Predictive Analytics for Wide Area Search and Rescue (PAWSAR): FY23 HADR Technical Investment Program
Abstract
When a disaster such as a hurricane strikes a large geographic area, specially trained urban search and rescue (USAR) teams may conduct a wide area search, looking for an unknown number of people within an unclear boundary. Allocating the right number of USAR teams to a response, identifying where they should search, and how to subdivide the search area are tasks performed by experienced personnel. In this paper, we describe an effort to establish a predictive analytics framework that can provide these decision makers with a quantified understanding of potential impacts, time, and resources required to conduct searches within a given area, and how an urban area may be best divided into subareas of equal work. Using data collected by USAR teams from past hurricane responses, we develop prototype models that 1) estimate structural damage based on other hurricane predictive models, 2) estimate the time required to search structures with a set of damage distributions, and 3) partition a city into areas of equal work. We demonstrate how these models can be used in a prototype operational pipeline that could be executed in sync with the USAR operational cycle. Future work entails validation of the models against future hurricane impacts and expanding the city partitioning algorithm to derive work estimation from the modeled damage and search time predictions. Numerous supporting analyses conducted as part of this work have also yielded insights into how USAR teams conduct wide area search operations. These insights may be immediately applicable to the community's perpetual continuous improvement model.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 15, 2023
- Accession Number
- AD1220014
Entities
People
- Chad L. Council
- Dieter W. Schuldt
- Jeff Liu
Organizations
- Massachusetts Institute of Technology