Online adaptive fuzzy logic controller using genetic algorithm and neural network for Networked Control Systems

Networked Control Systems are used for controlling remote plants via shared data communication networks such as Ethernet. These systems have found many applications in industrial, medical and space sciences fields. However there are some drawbacks in these systems, which make them challenging to design. One of the most common problems in these systems is the stochastic time delay. Packet switching in internet brings about the randomly varying time delay and consequently makes these systems instable. Convenient controllers such as PID and PI type controllers which are just matched with a constant time delay could not be a solution for these systems. Fuzzy logic controllers due to their nonlinear characteristic which is compatible with these systems are potentially a wise option for their control purpose. Fuzzy logic controller could become adaptive by means of neural networks and beneficial to deal with the varying time delay problem. Further, they do have more capabilities to tackle packet dropouts and dynamically system variables. This paper introduces a novel control method which addresses the varying time delay problem effectively. This novel method suggests an online adaptive fuzzy logic controller which has been controlled and adapted through the neural network. This method takes the advantage of the genetic algorithm to optimize the membership functions for its fuzzy logic controller. This designed controller is applied to an AC 400 W servo motor as a remote plant in order to control its position via Ethernet. The measurement of round-trip time (RTT) is used to estimate the online time delay as a parameter in online adaptive fuzzy logic controller. The rule-based table of designed fuzzy logic controller rotates in relation to this estimated time delay. The value of rotating is obtained from a trained neural network. Comparison of simulation results for different controllers indicates that this novel designed controller provides a better performance over the v- rying time delay. The proposed method follows the input easily, despite classical methods which result in an unstable system especially over the large time delays as large as 600 ms. Results get even more improved when genetic algorithm is applied to fuzzy logic controller.