Evolutionary optimization in the search for intrinsically robust circuit designs

A hybrid genetic algorithm is developed that allows the variability of components to be included in the design of analogue electronic circuits. During the local search stage of the hybrid genetic algorithm, the Taguchi method is used to determine the sensitivity of the neighbourhood of a proposed trial solution. The information gathered during local learning is used to guide the stochastic search towards intrinsically robust circuit designs. An example is given showing that circuit designs obtained from this method outperform those obtained from a traditional genetic algorithm.