A DT Machine Learning-Based Satellite Orbit Prediction for IoT Applications

Abstract

Satellite orbit prediction has important applications in the field of space situation awareness, such as space collision warning and observation scheduling. The expansion of space information network challenges the low delay, high-accuracy transmission and real-time response of satellite orbit prediction tasks. The traditional orbit prediction process is affected by the measurement error, the estimation error, the unmodeled orbit perturbation and other factors, resulting in low accuracy orbit prediction results. In order to meet the requirements of high accuracy requirements, we built a satellite digital twin system based on the Docker container to predict, optimize and control the satellite orbit status in low consumption. The proposed digital twin system uses the container technology to build each module, which makes the updating of the orbit prediction model more convenient. In addition, in the designed digital twin system, we present an orbit error prediction model based on machine learning. Compared with the traditional physical dynamic model, the proposed machine learning model can effectively correct the error value of orbit prediction and improve the accuracy of orbit prediction. Finally, the validity of the corresponding model is verified in the simulation environment.