A novel line of research in homeland security and computational modeling has demonstrated that human behavior is influenced by architectural design and suggests that even modest changes in a building’s design or redesign can have a major influence on how a crowd will behave during an emergency. The objective of this project is to develop algorithms and technologies needed for real-time crowd behavior forecasting as it relates to building design.
A critical component of a Virtual Sentry is the ability to autonomously monitor crowded places and make real-time interventions. Examples include crowd management, changes in security protocols, and adapting physical infrastructure. Inherent in the Virtual Sentry specifications is the ability to forecast crowd behavior accurately and efficiently in response to a myriad of possible interventions, extract meaningful insights, and balance competing objectives when aiding patrons and security directors in making real-time decisions where hundreds of lives may be at stake. The proposed research deliverables will serve as the cornerstone of an AI/ML-driven automated decision system for the proposed Virtual Sentry architecture that will be used as the basis of effective threat deterrence, risk assessment, and site architecture design.