 |
Center for Nonlinear Studies |
We present work in two problems in disaster relief. The first problems models the delivery of aid to rural areas. Because of their distance from supply warehouses and damaged infrastructure it is uncertain day-to-day whether a beneficiary location will be accessible. In some settings, relief organizations collaborate by assigning each beneficiary location to a single relief organization; organizations subsequently provide this assistance independently. We describe a two-stage stochastic integer programming model for assigning relief organizations to beneficiary locations to maximize the probability of successfully visiting beneficiaries, each of which have uncertain accessibility. Using sample average approximation to solve the model, we show preliminary results on randomly generated test cases and describe future work on the multi stage stochastic problem of managing individual relief organizations after they are assigned beneficiaries. The second problem is motivated by relief organizations providing local transportation of supplies in urban settings, where goods are delivered frequently and multi-stop routes are possible. We describe formulations and heuristics for the Capacitated Vehicle Routing Problem with Pickups, Deliveries, and Service Choice (CVRPPD-SD). In this problem, a fleet of vehicles must deliver a single good, such as surplus food, from pickup to delivery locations. Unlike the standard pickup and delivery problem, pickup and delivery locations are not paired a priori. Delivery locations are chosen from a candidate pool with a fairness objective, and routing minimizes the latest time of delivery. CVRPPD has seen limited work in the vehicle routing literature, and the fairness objectives and service choice have not been studied previously.
Host: Russell Bent, Energy and Infrastructure Analysis D-4/Decision Applications