Nonsmooth optimization methods for parallel decomposition of multicommodity flow problems (Q1308671)

Multicommodity network flow problems are tackled by using a resource directive decomposition technique. First, decision variables representing a feasible sharing of the resource among all different commodities are introduced. Then the problem is decomposed in mincost single commodity network flow problems. Finally, a master problem is solved for finding an improved sharing of the resources. Since the objective function of the master problem is non-differentiable, a bundle algorithm is used for solving it. The adopted decomposition scheme is particularly suitable for parallel implementation. The results are also extended to the nonlinear, separable case.