Representation and Evaluation Mechanisms for Evolutionary Design of Analog Circuits

Abstract

This work exposes the proposal of two essential elements when the automat-ed analog circuit design problem is approached by means of evolutionary computation: a representation and a fitness function mechanism. The analog circuit design problem is discussed, revealing in one hand, the need of a rep-resentation mechanism that exhibits the flexibility to adapt to very different topologies, remaining independent of the circuit´s complexity. And on the other hand, a fitness function that correctly leads the metaheuristics towards the building of a circuit with a specific behavior, while easily adapts to the representation mechanism. Subsequently a genetic algorithm is used to verify the efficiency of both mechanisms in combination with other routines in the design of one low-pass filter.