SCC-CIVIC-FA Track B: Community-Centric Pre-Disaster Mitigation with Unmanned Aerial and Marine Systems

点击翻译按钮获取中文摘要

Each year, floods, hurricanes, and wildfires result in over $125 billions of dollars of losses and loss of life. Unfortunately, Texas, the state with the greatest number of annual federally declared disasters, and over $100B in economic losses since 1980, is no exception. Often BIPOC and low-income communities are impacted the most. Low-cost ($1-12K) unmanned aerial systems (drones) and unmanned marine surface vehicles (robot boats), coupled with advances in artificial intelligence and geospatial software, could revolutionize how communities prepare, prevent, and minimize losses. However, Texas emergency managers currently lack the workforce and knowledge to investigate and implement these technologies in a meaningful way. Advances in disaster science are slow in part because researchers do not have access to comprehensive, longitudinal datasets to apply computer vision/machine learning (CV/ML) to the most pressing needs. This one-year, $384K pilot program under the direction of the Texas A&M Institute for a Disaster Resilient Texas will create a sustainable research-centric civic engagement cycle in three vulnerable communities: rural (Bryan), urban (Houston), coastal (Galveston). Emergency managers, working with research and development partners, will annually determine pressing needs. Approximately 90 students are expected to work in some form with five emergency management agencies, five universities including CMU and UC Berkeley, three companies, and two non-profits. The students, taken from the schools where 76% are economically disadvantaged, 23% African-American, and 57% Hispanic, will be trained to collect or process pre-disaster mitigation data. These activities will amplify their STEM and career certificate courses, robotics clubs, and incubator experiences. The data and data products will be immediately available to state and local pre-disaster mitigation agencies. Data in the first year can result in savings on the order of $21K per parcel by informing common planning decisions, such as protecting open space and buying out vulnerable housing. The research component will contribute to fundamental advances in disaster science, robotics, AI, and urban land use planning by providing access to data that can answer six fundamental research questions. It will create the largest comprehensive, longitudinal datasets of unmanned vehicle imagery for pre-disaster mitigation. The datasets will establish the trustworthiness of CV/ML for disaster science, develop new algorithms for recognition of vulnerabilities during different seasons and weather conditions, and further the fundamental understanding of transfer learning. Performance data will lead to an informatics-based model of sampling that captures the technical tradeoffs between accuracy, resolution, and frequency on identifying objects and scene understanding. This project is part of the CIVIC Innovation Challenge which is a collaboration of NSF, Department of Energy Vehicle Technology Office, Department of Homeland Security Science and Technology Directorate and Federal Emergency Management Agency.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.