Neural networks based adaptive controller design for unmanned aircraft

This paper presents an adaptive controller using neural networks (NN) for a unmanned aircraft (UA). The NN-based controller is designed to track angle of attack command signal of the reference model. The nonlinear UA dynamic model, which is approximated by NN identifier with backpropagation algorithm, is given. The NN-based controller is generated and trained off-line, while learning adaptively on-line. Furthermore, a modified performance function is introduced to smooth the network response. The tracking performance of the proposed controller is validated both with and without parameter uncertainties.