SUMMARYThis paper presents a direct adaptive recon"gurable #ight control approach and demonstrates its e!ectiveness via an application to an advanced tailless "ghter aircraft. The recon"gurable control law is based on a dynamic inversion controller in an explicit model following architecture. An on-line neural network is used to adaptively regulate the error between the desired response model and the actual vehicle response. An on-line control allocation scheme generates individual control e!ector commands to yield the moments commanded by the controller, while prioritizing critical axes and optimizing performance objectives such as maneuver load alleviation. An on-line system identi"cation module generates estimates of the vehicle's stability and control derivatives for use in control allocation and command limiting. The recon"gurable control laws are demonstrated by comparing their performance to a dynamic inversion control law when unknown failure/damage are induced.

This paper proposes a general structure for aircraft control at whose core are several blocks of dynamic inversion of the controlled system. This control law is then applied to a nonlinear model of NASA's High Angle-of-attack Research Vehicle (HARV) which is currently being studied jointly by NASA-Ames, NASA-AmesDryden, and NASA-Langley. The control law itself uses the complete nonlinear aero data base, and has no restrictions on the manner in which the control inputs enter the dynamic equations. An algorithm which handles surface limiting while ensuring a solution to the moment equation inverse is incorporated into the control law. The excellent tracking performance of the control law is demonstrated via the simulation of a couple selected supermaneuvers.

No abstract

Many current and future aircraft axe open-loop unstable, and have many control surfaces. When some of these control surfaces are damaged, the remaining control surfaces may still be capable of controlling the aircraft, if the controller is reconfigured quickly enough.To speed up the parameter identification, signals can be injected to ensure sufficient excitation. This signal injection must be done in a way that does not annoy the pilot. The signal injection must also not counter the goals of the control allocation, which used the redundant surfaces to obtain the desired control moments while optimizing some cost function. This paper describes a method for rapid identification of parameters. Signal injection is used in a way that has no effect on the ideal system, since the redundant surfaces are moved in ways that do not affect the rigid body motion of the aircraft. The injected signals also have little or no effect on the other optimization objectives such as load alleviation and drag reduction.

No abstract