posted on 2010-10-29, 11:47authored byIan J. Graham
This thesis describes research into the development of a Computer Aided Design
(CAD) tool that uses a Genetic Algorithm (GA) to generate and evolve original
design concepts through human interaction.
CAD technologies are firmly established in the later stages of design, and include
many applications of Evolutionary Algorithms (EAs). The use of EAs as generative and search tools for conceptual design is less evident in fields other
than abstract art, architecture and styling. This research gains its originality in
aiming to assist designers early in the design process, by creating and evolving
aesthetically interesting forms (objects).
The integration of GA software with a solid modelling system has enabled the
development of a prototype `Evolutionary Form Design' (EFD) system. Objects
are defined using a genetic data structure and constructed from various geometric
primitives and combinations of Boolean operators. The primitives interact in
ways that are not easily predicted, often creating novel shapes that are unlikely to
have been discovered through conventional means. Edge blending further adds to
objects' complexity and visual appeal. Populations of objects are subjected to a
`selective breeding' programme, directed through the user's allocation of scores,
and may also be guided by simple geometric targets. These factors determine
which objects are `fittest' and most likely to parent a new, hopefully improved
generation of objects. The challenge has been to turn the concept into a genuinely
useful tool, ensuring that desirable features are reproduced in subsequent
populations. The key to achieving this is the way objects are recombined during
reproduction. Work has included developing 4 novel routine for grouping the
individual primitives that form objects using a Teamforming algorithm.
Innovative, aesthetically interesting forms can be evolved intuitively and
efficiently, providing inspiration and the initial models for original design
concepts. Examples are given where the system'is used by undergraduates to
generate seating designs, and by the author, to create virtual sculptures and a
range of consumer product concepts.
History
School
Mechanical, Electrical and Manufacturing Engineering