To illustrate this form of Evolutionary Design, consider the design of a four-legged table. A typical approach to Evolutionary Design Optimisation would be to parameterise part of the table design - for example, the position and length of the legs - and use an EA to optimise the values of those parameters for some criteria - for example, maximise the stability of the table. As the figure shows, phenotypes could consist of eight parameters, with genotypes being 64 bits.
Phenotype:
Table consisting of fixed top and four legs defined by:
Length of leg 1, Distance of leg 1 from centre
Length of leg 2, Distance of leg 2 from centre
Length of leg 3, Distance of leg 3 from centre
Length of leg 4, Distance of leg 4 from centre
Genotype:
11010110 10101101 10101110 10011010 01101010 10001010 11110010 00101110 |
Length 1 Distance 1 Length 2 Distance 2 Length 3 Distance 3 Length 4 Distance 4
Figure. Evolutionary Optimisation of a table.
In this trivial example, the optimal solution is clearly a table design with all four legs the same length, and with the legs placed at the four corners of the table top, figure 1.20 (right).
As the example shows, Evolutionary Optimisation is capable of finding functionally optimal (or at least functionally good) permutations of the form of existing designs. However, it is incapable of changing the design concept (i.e. a table top resting on a single pedestal with a wide base will never be 'invented').