Genetic algorithm for the design of rectilinear branched distribution systems

A genetic algorithm for the design of layout of rectilinear branched distribution networks is presented. The genetic algorithm uses a solution coding scheme capable of capturing all potential optimal solutions. Because the traditional genetic algorithm operators of crossover and mutation produce infeasible solutions at an unacceptably high rate with the coding scheme, two new operators, recombination and perturbation, are implemented using an architecture specifically designed to exploit their unique characteristics. The method is demonstrated in 10 separate trials on a small problem involving 6 noeds.