Recurrent Interval Type-2 Fuzzy Control of 2-DOF Helicopter With Finite Time Training Algorithm

TY - GEN

T1 - Recurrent Interval Type-2 Fuzzy Control of 2-DOF Helicopter With Finite Time Training Algorithm

AU - Kayacan, Erdal

AU - Ahmadieh Khanesar, Mojbata

PY - 2016/6

Y1 - 2016/6

N2 - This study presents the decentralized control of a 2-DOF helicopter by designing a recurrent interval type-2 fuzzy neural network (RIT2FNN). The main aim of the proposed controller is to force the pitch and yaw angles follow a desired trajectory by using a finite time adaptation law. The proposed control signal is composed of two terms: the output of the RIT2FNN and the control signal generated by a conventional proportional-derivative (PD) controller. In the beginning, since the initial conditions of the RIT2FNN are randomly selected and may not be appropriate, the PD controller is responsible for the control of the system. However, the stable adaptation laws, which benefit from sliding mode control theory, train the parameters of the RIT2FNN. Since the adaptation laws are guaranteed to converge in finite time, the parameters of the RIT2FNN converge to their appropriate values. Meanwhile, the PD controller participates less in the control process and the RIT2FNN becomes the dominant controller of the system. The proposed control method is promising when dealing with highly nonlinear real-time systems which have to operate under uncertain working environment.

AB - This study presents the decentralized control of a 2-DOF helicopter by designing a recurrent interval type-2 fuzzy neural network (RIT2FNN). The main aim of the proposed controller is to force the pitch and yaw angles follow a desired trajectory by using a finite time adaptation law. The proposed control signal is composed of two terms: the output of the RIT2FNN and the control signal generated by a conventional proportional-derivative (PD) controller. In the beginning, since the initial conditions of the RIT2FNN are randomly selected and may not be appropriate, the PD controller is responsible for the control of the system. However, the stable adaptation laws, which benefit from sliding mode control theory, train the parameters of the RIT2FNN. Since the adaptation laws are guaranteed to converge in finite time, the parameters of the RIT2FNN converge to their appropriate values. Meanwhile, the PD controller participates less in the control process and the RIT2FNN becomes the dominant controller of the system. The proposed control method is promising when dealing with highly nonlinear real-time systems which have to operate under uncertain working environment.

KW - 2-DOF helicopter model

KW - Type-2 fuzzy neural networks

KW - adaptive intelligent control

KW - feedback error learning

KW - recurrent

KW - sliding mode control

UR - http://www.scopus.com/inward/record.url?scp=84991074980&partnerID=8YFLogxK

U2 - 10.1016/j.ifacol.2016.07.977

DO - 10.1016/j.ifacol.2016.07.977

M3 - Article in proceedings

SN - 2405-8963

VL - 49

SP - 293

EP - 299

BT - IFAC-PapersOnLine

PB - Elsevier

T2 - 12th IFAC International Workshop on Adaptation and Learning in Control and Signal Processing

Y2 - 29 June 2016 through 1 July 2016

ER -