Large-scale agent simulation of agent-based models

Agents are a very useful form of modelling and understanding complex natural systems. Agent-based simulations have been applied in many different domains, such as: economics, sociology, psychology, etc. While they offer a very flexible approach to modeling such systems, they often require significan...

Full description

Saved in:
Bibliographic Details
Main Author: Poh, Wei Li.
Other Authors: School of Computer Engineering
Format: Final Year Project
Language:English
Published: 2012
Subjects:
Online Access:http://hdl.handle.net/10356/48602
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
Description
Summary:Agents are a very useful form of modelling and understanding complex natural systems. Agent-based simulations have been applied in many different domains, such as: economics, sociology, psychology, etc. While they offer a very flexible approach to modeling such systems, they often require significant computational resources for execution. Usually, distributed computing techniques are widely used in the simulation and modeling of agents in order to reduce simulation or computational time. There are mainly 3 factors that affect computational time: node-to-node communication cost, migration cost of agents and the load of each node. Depending on the type of partition methods used, it would affect the nodes communication cost and the node’s load in which may impact to the simulation time. This project will investigate the effectiveness of different partition methods for distributed simulation of agent-based models (ABM) in discrete and continuous environments. The intention is to provide faster execution of the models. Case studies example used for discrete environment is Game of Life model (GoL) and Boids model for continuous environment. The results find that static partitioning of environment horizontally (SPH) provided the fastest execution time for discrete environment. For continuous environment, depending on the behavior of the agents, static partitioning of environment vertically (SPV) or grid-based partitioning may provide the fastest execution time.