Research has shown that neural networks can be used to improve on approxima
te dynamic inversion for control of uncertain nonlinear systems. In one arc
hitecture, the neural network adaptively cancels inversion errors through o
n-line learning. such learning is accomplished by a simple weight update ru
le derived from Lyapunov theory, thus assuring the stability of the closed-
loop system. This methodology is reviewed and extended to incorporate an im
portant class of neural networks with one sigmoidal hidden layer. An agile
antiair-missile autopilot is subsequently designed using this control schem
e. A control law based on approximate inversion of the nonlinear dynamics i
s presented. This control system is augmented by the addition of a neural n
etwork with on-line learning. Numerical results from a nonlinear agile anti
air-missile simulation demonstrate the effectiveness of the resulting autop
ilot.