The present invention discloses a hypersonic aerocraft neural network composite learning non-
backstepping control method. The technical problem is solved that a current hypersonic aerocraft control method is bad in practicality. The technical scheme comprises: performing transformation of an attitude subsystem strict
feedback form, obtaining an
output feedback form, employing a high-
gain observer to perform
estimation of newly defined variables, and providing basis for subsequent design of a controller; allowing the controller to consider the lump nondeterminacy of the
system, and only requiring one neural network to perform approximation, wherein the controller is simple in design and is convenient for
engineering realization; aiming at control of unknown cases of a
gain function, designing the controller based on the parameter
linearization expression mode; and introducing
system modeling errors, and constructing a neural network composite updating rule and a parameter adaptive composite updating rule to realize
fast tracking of a hypersonic aerocraft. The effective
estimation of unknown states is realized based on the high-
gain observer, the repeat design of the virtual controlled quantity is not needed so as to simple the design of the controller. the realization is easy, and the practicality is good.