Simulation of heuristics for agent behaviour in complex games with applications for ideation

Ruwe, Rüdiger; Kleer, Robin

doi:999910296649

Simulation of heuristics for agent behaviour in complex games with applications for ideation

Ruwe, Rüdiger (Author)

2013

ImpressumAachen : Publikationsserver der RWTH Aachen University 2013

Aachen, Techn. Hochsch., Diplomarbeit, 2013

Veröffentlicht auf dem Publikationsserver der RWTH Aachen University

Hauptberichter/Gutachter
Kleer, Robin (Thesis advisor)

Online
URN: urn:nbn:de:hbz:82-opus-45397
DOI: 10.18154/RWTH-CONV-143451
URL: https://publications.rwth-aachen.de/record/206219/files/4539.pdf

Einrichtungen

Lehrstuhl für Betriebswirtschaftslehre, insbesondere Technologie- und Innovationsmanagement (812710)

Inhaltliche Beschreibung (Schlagwörter)
Spieltheorie (Genormte SW) ; Idee (Genormte SW) ; Wettbewerb (Genormte SW) ; Heuristik (Genormte SW) ; Algorithmus (Genormte SW) ; Agent <Informatik> (Genormte SW) ; Simulation (Genormte SW) ; Mathematik (frei) ; Agenten (frei) ; gametheory (frei) ; idea contest (frei) ; algorithm (frei) ; simulation (frei) ; agent (frei)

Thematische Einordnung (Klassifikation)
DDC: 510
msc: 91A25

Kurzfassung
This thesis deals with the agent side of ideation (idea creation) games in a principal-agent setup. Ideation Games with four incentive schemes are modelled, and heuristics that model a behaviour that is favourable for the agent are found by means of simulation. A simulation environment for this and similar problems of applied algorithmic game theory, is designed, implemented and published at sourceforge. The heuristics are derived by a detailed analysis of the used game model and verified by simulation. For the flat incentive schemes and the performance collaborative scheme as specified in Chapter 2 free-riding is found to be a predominant effect and it can be shown that it is, indeed, in the interest of the agents to spend few resources on ideation. In the private performance incentive scheme as specified in Chapter 2, heuristics could be found that produce very favourable results for both principal and agent. The simulator that was developed to reach the objectives of this thesis can be quickly modified for further research.

OpenAccess:
PDF