Neural Guidance Control for Aircraft Based on Differential Flatness

Abstract

Increasing air traffic density and new strict environmental regulations have been driving researchers to develop more advanced guidance systems for civil transportation aircraft to perform complex maneuvers. In this paper, a neural guidance control scheme is proposed to make an aircraft perform more complex trajectories. Based on differential flatness, the inertial position of the aircraft can be one of the flat outputs for its flight guidance dynamics such that the corresponding guidance control input can be derived from the desired flight trajectory. However, the flatness property of the guidance dynamics implies that the guidance control input cannot be obtained analytically from the information of reference trajectories. To tackle the numerical difficulty due to its implicit flatness, the neural network is employed to construct the relationship between the flat output and the guidance commands submitted to the autopilot system for the aircraft to track the reference trajectories. An additional adaptive capability is necessary to compensate for the model approximations, disturbances, and neural-network limitations.