Behaviour of F learning automata as multicriteria routing agents in connection oriented networks

The use of intelligent algorithms based on learning automata (LA) has proved to be very efficient for traffic routing control. However, thus far these algorithms have been focused only on unimodal routing problem in connection or packet oriented networks. The problem of multicriteria routing in communication networks is inherently NP-hard. We present a novel heuristic approach to this problem based on the concept of learning automata in fuzzy environments. We call these automata the F type learning automata. Well known learning automata of P, Q and S types are special cases of F type automata. Among others we prove that these automata are nonabsorbing. A general routing scheme enabling implementation of different routing strategies including Best Quality Routing (BQR) and Least Cost Routing (LCR) has been defined and evaluated through simulation results obtained for connection oriented telecommunication network. The behaviour of different t and s norms has been investigated as well.