Agent-Based Models use individual computer objects as players or agents. Governed by a set of rules, these agents are turned loose in a computer-generated landscape to perform their appointed task such as trading, segregating, spreading disease or minority opinions, reacting, creating mayhem, bank fraud, or any number of other mischievous endeavors. Often, their resultant behavior has a remarkable similarity to observed reality and can also lead to an understanding of emergent behavior. Consequentially, Agent-Based Models (ABMs), combined with remote sensing imagery and geospatial information, will provide a powerful tool for the analysis and ultimate prediction of complex anthropogenic or natural disasters or events.
Agent-Based Modeling of Emergency Response
Any large-scale anthropogenic or natural disaster, such as a chemical spill, terrorist attack, fire, hurricane, or flooding, etc., impacts human behavior and vehicle movement in the affected area. The response of the affected population is driven by available information about the event. However, inattentiveness to public announcements via vehicle radios, listening to other audio media, and an initial lack of reliable information in the chaotic moments immediately after a disaster will produce an uninformed or misinformed public. For example, the sudden and unannounced nature of a disaster often results in uncertainty with regard to geographic location and extent of the event, resulting in inaccurate information worsened by inattention to public communication. Therefore, the uncertainties and lack of attention to the initial public announcements exacerbates the initial emergency response effort. The question of how communication networks might enhance or diminish the proliferation of information that would facilitate the evacuation of the population is being addressed. To answer these questions, we are developing a model of public and interpersonal communication via cell phones and their respective networks to begin a study of the role and impact of information as it passed rapidly through communication channels as individuals share in the context of initial repetitive public information during an evolving disaster response.
The purpose of this project is to develop methods for rapid analysis of video and sequential images of crowds of people to determine when a peaceful crowd may become panicked and potentially dangerous. The ultimate goal is a system that could quickly analyze the movement of a crowd to predict whether it is about to become chaotic and to identify potential anomalies, “hot-spots” or disruptive individuals within the crowd.
Ultimately, the final component test will be a practical implementation of the system in an urban environment, suitable for rapid deployment by law enforcement personnel. Utilizing wireless technology, unmanned aircraft and fixed observation nodes can be rapidly (and cheaply) set up to provide meshed data for analysis. This final step will be used to determine whether our algorithm is robust in a multi-path environment (wireless, cellular), with inherent time delays and visual resolution. If the system proves successful, it should be useful in monitoring events in real-time for the purpose of disaster response, law enforcement, and peacekeeping, as well as for analyzing and simulating crowd movement offline for the purpose of better disaster preparedness and urban planning.
Below is a video of the working model.