Emergent behaviour in virtual agents
An investigation into causes of emergent behaviour    Rhodes University, P.O. Box 94, Grahamstown, Republic of South Africa

 

Home

Abstract

Progress

Literature

Experimentation

Write up

Publications

Conferences

SKills gained

Milestones

About authors

 

Write up

The first section

This section include the title, abstract, keywords, declarations,  acknowledgements, table of contents, table of figures and the table of tables. The contents of this section, available in PDF format only,  are not final and may be changed any time.

Introduction

Our Chapter I of the thesis is presented in this section, containing the Introduction, problem statement,  research methodology, thesis contributions, and the thesis overview.

Background

Our Chapter 2, which summarises related work is presented in this section.

The path formation model

In this section, we discuss path formation as an example of emergent behaviour. We discuss an ants based path formation, flock inspired path formation and path formation using the generalised version. We conclude by extracting those sections of the models that cause path formation in each case. The extracted sections are tested for similarity and results recorded. A generalised version of the path formation model, inspired by both ACO and boids is presented as conclusions. This section forms Chapter 3.

Flocking, schooling and herds

In Chapter 4, we discuss flocks, schools and herds as another example of emergent behaviour. An ants based flocking/schooling model is presented. Similarly, a Reynolds based flocking model is also presented. We in the same way, extract those sections of the two models that cause flocking. We compare the results and a generalised version is deduced.

Application: generalisation vs

With the successful generalisation of routines causing specific behaviour in Chapter 3 and 4, an ants inspired area coverage model  is presented. Additionally, a flock inspired area coverage model is also presented. We demonstrate that the generalised version of the two models achieve area coverage too. This forms Chapter 5 of the thesis.

Other formations and tests

If them generalised version produce the same emergent behaviour, we should be able to adapt it to achieve other wanted behaviour.  In this section, we vary routine parameters and achieve other formations such as circles, triangle, rectangle and hopefully user defined formations, indicating how we can predict emergence from knowing its causes. This culminates Chapter 6.

Results analysis & discussion

 
Conclusions