Nonlinear Adaptive Control of High Performance Aircraft Using Neural Networks

The aerodynamic coefficients governing the forces and moments acting upon high performance aircraft such as the AFTI-F16 are inherently both nonlinear and uncertain. Accurate coefficients are absolutely vital for design, yet these parameters are analytically intractable. Hence, empirical estimates must be employed despite the associated experimentation error which plagues such techniques. This paper delves into possible nonlinear control methods to surmount these uncertainties while guaranteeing satisfactory performance. Specifically, the advantages and disadvantages of sliding control, adaptive spline interpolation, modified radial Gaussian neural networks, and activated Gaussian node neural networks will be considered. G-command performance will be judged with respect to both static and fluctuating parameter uncertainty. Results indicate that the use of neural networks yields significant performance advantages and superior parameter identification.