Fuzzy control of telecommunications networks using learning technique

Communications services and traffic have become varied with the development of communications and progress in telecommunications network control for call acceptance and routing has been desired. In implementing flexible control in communications networks, which are complicated systems, introduction of fuzzy control techniques which is an application of knowledge engineering is considered to be effective. To realize fuzzy control efficiently, it is necessary to establish appropriately membership functions that are quantitatively homologous with the target of control. However, finding and identifying such membership functions are difficult tasks. From this viewpoint, this paper proposes a method of tuning a membership function by learning and investigates call acceptance control in detail as an example. By this method, the input values to the fuzzy system and the revenues of the network are measured and optimal values are obtained automatically for various parameters of the membership functions by learning based on these data and the past tuning history. In this study, the efficiency of this tuning method is evaluated by simulation. As a result, it has been shown that by this method membership functions converge to appropriate forms and good control characteristics can be obtained.