Fully Tuned Radial Basis Function Neural Networks for Flight Control

Over the last four decades, adaptive control has evolved as a powerful methodology for designing feedback controllers of nonlinear systems. However, most of these studies assume that the system nonlinearities are known a priori, which is generally not applicable in the real world. To overcome this drawback, from 1990s, there has been a tremendous amount of activity in applying Neural Networks (NNs) for adaptive control. With their powerful ability to approximate nonlinear functions, neuro-controllers can implement the expected objectives by canceling or learning the unknown nonlinearities of the systems to be controlled. NNs are especially suitable for the adaptive flight control applications where the system dynamics are dominated by the unknown nonlinearities. Moreover, among different choices of network structures, Radial Basis Function Network (RBFN) has shown its potential for on-line identification and control, and hence arouses much research interest.

Control laws and design methods incorporating ANNs have been intensively studied in the area of aircraft flight control. In [17], Calise et al. have summarized some current research efforts of applying NN technology for flight control system design, with emphasis on nonlinear adaptive control. It has been shown that NN with on-line learning can adapt to aircraft dynamics which is poorly known or rapidly changing. However, in most of these applications, feedforward network with BP learning algorithm or its extensions has been the main paradigm, and there are only limited papers which explore the application of RBFN.