Vehicle Routing with Stochastic Demands and Partial Reoptimization

We consider the vehicle routing problem with stochastic demands (VRPSD), a in which customer are known distribution at route planning stage and revealed during execution upon arrival each customer. A long-standing open question on VRPSD concerns benefits of allowing, execution, partial reordering planned visits. Given practical importance this growing interest under optimal restocking, we study recourse policy as switch policy. The is canonical reoptimization that permits only pairs successive customers to be reordered. jointly preventive restocking introduce branch-cut-and-price algorithm compute priori plans. This features pricing routines where value functions represent expected cost-to-go along routes for all possible states decisions. To ensure tractability, adopt strategy combines elementary completion bounds varying complexity, solve without relying dominance rules. Our numerical experiments demonstrate effectiveness solving instances up 50 customers. Notably, they also give us new insights into reoptimization. enables significant cost savings over when come from an built deterministic approximation data, important scenario given difficulty finding solutions. smaller comparing solutions obtained both policies. As it appears, further may require joint reassignment visits among vehicles context permits.