Disaster Response Routing with Social Data in Stylized Demand Scenarios

摘要

Social media has changed the landscape of disaster response. In recent years, it has become somewhat of a lifeline for individuals in times of crisis, operating as a platform where users can notify others of their safety status, provide public updates on crisis situations, and alert their communities about unsafe conditions. This publicly shared social media data can be used by emergency managers as a quick way to obtain information detailing the location and needs of an impacted community. These data, however, are not initially verified and thus it becomes critical to ascertain the advantages and disadvantages of employing it in a disaster relief routing plan. A capacitated vehicle routing problem is presented in which traditional verified requests and social media unverified requests are available at the time of planning. Two alternative decision approaches for solving this problem are presented: one that integrates social data into the disaster response plan a priori and one that delays the integration of social data until a recourse phase. We present the results of a computational study comparing these strategies across a variety of stylized location and request distributions for the two types of demand.