SBIR Phase I: Machine Learning Enhanced Hardware for Optimization and Self-tuning of High-Finesse Optical Cavities

这个小企业创新研究（SBIR）第一阶段项目的广泛影响将是展示强大的计算技术的实用性，如机器学习，用于复杂光学仪器的自动校准。这对于部署在现场的仪器尤其重要，因为在现场，环境波动不断以不可预测的方式干扰组件。这些技术的成功实施将使复杂的激光器和仪器从实验室过渡到商业用途。这在不断发展的量子通信行业尤其有影响力。该项目将推动机器学习的发展，以自主校准这些敏感仪器。这个小企业创新研究第一阶段项目将研究使用机器学习来保持由光学腔制成的精美滤光片的对准。这些空腔是复杂光学仪器中的常见部件，迄今为止，还没有商业解决方案可以自动对准这些空腔。这是由于控制对齐的参数集的复杂性，以及训练系统选择最优路径以恢复对齐的挑战。该项目的主要交付成果将是一套基于模块化机器学习的优化算法和一个现成的商用30mhz带宽自调谐光滤波器硬件模块。由于提出的自适应控制策略不依赖于单一的底层模型，因此开发的框架将是模块化的，并且易于转移到其他依赖于具有不同参数和控制特征的光子排列的非线性系统中。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The broader impact of this Small Business Innovation Research (SBIR) Phase I project will be to demonstrate the utility of powerful computational techniques, like machine learning, for auto-alignment of sophisticated optical instruments. This is particularly important for instruments deployed in the field, where environmental fluctuations continuously perturb components in unpredictable ways. Successful implementation of such technologies would allow complex lasers and instruments to transition from the laboratory to commercial use. This could be particularly impactful in the growing quantum communication industry. This project will advance the development of machine learning to align these sensitive instruments autonomously. This Small Business Innovation Research Phase I project will investigate the use of machine learning to maintain alignment of exquisite filters made from optical cavities. To date, no commercial solution exists for auto-alignment of these cavities, which are common components in complex optical instruments. This is due to the complexity of the parameter sets governing alignment, and the challenge to train the system to choose the optimal path to restore alignment. The primary deliverables of this project will be a set of modular machine learning-based optimization algorithms and a commercial off-the-shelf 30MHz-bandwidth self-tuning optical filter hardware module. Since the proposed adaptive control strategy does not rely on a single underlying model, the developed framework will be modular and easily transferable to other nonlinear systems, which depend on photonic arrangements with varying parameters and control features.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.