Lithography Learning Control With Uncertainty Compensation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional learning control methods for ultra-precision lithographic apparatuses are dependent on accurate mathematical models, which are difficult to obtain for complex systems, and are sensitive to non-repetitive disturbances, limiting their effectiveness.
Innovation Solution
A learning control system and method that incorporates an extended state observer to approximate the control object as a two-order system with unit mass, compensating for model uncertainty and suppressing external random disturbances, using a learning control system with a movement trajectory generation unit, learning control unit, feedback control unit, and uncertainty compensation unit.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional iterative learning control method is used, then convergence performance is improved with accurate mathematical model, but modeling cost increases and device complexity increases due to flexible modality and complex dynamics
Solution Approach 1:
The patent transforms the complex lithographic apparatus into a simplified two-order system with unit mass through parameter transformation. By changing the representation parameters of the system model rather than modeling the actual complex dynamics directly, the control design becomes tractable while maintaining accuracy. The extended state observer compensates for the transformed uncertainties, achieving high control accuracy without complex modeling.
Solution Approach 2:
The patent uses a nominal mass module as a simplified, computationally inexpensive model instead of requiring an accurate complex model. This nominal model serves as a disposable approximation that is easily updated and computationally trivial, replacing the need for expensive and complex accurate modeling while still enabling effective control through the extended state observer compensation.
2Measurement precision
If conventional iterative learning control method is used, then learning performance is improved, but sensitivity to non-repetitive disturbance increases
Solution Approach 1:
The extended state observer acts as an intermediary between the control input and the actual system dynamics. It estimates the uncertainties and disturbances including non-repetitive components, and the uncertainty compensation unit uses these estimates to adjust the control signal. This intermediary mechanism filters out the harmful effects of non-repetitive disturbances while preserving the learning performance.
Solution Approach 2:
The patent implements a feedback mechanism through the extended state observer that continuously monitors the system behavior and estimates uncertainties. The estimated uncertainties are fed back through the uncertainty compensation unit to adjust the control input in real-time, creating a closed-loop system that actively compensates for non-repetitive disturbances and improves robustness.
3Speed
If accurate mathematical model is obtained, then convergence rate is improved, but modeling cost increases
Solution Approach 1:
The patent achieves fast convergence without expensive accurate modeling by transforming the system into a two-order system with unit mass. This parameter transformation simplifies the dynamics while the extended state observer compensates for the transformed uncertainties, achieving both fast convergence and low modeling cost simultaneously.
Solution Approach 2:
The patent replaces expensive accurate models with a cheap nominal mass model. This simplified model has trivial computational cost and can be easily updated, yet through the extended state observer and uncertainty compensation, it achieves convergence performance comparable to or better than expensive accurate models.
Data Source
AI summary
A learning control system includes a movement trajectory generation unit, a learning control unit, a feedback control unit, and an uncertainty compensation unit. The movement trajectory generation unit includes a movement trajectory generator; the movement trajectory generator is configured to generate a reference movement trajectory; a position measured signal is subtracted from the reference movement trajectory to obtain a position error signal which is input to the learning control unit; the learning control unit is configured to generate a feed-forward signal which is added with the position error signal to obtain a corrected error signal; and the corrected error signal is input to the feedback control unit; the feedback control unit includes a feedback controller configured to generate a feedback control quantity which is input to the uncertainty compensation unit; and the uncertainty compensation unit is configured to generate the position measured signal.