Optimal type II Fuzzy neural network controller for eight-rotor MAV

This paper focuses on modeling and intelligent control of the new eight-rotor MAV which is used to solve the problem of low coefficient proportion between lift and gravity for QuadrotorMAV. The dynamic and kinematical modeling for the eight-rotor MAV.Neuro-Fuzzy adaptive controller is proposed which is composed of two type-II fuzzy neural networks (T-IIFNNs) and one PD controller: The PD controller is adopted to control the attitude, one of the T-IIFNNs is designed to learn the inverse model of eight-rotor MAV on-line, the other one is the copy of the former one to compensate for model errors and external disturbances, both structure and parameters of T-IIFNNs are tuned on-line at the same time, and then the stability of the eight-rotor MAV closed-loop control system is proved using Lyapunov stability theory. Meanwhile, in order to reduce the computation work, the type-reduction and model construction process have been improved. For the issue of type reduction, a novel improved EKM algorithm is developed for improving the EKM algorithm. The proposed algorithm provides two improvements on the EKM algorithm. For the issue of rules redundant, the concept of normalized difference is proposed to describe the change of adjacent singular value so as to reflect the essential differences between redundant rules and important rules. Then the number of effective singular can be determined according to its critical point, and the type-2 fuzzy model is constructed with rules located by TLS decomposition. Finally, the validity of the proposed control method has been verified through real-time experiments. The experimental results show that the performance of Neuro-Fuzzy adaptive controller performs very well under sensor noise and external disturbances.