A method for controlling steady flight of an unmanned aircraft, comprising the following steps: acquiring real-time flight operation data of an unmanned aircraft itself by means of a posture sensor, a position sensor and an altitude sensor loaded on the unmanned aircraft, and correspondingly performing parsing processing, by means of a loaded processor, on a kinematic problem of the aircraft, so as to establish a kinetic model of the aircraft (S1); according to a multi-layer annihilation neurodynamic method, designing a controller for the unmanned aircraft (S2); using the acquired real-time operation data of the aircraft and target posture data to solve an output control quantity of an electric motor of the aircraft by means of a designed multi-layer annihilation neural network controller (S3); and delivering a solving result to a motor speed governor of the aircraft, and realizing control over the motor power according to a relationship between the control quantity solved by the controller and the motor power of a multi-rotor unmanned aerial vehicle, so as to control the motion of the unmanned aircraft (S4). Based on a multi-layer annihilation neurodynamic method, the correct solution to a problem can be approximated rapidly, accurately and in real time, and a time-variable problem can be significantly solved.