Adaptable and interactive 3D modelling using genetic algorithms

Abstract

This thesis discusses the process of designing, implementing and testing a creative AI system that generates 3D models for the purpose of video game development. The system utilizes a combination of an interactive genetic algorithm and a distance weighted K-nearest neighbour regressor. Regression is employed as a learning mechanism to enable the system to rate designs based on the overall user design preferences absorbed through user interaction. The results from testing the implemented sytem are discussed, and possible changes and improvements to the system based on the produced results are also covered.