Show simple item record

Towards Rigorous Agent-Based Modelling

Linking, Extending, and Using Existing Software Platforms

dc.contributor.advisorKurth, Winfried Prof. Dr.
dc.contributor.authorThiele, Jan C.
dc.titleTowards Rigorous Agent-Based Modellingde
dc.title.alternativeLinking, Extending, and Using Existing Software Platformsde
dc.contributor.refereeKurth, Winfried Prof. Dr.
dc.description.abstractengAgent- or individual-based modelling is a modelling approach where the heterogeneity of entities, i.e. agents, matters. In the past ten to twenty years agent-based models became increasingly popular and were applied in many different research areas, from computer science over sociology and economy to ecology. Built by rule systems instead of differential equations they cannot be solved analytically but have to be implemented as computer software and analysed by running simulations. As the field of agent-based modelling is relatively young compared to, for example, mathematical modelling with differential equations, established standards in developing, implementing, describing, and analysing are missing or currently being arising. Instead, many modellers build, implement and analyse their models from scratch and reinvent the wheel. The work at hand aims to support the process of establishment of standards in agent-based modelling. After a short general introduction in the first chapter, the second chapter gives an introduction into the history of agent-based modelling in different research areas, discusses open issues in agent-based modelling, presents the most important toolkits/languages/Integrated Development Environments (IDE) for implementing agent-based models, and closes with a deeper look on the IDE/language NetLogo and some extensions developed here. In the third chapter a framework for building and analysing agent-based models by linking two existing and well-known toolkits/languages, NetLogo and R, is described. Such a seamless integration of an agent-based modelling environment with a statistics software enables the modeller to design simulation experiments, store simulation results, and analyse simulation output in a more systematic way. It can therefore help close the gaps in agent-based modelling regarding standards of description and analysis. The fourth chapter of this theses provides a "cookbook" of many important methods for calibration of agent-based models as well as for sensitivity analysis. Such a comprehensive overview of well-known and established techniques enables the modeller to become aware of existing methods, learn what they can deliver and where to apply them. Furthermore, the recipes contain application examples implemented and adaptable to other models implemented in NetLogo under the use of the framework introduced in the third chapter. A key feature of science - the replication of experiments - is discussed in chapter five with focus on the field of agent-based modelling in ecology. It should encourage the community to replicate models and publish the replications. Replication of models fulfils different purposes: it uncovers implementation-dependent differences in model results, it shows lacks in documentation and/or documentation protocols as well as robustness tests, and it is a first step towards community-tested standard models or model components. The work closes with an integrated discussion and outlook on open
dc.contributor.coRefereeGrimm, Volker Prof. Dr.
dc.contributor.thirdRefereeLytinen, Steven Prof. Dr.
dc.subject.engagent-based modellingde
dc.subject.engindividual-based modellingde
dc.subject.engrigorous abmde
dc.subject.engstandard softwarede
dc.affiliation.instituteFakultät für Mathematik und Informatikde
dc.subject.gokfullInformatik (PPN619939052)de

Files in this item


This item appears in the following Collection(s)

Show simple item record