An on-line approach for ANFIS modelling and control of a flexible manoeuvring system

This paper presents an on-line nonlinear dynamic modelling and control approach based on adaptive neuro-fuzzy inference system (ANFIS) for a twin rotor multi-input multi-output system (TRMS), in the vertical plane motion. The TRMS can be considered as a flexible aerodynamic test rig that resembles the behaviour of a helicopter in hovering mode. The TRMS and similar manoeuvring systems are often subjected to random disturbances arising from various sources such as driving motors and external environmental sources. For such highly nonlinear systems with varying operating conditions, adaptive control approaches are suitable tools to cope with plant uncertainties. A model inverse control of the TRMS is developed using on-line ANFIS learning algorithm. The consequent and antecedent parameters of a first order Takagi-Sugeno fuzzy inference system are optimised on-line using recursive least squares and gradient descent algorithms, respectively. The optimal initialization of the ANFIS parameters is achieved through an off-line training process. The developed strategy is compared to other control laws in terms of tracking performance and disturbance rejection. The obtained simulation results demonstrate the efficiency of the on-line control scheme.