A real time learning control for broadband ISDN

In such integrated services networks, the grades of service (GOS) is classified into call level and cell level. To satisfy both GOSs in the ATM (asynchronous transfer mode) networks for BISDN, the learning approach seems to be suitable, because the precise characteristics of the source traffic are unknown and this approach can be adapted for such complex systems. The authors propose a learning control, using the method of potential functions to identify the system characteristics and to estimate the blocking probability for the call-level GOS. As for maintaining the GOS for cell level, in order to estimate the cell delay for the cell-level GOS, in real time, they propose the equivalent MMPP (Markov modulated Poisson process) method. A numerical example for a video/data system is also given to show the effectiveness of this learning control