Solving multicommodity integral flow problems by simulated allocation

Abstract

The paper discusses a stochastic approach, called Simulated Allocation (SA), to the capacitated and uncapacitated multicommodity integral flow allocation problems. Since such problems are in general NP?complete, effective heuristic approaches are of interest, and SA is one of them. The idea of SA has its origin in discrete event simulation of alternative call routing in telecommunication networks. We describe an implementation of SA and illustrate its effectiveness by means of numerical examples including multiple knapsack problem and network design. Numerical experiments indicate that SA is capable of bypassing local minima of the problem objective function, and allows for achieving reasonable solutions in a reasonable time.