Distributed adaptive routing control in communication networks using a temporal fuzzy classifier system

The application of a novel temporal fuzzy classifier system to the problem of distributed adaptive routing control in packet-switched communication networks is described. The proposed fuzzy classifier system augments classifiers with temporal tags for the purpose of explicitly incorporating a representation of time. This attribute is important for learning in environments with temporal features. It is also a desirable property in control of systems where state measurements are delayed or only available occasionally, and when the effects of control actions are delayed. Distributed adaptive routing control is one such problem. Experimental results are presented which demonstrate that temporal fuzzy classifier controllers can be evolved, which outperform extant routing algorithms in communication network simulations. Reasons for the improvement in performance are suggested and discussed.