Research On Dynamic Scheduling Methods For Post-disaster Emergency Resources Under Uncertain Conditions

In order to deal with major natural disasters that usually occur in sudden and to minimize the loss of property and to protect the lives of people,we need to make scientific and efficient emergency resource scheduling plans in the first time after the occur of a disaster,so that to provide enough support on emergency facilities,relief personnel,tools,supplies,and so on.From the properties of post-disaster environments,the following two aspects should particularly be considered:The first is uncertainty.Major natural disasters seriously damage the traffic,communications and other infrastructure,leading to difficulties in the transmission and collection of disaster information,and making it difficult for the decision-makers to comprehensively and accurately grasp the disaster information in a short time.In this paper,a robust optimization method is used to deal with various uncertainties,such as demand,supply and transportation time.The second is dynamic.In post-disaster stage,some information such as demand normally keeps varying especially after secondary disasters occur,and uncertain information is also gradually revealed.These unceasingly updated information itself is dynamic to decision-makers.In addition,as post-disaster relief work usually needs a long time to finish,the available levels of various emergency resources,the demands of affected areas,and the infrastructure status of the disaster areas will change during this time.In this paper,the model predictive control method is innovatively introduced to deal with the dynamic of disaster information.According to the above two aspects that should be considered,this paper mainly carried out the following research works:Firstly,the uncertainty of parameters such as demand should be considered in the formulation of emergency resource mobilization-transportation scheduling plans.Taking into account the requirements of mobilizing the material supplying and medical capacity,as well as forward-delivery of the material and the backward-evacuation of the wounded,a multi-periods,multi-supply areas,multi-demand areas,multi-materials,multi-modes,transshipment permitted,limited-capacity emergency resource mobilization-transportation scheduling model is established to minimize the total cost.On this basis,two different robust optimization methods are adopted to deal with the uncertainty of demand parameters.The experimental results show that the solution based on Bertsimas is better than that based on Ben-Tal in most cases when the model is validated.Although the robust optimization model based on Ben-Tal has a great advantage in reducing the mobilization-transportation cost,but the primary goal after the earthquake is to save the wounded and reduce casualties.Therefore,the robust model based on Bertsimas will be more suitable for the actual situation after the earthquake.Secondly,when making a plan for emergency resource transportation,it is necessary to consider the inaccuracy of key parameters(i.e.,supply and demand)that gotten by forecast and the evolution of these parameters that exhibited with time progresses.A multi-period emergency transportation model whose objective is to minimize the total weighted unsatisfied demands on material and unserved wounded,in which both the delivery of relief commodities and the evacuation of critical population have been considered.Then,a rolling horizon framework based on the model predictive control(MPC)method is innovatively introduced.Furthermore,a revised relief transportation model that based on the MPC method is presented,as well as the corresponding adjustment policies to adapt to the evolution of supply and demand and to satisfy the real-time adjustment requirement caused by the inaccuracy of prediction on supply and demand.The experimental results show that,compared with the traditional method,the proposed method can effectively decrease the amount of total weighted unmet demand,and obviously eliminate the influence on the optimization objective caused by the inaccuracy of predictions.Thirdly,based on the two studies mentioned above,considering the uncertainty and dynamic of the post-disaster environment and the real-time adjustment requirements of the existing distribution plans according to the deviation between the predicted values and the actual(or observed)values of the input data,a rolling horizon-based framework that is based on the robust model predictive control(RMPC)approach is proposed to obtain robust relief distribution plans and adjust them in accordance with updated real-time information.The numerical experiments show that the impact incurred by the uncertainty of the demand is more obvious than that incurred by the uncertainty of the supply when applying the RMPC mechanism,and the introduction of RMPC can effectively mitigate the impact on the objective value when varying the setting of the budget of uncertainty.Besides,the accuracy of the expected value exhibits stronger effects on the optimization objective than the variability range when applying the MPC mechanism.Additionally,the advantage of RMPC model is obvious compared with the robust model,while its advantage is limited compared with MPC method.This study provides valuable solutions that can reduce the disturbance of uncertainty and dynamic on post-disaster emergency transportation plans,and will provide a good guide on making transportation plans on emergency resource.