produces the estimate of the system state when the net reaches its stationary point. Simulation results show that the proposed approach performs similarly to the Kalman filter. Due to the parallel computational mode of the neural net, the proposed approach is more attractive for real-time implementation, from the computational point of view, than classical estimators.<>