Using FLAME Toolkit for Agent-Based Simulation: Case Study Sugarscape Model

TL;DR

This paper demonstrates how the FLAME toolkit can be used to simulate and analyze the Sugarscape agent-based model, revealing the impact of initial conditions and computational performance on social simulation outcomes.

Contribution

It extends the Sugarscape model by implementing it in FLAME, providing insights into initial state effects and processing variance on multicore architectures.

Findings

Initial starting states significantly influence simulation results.

Simulation time varies notably across different initial conditions.

Limitations of FLAME affect model accuracy and scalability.

Abstract

Social scientists have used agent-based models to understand how individuals interact and behave in various political, ecological and economic scenarios. Agent-based models are ideal for understanding such models involving interacting individuals producing emergent phenomenon. Sugarscape is one of the most famous examples of a social agent-based model which has been used to show how societies grow in the real world. This paper builds on the Sugarscape model, using the Flexible Large scale Agent-based modelling Environment (FLAME) to simulate three different scenarios of the experiment, which are based on the Sugar and Citizen locations. FLAME is an agent-based modelling framework which has previously been used to model biological and economic models. The paper includes details on how the model was written and the various parameters set for the simulation. The results of the model simulated are processed for three scenarios and analysed to see what affect the initial starting states of the agents had on the overall result obtained through the model and the variance in simulation time of processing the model on multicore architectures. The experiments highlight that there are limitations of the FLAME framework and writing simulation models in general which are highly dependent on initial starting states of a model, also raising further potential work which can be built into the Sugarscape model to study other interesting phenomenon in social and economic laws.