Abstract:The WIYN 3.5m Telescope at Kitt Peak National Observatory hosts a suite of optical and near infrared instruments, including an extreme precision, optical spectrograph, NEID, built for exoplanet radial velocity studies. In order to achieve sub ms precision, NEID has strict requirements on survey efficiency, stellar image positioning, and guiding performance, which have exceeded the native capabilities of the telescope's original pointing and tracking system. In order to improve the operational efficiency of the telescope we have developed a novel telescope pointing system, built on a recurrent neural network, that does not rely on the usual pointing models (TPoint or other quasi physical bases). We discuss the development of this system, how the intrinsic properties of the pointing problem inform our network design, and show preliminary results from our best models. We also discuss plans for the generalization of this framework, so that it can be applied at other sites.

What problem does this paper attempt to address?

### Problems the paper attempts to solve This paper aims to solve the problem of insufficient precision in the pointing and tracking systems of the WIYN 3.5 - meter telescope, especially when high - precision radial velocity measurements are made using the NEID spectrometer. Specifically: 1. **Insufficient pointing precision**: - The current pointing systems (such as TPoint) rely on traditional regression models, which need to be recalibrated regularly and have significant time - dependent errors at certain azimuths. - These errors lead to a time - consuming target acquisition process and sometimes complete failure. 2. **Inefficient multi - stage target acquisition process**: - The target is first acquired by the Star Tracker Camera installed on the telescope and then iteratively adjusted onto the NEID scientific fiber through the secondary guiding camera multiple times. - This process is not only time - consuming but also requires operator intervention, reducing the observation efficiency. 3. **Limitations of the existing system**: - Existing systems such as TPoint need to be recalibrated many times a year, and each calibration takes a large amount of time (about half a night) to acquire reference stars. - The existing system cannot systematically track and report pointing errors, making it difficult for users to predict possible challenges or mechanical problems in advance. ### Solutions To solve the above problems, the research team has developed a new method based on deep learning, especially using Recurrent Neural Networks (RNN) to generate adaptive and accurate pointing models. The advantages of this method include: - **Higher pointing precision**: The new model can predict the position of the telescope within a range of about 0.05°, while the typical error of the existing system is about 0.2°. - **No need for additional observation time**: The required data can be obtained through normal operations, and the model can be recalibrated offline during the day, thus saving precious observation time. - **More uniform error distribution**: The new model reduces the dependence on specific azimuths and improves the overall pointing precision. In addition, the research team also plans to extend this framework to other observatories to improve the operational efficiency of more telescopes. ### Summary The main goal of this paper is to significantly improve the pointing precision and observation efficiency of the WIYN telescope by introducing a new deep - learning - based pointing system, so as to better support the high - precision radial velocity measurement tasks of the NEID spectrometer.

Deep learning solutions to telescope pointing and guiding

4-K-resolution Minimalist Optical System Design Based on Deep Learning.

Machine Learning based Pointing Models for Radio/Sub-millimeter Telescopes

Neural Network Based Point Spread Function Deconvolution For Astronomical Applications

Neural Network Based 3D Tracking with a Graphene Transparent Focal Stack Imaging System

Point spread function estimation for wide field small aperture telescopes with deep neural networks and calibration data

Telescope performance real-time monitoring based on machine learning

The R2D2 Deep Neural Network Series Paradigm for Fast Precision Imaging in Radio Astronomy

Deep Neural Network closed-loop with raw data for optical resident space object detection

Deep learning-based deconvolution for interferometric radio transient reconstruction

Perception of Misalignment States for Sky Survey Telescopes with the Digital Twin and the Deep Neural Networks

Phasing segmented telescopes via deep learning methods: application to a deployable CubeSat

Direct Exoplanet Detection Using Deep Convolutional Image Reconstruction (ConStruct): A New Algorithm for Post-Processing High-Contrast Images

Detection and Classification of Astronomical Targets with Deep Neural Networks in Wide Field Small Aperture Telescopes

Application of Physics-Informed Neural Networks in Removing Telescope Beam Effects

Neural-network-based wavefront solution algorithm for a wide field survey telescope

Stable fiber-illumination for extremely precise radial velocities with NEID

Direct Exoplanet Detection using Convolutional Image Reconstruction (ConStruct): A New Algorithm for Post-processing High-contrast Images

Model Optimization for Deep Space Exploration via Simulators and Deep Learning