A Network Based Learning Architecture for Fuzzy Logic Controllers

Tuning the parameters of fuzzy logic controllers is one of the most important parts of the design of these controllers and it has been extensively explored by researchers. Various techniques and algorithms have been utilised to fine tune the controller parameters. Designing controllers with the ability to retain and share the tuned parameters with other controllers have potential advantages on reducing the time required in tuning process. So far, however, there has been no research on the design of networked fuzzy controllers with the ability to retain the knowledge gained in tuning process and to provide a communication facility to enable the exchange of the acquired knowledge between controllers through a network. By expanding a previous work of the authors in auto-tuning fuzzy logic controllers, this paper proposes an original architecture for designing a network of fuzzy logic controllers with the capabilities of auto tuning and sharing parameters. To improve the performance, each controller automatically and progressively tunes its parameters and retains the acquired knowledge in its memory for the future when a similar set-point is assigned to the controller. At the same time the acquired knowledge is shared with the rest of the controllers on the network where the controllers can benefit from it in tuning their parameters. The result demonstrate that this method has a substantial impact on minimising the time required to tune the controllers.