Multiobjective coevolution of artificial neural network ensembles for game artificial intelligence
The retail sales of computer and video games have grown enormously during the last few years, not just in United States, but also throughout the rest of the world. This is the reason why a lot of game developers and academic researchers have focused on game related technologies, such as graphics, au...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2013
|
| Subjects: | |
| Online Access: | https://eprints.ums.edu.my/id/eprint/11599/1/ph000000064.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The retail sales of computer and video games have grown enormously during the last few years, not just in United States, but also throughout the rest of the world. This is the reason why a lot of game developers and academic researchers have focused on game related technologies, such as graphics, audio, physics and machine intelligence with
intention of creating newer and more attracting games. The main objective of this thesis is the investigation of the potential contributions of advanced yet general computational intelligence techniques to create an autonomous game agent or controller to solve the problems in the game environments with the appropriate level of intelligence. The game agents must be able to react immediately to the dynamic changes of the environments
by selecting appropriate responses according to the situations. Hence, this research focuses on automatic generation of highly skilled intelligent game controllers through a systematic investigation of the application of evolutionary multiobjective optimization, neural network ensembles, coevolutionary approaches, and more importantly through the hybridization of these techniques. The prey and predator game genre of Ms. Pac-man game was used as a test-bed, because this game provides Significant challenges for
testing and assessing the novel machine intelligence techniques. The experimental results indicate that the best-performing game controller could be achieved through the
hybridization of evolutionary multiobjective optimization, neural network ensembles and coevolutionary techniques. |
|---|