Title and Abstract
Title: Machine Learning-Enabled Orbit Prediction and Earth Observation
Abstract: Since the launch of the first satellite (Sputnik 1) in 1957, humans have created a lot of objects in orbit around Earth. The number of space objects larger than 10 cm is presently more than 21,000, the estimated population of objects between 1 and 10cm is about 500, 000, and for objects smaller than 1cm the number exceeds 100 million. Both the number of space objects and the number of conflicts between these objects are increasing exponentially.
The first part of this talk overviews the research we have been pursuing on to address the challenges posed by the growth of space debris. We will first introduce the Modified Chebyshev-Picard Iteration (MCPI) Methods, which are a set of parallel-structured methods for solution of initial value problems and boundary value problems. The MCPI methods have been recommended as the “promising and parallelizable method for orbit propagation” by the National Research Council. The talk will then highlight our recent results to develop a methodology to predict RSOs trajectories both higher accuracy and higher reliability than those of the current methods. Inspired by the learning theory through which the models are learnt based on large amounts of data and the prediction can be conducted without explicitly modeling space objects and space environment, we are working on a new orbit prediction framework that integrates physics-based orbit prediction algorithms with a learning process.
The second part of this talk will present a novel method to guide an Earth observation satellite equipped with maneuverable sensors for remote sensing, using measuring total column ozone as the application. The resulting task planning is intelligent since it achieves the goal not only to find the extreme values but also to understand what happens across the whole studied area. We develop a Bayesian Optimization-based method to achieve the goal.