In this paper we present a co-evolutionary model of design in which potential solutions to a design problem evolve in parallel with the problem description. This computational model is based on the observation that creative designers often refine and revise the design requirements of a particular problem at the same time as they generate and propose an evolving series of potential solutions to the problem. Genetic algorithms guide the search for a solution using a fixed fitness function, and revisions to the crtieria for the best solution involve manually modifying the fitness function. In our model of co-evolutionary design, the fitness function is automatically changed as the problem space and solution space coevolve. In the paper we describe the model in general, show how we have applied it to the design domain of structural engineering, and present some preliminary experimental results.
[1]
Josiah Poon,et al.
Co-evolution in design: A case study of the Sydney Opera House
,
1997
.
[2]
Mary Lou Maher,et al.
Creative Design Using a Genetic Algorithm
,
1994
.
[3]
Mary Lou Maher,et al.
Formalising Design Exploration as Co-Evolution
,
1996
.
[4]
Kenneth A. De Jong,et al.
A Cooperative Coevolutionary Approach to Function Optimization
,
1994,
PPSN.
[5]
Jan Paredis,et al.
Coevolutionary computation
,
1995
.
[6]
Melanie Mitchell,et al.
An introduction to genetic algorithms
,
1996
.
[7]
Mary Lou Maher,et al.
Modeling design exploration as co-evolution
,
1996
.
[8]
Mary Lou Maher,et al.
A Model of Co-evolutionary Design
,
2000,
Engineering with Computers.