A Stochastic Approach to the Vehicle Allocation Problem with Uncertainty in Demands

The Vehicle Allocation Problem models the allocation of vehicles in a space-time network in road freight transport. A two-stage extension with a heterogeneous fleet and uncertain demand has been proposed: travel commitments are defined in the first stage; in the second, after the scenario, trips are adjusted and the chartering of extra vehicles is admitted. The model was described and experiments were carried out in small and large instances with up to 200 Monte Carlo scenarios, evaluating times, optimal profit RP, VSS and EVPI in absolute terms and relative to RP, with sensitivity to the profit of extra vehicles. In small instances, the solution is usually fast; In large cases, the effort is in the order of minutes, with practical unfeasibility above 100 scenarios. Relative gains are moderate and grow with less attractive chartering: averages of VSS and EVPI in the order of 0.5-3% over RP; with maximum values of 6% and 13%, respectively.