Evolutionary algorithms for design

Genetic algorithms are effective optimization tools which work well using bit string encoding, but suffer from intensive computations and limited accuracy. To avoid the added problems of converting between real numbers and bit strings, the authors propose a new evolutionary algorithm which can work in real number space while preserving the structure of genetic algorithms as much as possible. This modified algorithm, which can be viewed as an extended genetic algorithm working in real number space, eliminates the problems of quantization and conversions between real and binary representations, and in addition offers a true global search in real number space. The new algorithm is described and illustrated by applying it to the design of an acoustical array to achieve a reception pattern with uniformly low sidelobes.