The Traveling Salesman Problem with Imperfect Information with Application in Disaster Relief Tour Planning

Abstract

Many in the disaster response community have begun to explore ways to use information posted on social media platforms to identify a larger set of needs in a shorter amount of time following a disaster. However, needs communicated through social media platforms have initially not been verified so many within the emergency response community remain skeptical over the usefulness of such information. Consequently, as emergency managers consider whether to incorporate social media data in disaster planning efforts, a key tradeoff must be assessed. Confidence in the accuracy of needs to which resources are allocated is increased when information discovered on social media is ignored, but there is potential to leave populations that have not yet been discovered through traditional means unassisted. This paper introduces a new problem framework that describes a formal method for quantitatively assessing the impact of including unverified information in disaster relief planning. The usefulness of the framework is demonstrated in the context of the traveling salesman problem. A decision approach that considers social media information is compared to one that does not on the basis of total response time of resulting tours. A case study that considers variations in report accuracy and quantity for uniformly distributed demand instances is presented.