Hemant Purohit
George Mason University, VA·Computer Science
During hurricanes and floods, evacuation and public shelters save lives. Still, emergency managers face hard choices: which shelters to open and when, how to allocate resources, and how many people will come.
- Fields
Computer Science
Engineering
Social Science
- Topics
emergency management
evacuation
public sheltering
- Grants
- 1 known grant, through Sep 2028
What they work on
- Build a dynamic knowledge graph using probabilistic graphical models that combines sensor data, citizen movement and infrastructure interdependencies such as power and transportation to predict risks to emergency shelters.
- Develop a neurosymbolic AI planning framework that accepts emergency managers' safety constraints and explains shelter resource allocation decisions in natural language.
- Test PCExplorer with the Virginia Beach Office of Emergency Management, including personalized shelter messages to residents and a feedback loop that captures their actions.
More about the research
Current decision tools rely on weather forecasts, flood risk assessments and past shelter use. They cannot track how residents' needs and movements change, or how failures in linked infrastructure such as transportation and power could disrupt shelters. Some residents are also less likely to seek shelter because of transportation problems, past experiences or doubts about services.
Why it matters
Coastal cities and emergency managers could use this system to decide which shelters to open, how to allocate resources and how to reach residents during floods and hurricanes. Clearer, explainable AI recommendations and personalized messages could encourage more people to seek shelter. This may reduce harm and help communities become more resilient to severe weather.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Brittany Brand
Boise State University, ID·Social Science
Identify the factors that influence whether households in four wildland-urban interface communities adopt mitigation, stay-and-defend preparedness, or evacuation readiness, extending the Protective Action Decision Model.
Elias Bareinboim
Columbia University, NY·Computer Science
Develop causality-integrated methods for online and offline policy learning, interventional planning, imitation learning and knowledge transfer so autonomous agents decide robustly with less data, building on structural causal models.
Tristan Lin
University of North Texas, TX·Social Science
Track how coastal households in 100-year flood zones turn intentions into flood protection behaviors, surveying them six times over three years based on Paton's Social-Cognitive Preparation Model.
The school
George Mason University: 15 labs on PI Pages, $282M a year on research, 297 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.