Abstract: First we introduce the current research status of evolutionary design for analog circuits briefly, and then we propose a multi-objective adaptive Genetic Algorithm and discuss it in detail. It features an encoding-decoding scheme based on typical structures of circuits and standard values of components(i.e. discrete values in common use), a multi-objective fitness evaluation method using Pspice simulation and dynamic adjusting of weight vectors, and a genetic parameter adaptiation technique for considering the different effect of each locus on fitness and tracking the course of evolution. Some experiments on active filters show that with our approach a suitable type and an optimal value of each component in the circuit can be found automatically, so an effective circuit with a response very close to the design target can be developed, and only modest computational resource is required that a PC can afford.

Key words: evolutionary circuit design, genetic algorithm, evolvable hardware, optimization