Precise interception of a maneuvering target by a missile

This paper presents a new error critic for neurocontrol of highly nonlinear systems. The unavailability of the target outputs of a neural network controller constitutes a difficulty in tuning the adjustable weights and biases and deserves a careful consideration to improve the tracking performance. The method described aims to drive the system into a predefined sliding regime through the use of a novel error measure. The controller is an ordinary feedforward neural network, the tuning strategy is the error backpropagation, and the contribution of the paper is the proof of precise tracking by using the proposed error critic in such a standard neurocontrol configuration. The results obtained justify the performance claims of the paper without requiring the analytical details of system under control