Automated design of compound lenses with discrete-continuous optimization

2025-09-27

AI Summary

This work addresses the longstanding challenge of jointly optimizing continuous parameters (e.g., surface curvatures) and discrete parameters (e.g., lens count, material types) in compound lens design. We propose the first hybrid framework integrating gradient-based optimization with a customized Markov Chain Monte Carlo (MCMC) sampler. Our method enables autonomous evolution of optical topology—including layer count and material combinations—via cross-dimensional mutation and paraxial projection, eliminating reliance on manual intervention. While preserving efficient gradient updates for continuous variables, it simultaneously supports global exploration over discrete configuration spaces. Experiments across diverse imaging tasks demonstrate that our approach significantly advances the Pareto frontier in sharpness and throughput, surpassing the existing speed–sharpness trade-off barrier in automated optical design.

Technology Category

• Search and Optimization: Mixed Discrete/Continuous Search
• Machine Learning: Optimization
• Constraint Satisfaction and Optimization: Mixed Discrete/Continuous Optimization

Application Category

• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

We introduce a method that automatically and jointly updates both continuous and discrete parameters of a compound lens design, to improve its performance in terms of sharpness, speed, or both. Previous methods for compound lens design use gradient-based optimization to update continuous parameters (e.g., curvature of individual lens elements) of a given lens topology, requiring extensive expert intervention to realize topology changes. By contrast, our method can additionally optimize discrete parameters such as number and type (e.g., singlet or doublet) of lens elements. Our method achieves this capability by combining gradient-based optimization with a tailored Markov chain Monte Carlo sampling algorithm, using transdimensional mutation and paraxial projection operations for efficient global exploration. We show experimentally on a variety of lens design tasks that our method effectively explores an expanded design space of compound lenses, producing better designs than previous methods and pushing the envelope of speed-sharpness tradeoffs achievable by automated lens design.

Problem

• Automates joint optimization of continuous and discrete lens parameters
• Enables automatic topology changes without expert intervention
• Expands compound lens design space for improved performance tradeoffs

Innovation

• Jointly optimizes continuous and discrete lens parameters
• Combines gradient optimization with Markov chain Monte Carlo
• Uses transdimensional mutations for global design exploration