Simulation and Distributed Architecture of Multi Agent-Based Behavioral Economic Landscape (MABEL) Model within SWARM

Abstract

The Multi-Agent−based Behavioral Economic Landscape (MABEL) model introduces a distributed modeling architecture framework that supports the simulation of land-use changes over time and space over large regions. The model is based on the Swarm modeling software package, which is supported by a unique client-server framework with multiple interfaces built around a geographic information system (GIS), statistical analysis and database (SPSS), and Bayesian network software. The model architecture supports an integrated simulation environment with remote data retrieval, distributed and parallel scenario simulations, centralized decision-making algorithms, graphic displays for both client and server model components, and analysis capabilities. On the client side of MABEL, computational agents represent Bayesian relations among geographic, environmental, human, and socio-economic variables, with respect to land-use changes occurring across landscapes. A multi-agent simulation environment is created within Swarm, which simulates the buying, selling, and keeping of land by different types of agents. Agents are allowed to participate in an abstract market model. The characteristics of the server side of MABEL include (1) remote data retrieval via multiple interfaces with GIS software (ArcGIS and Arcview) and statistical database software (SPSS), and (2) coordinated agent decision making that allows for decision requests of agents from clients to be made to centralized Bayesian network agent profiles located on the MABEL server.