Remote controller design of networked control systems based on self-constructing fuzzy neural network

A self-constructing fuzzy neural network (SCFNN) is proposed in this paper to design remote controller in networked control systems (NCSs). The structure and parameter learning phases are preformed concurrently in the SCFNN. The structure learning is used to obtain a proper fuzzy partition of input space, while the parameter learning is used to adjust parameters of the membership function and weights of the consequent part of the fuzzy rules based on the supervised gradient descent method. The initial SCFNN consists of input and output nodes only. In the learning process the nodes of the middle layers, which correspond to the membership functions and the fuzzy rules, are created gradually, so a set of fuzzy rules is achieved dynamically. Numerical results on a test system using Profibus-DP network are presented and compared with results of the modified Ziegler-Nichols method. The results show the effectiveness of SCFNN in designing remote controller for NCSs without any prior knowledge on network-induced delay.