Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.654459
Title: Generating anthropomorphic NPC controllers with genetic profiling
Author: Davies, M. I.
Awarding Body: University of Essex
Current Institution: University of Essex
Date of Award: 2013
Availability of Full Text:
Access from EThOS:
Abstract:
This thesis highlights the research and findings of a PhD project designed to investigate whether mimicry!, Genetic Programming and computer game technologies could be used to create controllers for artificial humans that encourage anthropomorphic behaviour. From an initial broad research focus, the project eventually concentrated on the anthropomorphic attributes of behaviours demonstrated by Non-Player Characters (NPCs) used by computer games. In a deliberate attempt to uncover a viable alternative to existing mechanisms, the project sought to apply Artificial Intelligence systems more commonly used in academic Computer Science research to a computer game context. As part of this research, a new Artificial Intelligence mechanism was developed, (UK Patent Application No. 1012243.0), which combined elements of Genetic Programming and mimicry. The new mechanism was subsequently dubbed 'Genetic Profiling' and forms the core research presented herein. To further investigate the Genetic Profiling mechanism a number of environments were required either to act as evaluation test-beds or as an integral part of the new Artificial Intelligence strategy by supplying raw behavioural data from real people. The iWorld concept outlined in this document categorises these augmented environments, which could exist as physical, virtual or mixed-reality entities. Several iWorlds were created for use by this project, including the deployment of a Massive Multiuser Online (MMO) environment. Each was used to evaluate different implementations of the Genetic Profiling mechanism, including several possible strategies for generating and representing artificial behavioural data. Most of these iWorlds were customised to evaluate the performance of generated NPC controllers with special regard to a specific aspect of computer game design. In addition to extensive quantitative testing, the anthropomorphic attributes of artificial Genetic Profiling NPC controllers were evaluated qualitatively, using a bespoke testing strategy inspired by the 'Turing Test' experiments. Furthermore, to provide additional grounding and justification, models and theories from Psychology concerning human personalities were adapted to allow better representation and analysis of generated behaviours. Inspired by this research, the project proposes a hierarchical architecture describing the structure and contents of artificial human personalities generated using the Genetic Profiling mechanism. By grounding the artificial behaviours produced by Genetic Profiling in this way it was possible to identify sets of rudimentary Traits in artificial personalities, based upon the behavioural content generated for NPC controllers. At the time of writing, in addition to a patent application the research performed by this project and results from some of its investigations have provided content for a number of peer-reviewed international conference and journal papers. Essentially, the research presented in this document details the first-steps of a new paradigm for potentially allowing NPCs present in computer games to behave more like human players. As such, based upon this initial ground-world, there are several possible directions and augmentations that could potentially be made to the Genetic Profiling mechanism, should this project be continued in the future.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.654459  DOI: Not available
Share: