NONLINEAR FLIGHT CONTROL DESIGN FOR THE X-38 USING CMAC NEURAL NETWORKS

In this paper a control system architecture for attitude control of the X-38 reentry vehicle is proposed. At the core of the design is a nonlinear controller based on dynamic inversion. A CMAC neural network is used for in-flight compensation of uncertain model dynamics. Network approximation errors are accounted for by adding an adaptive bounding term to the control law. Boundedness of all signals in the closed loop is proven via Lyapunov theory. This control architecture is chosen to improve the robustness of the baseline inversion controller. Further, it incorporates some flexibility into the design with regard to controller adaptations to varying vehicle models and mission requirements. Simulation results show that improved tracking performance is achieved by means of CMAC network augmentation.