Authors: Dandong Yin*, Department of Geography and GIS, University of Illinois at Urbana-Champaign, Yan Liu, Department of Geography and GIS, University of Illinois at Urbana-Champaign, Shaowen Wang, Department of Geography and GIS, University of Illinois at Urbana-Champaign
Topics: Cyberinfrastructure, Geographic Information Science and Systems, Transportation Geography
Keywords: Agent-based modeling, traffic simulation, cyberGIS, distributed computing
Session Type: Paper
Start / End Time: 8:00 AM / 9:40 AM
Room: Grand Ballroom A, Astor, 2nd Floor
Presentation File: No File Uploaded
In many emergency scenarios, mass evacuation is necessary to cope with severe public threats within certain spatio-temporal ranges, such as hurricanes, wildfires and terrorist attacks. To study possible consequences and better understand complex phenomena like mass evacuation, agent-based models (ABMs) have been widely adopted by previous work. Many existing models simulate the behavior of individuals, hence posing computational challenges when applied to large urban areas. A key strategy to resolve the computational challenges is to partition transportation networks into smaller regions and handle corresponding computational loads by taking advantage of advanced cyberinfrastructure. In this study, a novel network partition algorithm is developed to improve the scalability of agent-based models for mass emergency evacuation by establishing a cyberGIS-enabled computational framework for exploiting unique spatial movement patterns of emergency evacuation. Experiments show that our algorithm outperforms a popular network partition algorithm for microscopic traffic simulation in terms of load-balancing and communication reduction. Furthermore, the entire computational environment and software codes of this framework are built into a Docker image with a series of Jupyter notebooks, which enables the reproducibility, validation and future extension of the framework by broad cyberGIS and spatial modeling communities.