Neuro-fuzzy Controller of Rigid-Flexible Manipulators with Intelligence Membership Function

This paper presents a new Neuro-fuzzy control system to control rigid-flexible manipulators. Enhancing the performance of fuzzy controller and intelligence in fuzzy and non-fuzzy units are the goals of this research. Proposed control system includes a fuzzy controller in the feedback and a neural network is the feed-forward. The network has the responsibility of estimating the inverse dynamic of device and then, the production of control command. Updating weighting coefficients of network is done on line using the fuzzy controller output. On the other hand, two dynamic recurrent neural networks are used for making fuzzy unit intelligent. Networks are responsible for regulating the main factors of membership functions in the fuzzy controller. The input of these networks is error and error change rate and their weights are done by using an error back-propagation algorithm. To verify the effectiveness of the proposed method, simulation is conducted for skilled manipulators with three interfaces which the end interface is flexible. System responses to step input and sinusoidal input are separately obtained for fuzzy controllers and proposed controller and compared. Comparison and studies indicate the effectiveness of the provided method.