A Comparison of Encoding Methods for Automated Design of Optical Architectures

Commercial ray-tracing programs can optimize system parameters for a given optical architecture (e.g., radii, thickness, and spacing of lenses); however, design of the underlying system architecture (e.g., number of lenses, their type, and their order in the optical train) remains an expensive trial-and-error approach driven by prior experience and human intuition. Work on automating the architecture design process has had some success but the problem remains open. We compare the efficacy of novel methods for encoding the design of optical systems on a simple imaging objective.