24 October 2026 in SSRN Electronic Journal
Conventional zoom optical design typically begins with paraxial analysis to establish an initial configuration, while subsequent refinement and aberration correction rely on the experience of optical designers. This paper proposes an automated framework for zoom optical design that combines particle swarm optimization (PSO) with differentiable optical optimization. The framework first performs a global search in the paraxial domain to obtain an initial configuration that satisfies the target focal-length specifications. Starting from this configuration, differentiable optimization progressively expands the field of view and aperture through a curriculum-learning strategy while optimizing image quality. An adaptive illumination-based fine-tuning strategy is subsequently introduced to automatically adjust the clear apertures subject to prescribed relative-illumination constraints, thereby improving marginal-field image quality. The framework requires the target specifications and group configuration as inputs and automatically generates optimized zoom systems. Two optical systems with different group configurations were successfully designed using the proposed framework. At 0.7 normalized field, the adaptive illumination module improved the mean MTF at the target spatial frequency by 24% and 15% in Cases 1 and 2, respectively. These results demonstrate the effectiveness and potential applicability of the proposed framework for automated zoom optical design.