Algoritmos para atribuição de tráfego em redes de filas finitas dependentes do Estado

Abstract

The traffic models must be detailed enough to capture the fundamentaldetails involved. Also these models must be simple enough to fit well intoan optimization framework. In this dissertation we investigate a traffic assignment model, the classical System Optimum (SO) model, which is based on finite queueing networks. The queueing model chosen represents quite well those situations with congestion effects in which the speed of an user decays with the increase of the number of users simultaneously in the system. We obtained results with and evolutionary heuristics that indicated that the traffic assignments make sense and are robust.