Satellite Maneuver Detection Using Machine Learning and Neural Network MethodsBehaviors
Nicholas Perovich, Zachary Folcik, Rafael Jaimes · 2022 IEEE Aerospace Conference (AERO) · 2022
This study explores methods to improve the accuracy of satellite maneuver detection by comparing traditional statistical methods with machine learning and neural network-based methods. The two alternative methods explored are a Random Forest algorithm and a Deep Neural Network. Simulation data is generated and appropriate features are selected and computed. Data is split into training and validation. Next, alternative methods are trained and their performance is validated. Optimization techniques are then applied to achieve better results. Finally, the three methods are compared to each other. Conclusions and future work are described.