Gnc-95 Neural Networks for Stable Adaptive Control of Air-to-air Missiles

A recently developed approach to the control of uncertain nonlinear systems uses neural network technology to improve upon feedback linearization. In the proposed architecture, a neural network adaptively cancels linearization errors through on-line learning. One feature which distinguishes this approach is its use of Lyapunov stability theory to guarantee closed-loop stability. In this paper, the authors apply this methodology to design a longitudinal missile autopilot. First, a simple scheme for controlling angle of attack by approximate inversion of the longitudinal dynamics is proposed. This nonlinear control system is then augmented by the addition of an on-line neural network. Finally, the resulting control law is demonstrated by nonlinear simulation of an air-to-air missile.