Survey of City-Wide Homelessness Detection Through Environmental Sensing

Abstract

The growing homelessness crisis in the U.S. presents complex social, economic, and public health challenges, straining shelters, healthcare, and social services while limiting effective interventions. Traditional assessment methods struggle to capture its dynamic, dispersed nature, highlighting the need for scalable, data-driven detection. This survey explores computational approaches across four domains: (1) computer vision and deep learning to identify encampments and urban indicators of homelessness, (2) air quality sensing via fixed, mobile, and crowdsourced deployments to assess environmental risks, (3) IoT and edge computing for real-time urban monitoring, and (4) pedestrian behavior analysis to understand mobility patterns and interactions. Despite advancements, challenges persist in computational constraints, data privacy, accurate environmental measurement, and adaptability. This survey synthesizes recent research, identifies key gaps, and highlights opportunities to enhance homelessness detection, optimize resource allocation, and improve urban planning and social support systems for equitable aid distribution and better neighborhood conditions.