摘要：AbstractWhen a region is hit by a severe disaster, humanitarian supplies must be provided to victims/evacuees efficiently throughout the entire disaster and post-disaster periods. The emergency packages include but not limited to food, water, sanitation supplies, medicine, medical equipment, etc. Delivery of the humanitarian aids from suppliers to shelters must be done within certain time limits. Though helicopters are used for transporting a fraction of the daily requirements, the capacity of small aircrafts limits the throughput of this mode. Ground transportation modes are still playing a major role in the humanitarian logistics. Because of the severe weather during and after disaster, e.g. storm after hurricane or after-shock after an earthquake, it is common to have failures on the road and infrastructure, such as flooding, surface cave-in and sedimentation, which may delay the traffic or even make part of the network unusable. The expected reliability of a route is one of the main variables when planning a trip. The route choice coheres with the choice of the departure time to reach a destination in time with an acceptable probability. To increase the possibility of providing supply to evacuees under uncertainty without disruptions is a challenging problem. An efficient and reliable routing and scheduling model needs to be developed for both disaster and post–disaster conditions. In this study, we address a sub-problem of the general humanitarian supply-chain problem, which is the humanitarian response planning for a fleet of vehicles with reliability considerations. Routing and scheduling of humanitarian supply transportation is formulated as a mathematical model. To apply this routing and scheduling method in real operations with on-line information, efficient algorithms are necessary. A genetic algorithm based heuristic is proposed to solve the problem in reasonable computational time. The performance of the mathematical optimization model and heuristic algorithm are evaluated using test networks. The results show that the proposed approach can provide prompt delivery while reducing the risk of undesirable delay caused by uncertainty. The algorithm can provide high quality solution within short computational time to fulfill the on-line operation requirement.
