Optical System Designing System with Reinforcement Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Optical system design processes are inefficient due to reliance on manual parameter adjustments and limited optimization methods, which result in a high number of design plans that may not effectively meet target specifications for optical performance and cost considerations.
Innovation Solution
An optical system designing system utilizing reinforcement learning with a learned model that adjusts parameters such as the number of lenses, lens materials, cementing, and stop locations, and performs aberration optimization using Bayesian optimization to compute design solutions that meet target values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If manual parameter adjustments and limited optimization methods are used, then optical designers can evaluate designs from various viewpoints, but the design process becomes inefficient and requires creating a high number of design plans
Solution Approach 1:
The patent replaces manual mechanical optimization processes with reinforcement learning algorithms. The optical design optimization that previously required iterative manual parameter adjustments is now performed by an automated RL agent that learns optimal design configurations through trial and error, substituting human mechanical operations with intelligent automated systems.
Solution Approach 2:
The reinforcement learning agent performs self-learning and self-optimization without requiring continuous human intervention. The system automatically evaluates design plans, learns from outcomes, and improves its optimization capability over time, enabling the design process to serve itself rather than relying on external expert guidance for each iteration.
2Reliability
If optical designers create many design plans using optimization functions, then promising design plans can be identified, but the number of design plans increases significantly
Solution Approach 1:
The reinforcement learning approach fundamentally changes the optimization parameters from traditional manual adjustment variables to learned policy parameters. The RL agent learns optimal parameter configurations directly from design specifications, transforming the optimization process from exploring many possible designs to directly computing near-optimal designs through learned parameter transformations.
Solution Approach 2:
The reinforcement learning model acts as an intermediary between design specifications and optimal design configurations. Instead of directly manipulating optical parameters or evaluating numerous design plans, the RL agent serves as a intelligent mediator that translates requirements into optimized designs, reducing the need to evaluate large numbers of intermediate designs.
3Manufacturing precision
If traditional optimization methods are used, then optical performance can be improved, but the process relies on designer experience and knowledge
Solution Approach 1:
The patent substitutes human expert knowledge and experience with reinforcement learning algorithms. The implicit knowledge that optical designers acquire through years of practice is captured and encoded in the RL agent's learned policies, replacing the need for human expertise with an automated system that has learned optimal design strategies through training.
Solution Approach 2:
The optimization process transitions from experience-based parameter selection to data-driven parameter optimization. The RL agent learns optimal parameter configurations through training on design data, transforming the optimization from a knowledge-intensive process to an algorithm-driven process that systematically explores the design space based on learned patterns rather than human intuition.
Data Source
AI summary
An optical system designing system for designing an optical system through reinforcement learning has a storage unit storing at least information relating to a learned model, a processor and an input unit that inputs optical design information and a target value to the processor. The learned model is a learning model configured as a function whose parameters have been updated in such a way as to compute a design solution towards the optical design information of the optical system that is based on the target value. The processor executes a macro process of at least one of the actions of changing the number of lenses included in the optical design information, changing a lens material, changing cementing of lenses, changing the location of a stop, and selecting a spherical lens or an aspherical lens and performs an optical system optimization process using weights for aberrations computed by Bayesian optimization as correction values. The processor computes the design information after the execution of the macro process and a reward value based on the target value. Then, the processor computes an evaluation value based on the optical design information and the reward value and computes a design solution based on the target value.