An adaptive extended Kalman filter using artificial neural networks

Develops an adaptive state-estimation technique using artificial neural networks, referred to as a neuro-observer. The neuro-observer is an extended Kalman filter structure that has its state-coupling function augmented by an artificial neural network that captures the unmodeled dynamics. The neural network of the neuro-observer trains on-line using an extended Kalman filter training paradigm. Improvement in the system model then provides for a more accurate state estimate in the feedback loop, thus enhancing the control signal so that the system behaves in a closer to optimal fashion.