Collaborative Research: Advancing Supply Prepositioning for Emergency Preparedness by Social Sensing

可预见的灾害，如飓风、暴风雪和干旱，可在其登陆前几天发现。为了减轻这些灾害的后果，人道主义机构和政府机构可以利用预报信息开展备灾活动。然而，一个尚未解决的挑战是获得准确的估计时空异质供应需求的供应预置。为了应对这一挑战，该研究计划采用一种方法，将社交媒体和模拟数据整合起来，以预测对人道主义物资的需求。预计由此产生的模型和原型工具将有助于支持改善与公众的沟通，优化供应预置战略，并促进决策，以帮助预防或减轻极端灾害的影响。这项研究还将导致新的教育模块和软件，用于教学本科生和研究生如何设计，分析和实施人道主义物流系统。这项研究将调查整合社交媒体数据和基于代理的模拟的潜力，以减少与灾后救援物资需求相关的不确定性。该项目将基于社会媒体分析和基于代理的模拟的集成开发一种新的社会感知方法。可信信息将从社会媒体中提取一个新的时空贝叶斯方法。其次，本项目将建立一个新的预测驱动的多商品供应预置模型，将马尔可夫和决策理论的观点在随机优化框架。第三，供应预置的结果将通过社会交付来实现，这是一种新兴的交付计划，利用车队和志愿者车辆的交付能力。不同的优化分解技术将被检查和开发，以处理所考虑的问题的大规模性质。为了检验这些方法的有效性，将进行一系列案例研究。

Foreseen disasters, such as hurricanes, snowstorms and droughts, can be detected several days prior to their landfall. To ameliorate the consequences of these disasters, humanitarian and governmental agencies can utilize forecast information to conduct preparedness activities. However, one unresolved challenge is to obtain accurate estimation of spatial-temporal heterogeneous supply demand for supply prepositioning. To tackle this challenge, this research plans an approach which integrates data from social media and simulation to predict demand for humanitarian supplies. Resulting models and prototype tools are expected to contribute to support for improving communication with the public, for optimizing supply prepositioning strategies, and for facilitating decision-making to help prevent or ameliorate the impacts of extreme disasters. This research will also lead to new educational modules and software useful for teaching undergraduate and graduate students how to design, analyze and implement a humanitarian logistics system. This research will investigate the potential for integrating social media data and agent-based simulation to reduce uncertainty associated with the demand for post-disaster relief supplies. This project will develop a new social sensing approach based on the integration of social media analysis and agent-based simulation. Trustworthy information will be extracted from social media by a new spatio-temporal Bayesian method. Second, this project will build a new forecast-driven multi-commodity supply prepositioning model that incorporates Markovian and decision theoretic perspectives within a stochastic optimization framework. Third, the outcome of the supply prepositioning will be realized by social delivery, an emerging delivery scheme that leverages delivery power from both fleet and volunteers' vehicles. Different optimization decomposition techniques will be examined and developed to handle the large-scale nature of the considered problem. To test the validity of the approaches, a range of case studies will be conducted.