How to Use Machine Learning to Automate Aligning Optical Components in Experiments

Project Description

The goal of this project is to develop a machine learning-based approach for automated laser beam alignment using mirrors. By leveraging optical simulations, we aim to train algorithms that can optimize mirror adjustments to achieve precise beam positioning. This project will explore various ML techniques, such as reinforcement learning and gradient-based optimization techniques, to determine effective alignment strategies. The ultimate objective is to compare machine-learned approaches with traditional human methods and assess their scalability for more complex optical alignment tasks.

 Mentors

Tianyu Wang, PI Weiru Fan

1. Develop a machine learning-based approach for automated laser beam alignment
2. Explore different ML techniques (e.g., reinforcement learning, Bayesian optimization, Gradient-based methods) for optimizing optical path alignments
3. Compare ML-derived alignment strategies with traditional human methods
4. Evaluate the scalability of ML-based alignment for more complex optical setups
1. Apply machine learning to automate laser beam alignment and compare results with human strategies
2. Frame optics problems as ML optimization tasks, defining state, action, and reward functions
3. Explore ML techniques like reinforcement learning, genetic algorithms, and Bayesian optimization
4. Understand laser beam alignment principles and mirror adjustments for beam propagation
5. Gain hands-on experience with optical simulation tools like pyOpTools and Phydemo Ray Optics Simulator
7. Generate and analyze synthetic data for ML model training and performance evaluation
8. Develop problem-solving skills to determine effective alignment strategies
9. Compare simulated alignment approaches with traditional human-designed methods
10. Learn to scale ML-based solutions from simulations to real-world optical setups
11. Interpret alignment results, quantify accuracy, and refine models based on data analysis

Timeline

Week 1: Get familiar with the lab environment; safety training.
Week 2: Read literature and summarize prior methods; set the goal for learning relevant ML methods.
Weeks 3-4: Try to reproduce results from prior publications.
Weeks 5-9: Try to propose improvements over prior results and implement the new schemes in the lab.
Week 10: Summarize the results.