An algorithm for the traffic assignment problem

An optimization algorithm for the traffic assignment problem is presented. It is based on determining, for the flow corresponding to each origin at a time, a set of disjoint circuits with the most negative total marginal cost, followed by a unidimensional optimization along these circuits. The proposed method was compared to other well known approaches on randomly generated networks. In most cases the proposed method proved to be highly efficient, particularly when appropriate tighter stopping rules were used. In this sense, it was seen that when stopping rules were based on insufficiently tight tolerances, substantial errors could appear in the equilibrium traffic assignment.