AI 中文总结
本研究将詹姆斯·韦伯空间望远镜的工程遥测用作机会波前传感器，采用两阶段梯度提升回归器，利用在轨数据实现了18个镜段中13个的高统计显著性光程差推断，为大型天文台提供了低开销的波前估计方案。
AI 中文摘要
空间望远镜通过周期性、资源密集型的波前校准维持光学对准，导致两次校准之间的光学状态处于未观测状态。我们提出将星载工程遥测作为一种机会波前传感器。具体而言，詹姆斯·韦伯空间望远镜已记录的高时间分辨率热信号和指向信号，无需专门测量即可足以纳米级精度恢复空间分辨的光程差。我们的框架采用两阶段梯度提升回归器，在低维主成分基上预测每个镜段的光程差残差。该模型使用六个月在轨捕获的数据进行训练和评估。基于此有限数据，18个镜段中有13个的光程差推断具有高统计显著性；在6个镜段中，模型实现了50%以上的解释方差。这些结果证明了遥测驱动波前估计作为低开销传感方式的可行性，该研究表明其有望为大型分块式天文台的连续监测和校准调度提供支持。
英文摘要
Space telescopes maintain optical alignment through periodic, resource-intensive wavefront calibration, leaving the optical state unobserved between corrections. We propose treating onboard engineering telemetry as an opportunistic wavefront sensor. Specifically, the high-cadence thermal and pointing signals already recorded by the James Webb Space Telescope may suffice to recover spatially resolved optical path difference at nanometer precision, without dedicated measurements. Our framework uses a two-stage gradient boosting regressor. It predicts optical path difference residuals for each mirror segment in a low-dimensional principal component basis. The model is trained and evaluated using data captured on-orbit during six months. Based on this limited data, optical path difference inference has high statistical significance in 13 out of 18 mirror segments. In 6 mirror segments, the model achieves explained variance above 50\%. These results demonstrate feasibility for telemetry-driven wavefront estimation, as a low-overhead sensing modality. The study suggests a potential to support continuous monitoring and calibration scheduling for large segmented observatories.
Comments10 pages, 10 figures. Accepted to ICCP 2026
