A neural state estimator with bounded errors for nonlinear systems

A neural state estimator is described, acting on discrete-time nonlinear sy stems with noisy measurement channels. A sliding-window quadratic estimatio n cost function is considered and the measurement noise is assumed to be ad ditive. No probabilistic assumptions are made on the measurement noise nor on the initial state. Novel theoretical convergence results are developed f or the error bounds of both the optimal and the neural approximate estimato rs. To ensure the convergence properties of the neural estimator, a minimax tuning technique is used. The approximate estimator can be designed off li ne in such a may as to enable it to process on line any possible measure pa ttern almost instantly.