A learning control method for communications traffic system

The convergence of a learning control method for resource assignment control in integrated services telecommunications networks is proved. General black-box modeling is used to cope with the vast variation in traffic characteristics. To reduce the heavy calculational burden required by general modeling, a learning control approach is introduced. The approach promotes efficient learning by adopting the method of potential functions for online system identification and adaptive learning region adjustment. The method´s applicability to arbitrary traffic distributions is demonstrated for the problem of trunk reservation control of multispeed circuit-switched links