InterJournal Complex Systems, 1745
Status: Accepted
Manuscript Number: [1745]
Submission Date: 2006
Complexities, Catastrophes and Cities: Unraveling Emergency Dynamics
Author(s): Giuseppe Narzisi ,Venkatesh Mysore ,Lewis Nelson ,Dianne Rekow

Subject(s): CX.44



A large city or a city-state is an example of an evolving and self-organizing complex environment that efficiently adapts to different and numerous incremental changes to its social, cultural and technological infrastructure. Yet, it is not immediately apparent whether the current urban architectures are also robust against sudden, rare and punctuated catastrophic events. From a practical viewpoint, one may wonder if there is a methodical and algorithmic approach to plan against such events, in order that a large urban structure can recover from the effects of a disastrous event quickly and efficiently. In principle, by combining powerful ideas from game theory, simulation and multi-objective optimization, a suitably powerful approach could be devised. A realistic response simulation of an urban environment requires a large number of actors, each with their own skills, objectives, behaviors and resources, to be able to interact and coordinate their efforts in order to manage the outcome of a disaster. In this paper, we demonstrate how Agent-Based Modeling (ABM) can serve as a powerful simulation technique for analyzing such complex scenarios. We describe a new multi-agent based simulation framework, built atop the Java version of RePast 3.1, able to model and simulate catastrophic scenarios characterized by one-time exposure (e.g., chemical agent, bomb explosion, food poisoning), with additional analysis performed using the trace-analyzing linear temporal logic model-checker Simpathica / XSSYS. One of the main issues in ABM, which remains an art more then a science, is building models at the appropriate level of granularity necessary to capture the dynamics of interest without making analysis infeasible. Our efforts in this direction have resulted in a system with the following features: (1) large number of agents, belonging to five different classes: Person, Hospital, On-Site Responder, Ambulance and Catastrophe; (2) large number of parameters for describing the agentsí behavior and interaction; (3) several communication channels for information (health / resource levels, hospital operation mode, etc.) exchange between similar and differing agents; (4) modeling of the Person agent as selfish and bounded rational, with stochastic personality traits emulating panic behavior; (5) realistic models of medical / responder units and Catastrophe agent effects (disease prognosis and dosage response), validated by experts from the NYU Center for Catastrophe Preparedness and Response; (6) incorporation of topological and transportation constraints, via real urban geographic data integration (specifically, of Manhattan, New York); and finally, (7) computer software for parallel and distributed concurrent computing on clusters of workstations. Simple rules of behavior are seen to produce uncanny emergent dynamics with unpredictable interdependence. The complex interactions between the affected population and the available resources of a response plan have remained poorly understood, are still beyond the analytical capability of traditional modeling tools, and have resisted any systematic investigation until now. The Agent-Based Modeling paradigm, in conjunction with statistical analysis, multi-objective optimization and model-checking of agent-traces, offers a novel way to understand, plan and control the unwieldy dynamics of a large-scale urban emergency response. Thus, the ABM approach, combined with traditional tabletop exercises, holds immense promise as an aid to refine public health policies governing both catastrophe preparedness and response.

Retrieve Manuscript
Submit referee report/comment

Public Comments: