Abstract: Aiming at the characteristics of non-linear, strong coupling, under-drive and multi-input multi-output of eight-rotor aircraft, the mathematic model is deduced by combining the eight-rotor dynamic characteristics. The relationship between the input and the output of each channel is decoupled by using the model. The RBF neural network adaptive PID control strategy is proposed, which can adaptively adjust the PID parameters according to the control effect, and has the ability of autonomous learning and self-adaptability. Through the matlab simulation model of eight-rotor and its simulation, the RBF neural network adaptive PID and traditional PID control effect are compared and analyzed. The results show that the former control effect is better than the latter, the system fastness, robustness and stability has been greatly improved.