A Doublet Lens Design using a Genetic Algorithm

摘要

This paper describes the automated optimization of a simple optical doublet lens using a genetic algorithm. The genetic algorithm, used in conjunction with a simple ray-tracing model, seeks to find optimal lens designs with minimal designer intervention. The characteristics of the genetic algorithm and factors considered when applied to a lens design are discussed. The chromosome representation for a particular lens combination determines the effective "resolution" with which the genetic algorithm can find and differentiate local optima from the global optimum. The importance of matching granularity and gene decoding are discussed together with their effect upon search convergence toward the global optimum. A successful convergence process is determined by the selection and breeding strategies. Several strategies were examined and most found ineffective due to genetic drift within the population. Fitness proportional selection was found most effective, as was double point crossover breeding. With minimal designer input, a simple doublet lens was automatically designed.