CyberGIS Toolkit Home Page

Parallel Agent-Based Modeling (PABM)

Agent-based models (ABM) have been widely used as computational approach to simulating complex dynamic phenomena in ecological, geospatial, and social systems. Agents are basic problem-solving entities that can sense and react to stimuli obtained from themselves, other agents, and their interacting environments. The problem-solving capabilities of ABM lie in the explicit representation of agents, environments, and interactions among agents and between agents and environments. However, this type of representation often leads to significant data and computational challenges and, thus, requires high performance computing support.

PABM of the CyberGIS Toolkit is an illustrative application to demonstrate a spatially-explicit approach (Shook, Wang et al. 2013) designed to tackle a major computational bottleneck in employing ABM to model and simulate complex spatial dynamics: runtime inter-process communication. In this approach, runtime inter-process communication is conceptualized into four categories: entity interaction, entity transfer, simulation management, and parallelization. To address the scalability and computational performance challenges within these categories, a communication framework was developed to address agent organization and operation, domain decomposition, load balancing, and efficient data exchange through ghost zones, respectively.

Download

Publications

If you publish your work in which CyberGIS PABM is used, please cite the following publication:

  • Shook, E., Wang, S., and Tang, W. “A Communication-Aware Framework for Parallel Spatially Explicit Agent-Based Models.” International Journal of Geographical Information Science. doi

Install

Please refer to the INSTALL document in source package on how to install PABM. PABM uses IPM to gather performance information. IPM depends on PAPI and MPI.

Examples

Usage: mpirun -np [number of processes] PABM

Example output

NMI Build and Test Status

References