An intelligent deep learning approach to spacecraft attitude control: The case of satellites

Abstract

This paper presents a method using artificial intelligence (AI) and deep learning to transform an on- board camera into an intelligent star sensor. This method overcomes the problems and limitations of conventional star sensors, such as cost, size, weight, power consumption, camera constraints and orientation accuracy. The method proposes an intelligent processing module embedded in the camera to determine the attitude of the spacecraft, and uses two convolutional neural network models to improve the resolution of star images and determine the attitude of the spacecraft from these images. This approach guarantees high spacecraft orientation accuracy and real-time processing, contributing to the success of space missions. The off-line phases of collecting and pre-processing stellar images, building intelligent models and labeling training data are described in detail. The results show that this method offers improved performance over conventional star sensors, and presents significant contributions to scientific research and technological innovation in the field of space remote sensing, contributing to sustainable development and understanding of the universe.