Stochastic Programming Models. Applications to Logistics, Bikesharing and Production Management

This thesis addresses four problems characterized by the presence of uncertainty. The first two problems analyze a distribution system with uncertain demand in which transshipment and backordering are allowed, the third problem studies the allocation and rebalancing activities in a bikesharing system under uncertain bike demand, and the fourth problem deals with workforce planning decisions considering workers’ stochastic learning curves. For all these applications, stochastic programming formulations are proposed and the importance of considering uncertainty explicitly in the models is assessed. For the two distribution problems, complexity results, optimal policies for special cases, and a rolling horizon heuristic are proposed. For the bikesharing problem, a comparison of the solution of the stochastic program with the one of the real implemented system is presented. For the workforce planning problem, a rigorously designed computational study and statistical analysis are used to derive managerial insights. Finally, a methodology to obtain monotonic chains of lower bounds for stochastic programs is proposed and preliminary results are presented.