Smoothing the response of a gain scheduled autopilot for flexible flight vehicle by online tuning based on neural networks

Abstract

The challenging task of designing an auto pilot for non linear time varying flexible flight vehicle system under conflicting design requirements requires a gain scheduled controller throughout the flight. But still the final solution is bound to end up in some compromises with proper gain scheduling. The switching from one gain to another gain value creates disturbances in the flight and therefore this transition of gains must be smoothened. This is achieved by introducing an adaptive pid controller based on neural network as a supplementary to the main controller.