Systematic optical axis prediction based on machine learning for SONG pointing tracking model

The SONG telescope is part of the global SONG program, which includes 1-meter telescopes. It is located at the Lenghu Observatory in Qinghai, China. It is designed to serve two main scientific goals in stellar physics research: the detection of exoplanets by microgravitational lensing methods based on the Lucky Imaging Technique (LIT), and the study of the internal structure of stars using astroseismology methods based on apparent velocity. Telescope pointing accuracy is critical to scientific research, and high-quality data can provide more accurate and reliable results, thus advancing astronomical science. The telescope pointing error is the deviation between the actual pointing of the telescope and the expected pointing during observation. Considering the mechanical structure, driving system, atmosphere effects, sensors, and feedback errors of the telescope, the telescopes are often required to use pointing models to correct these errors. This article proposes a concept verification based on machine learning to reduce the direction error of the SONG Telescope. Using recent historical pointing data, the machine learning algorithm XGBoost is applied to train the model, which can effectively help to improve the precision of telescope pointing, thus enhencing the quality of observational data. At the same time, its results will provide effective information for the operation of the telescope in the future.