SF24B-T007 - Finding Guaranteed RL Control for Satellite Systems

Neoskye proposes an innovative solution to enhance satellite control systems' stability, adaptability, and efficiency by developing and deploying EquilibriaNet. This advanced technology integrates sophisticated machine learning techniques with state-of-the-art control algorithms to address the challenges of nonlinear and dynamic space environments. By leveraging model-free deep reinforcement learning (DRL) and zeroth-order optimization, EquilibriaNet ensures robust performance and adaptability without relying on predefined models.The project will focus on three critical technical objectives: Algorithm Modification and Model Development, Model Training, and Testing and Validation. The first objective involves refining control algorithms and creating a machine-learning model for precise satellite maneuvering. The second objective is to train this model using diverse datasets, including historical and synthetic satellite data, to ensure its capability to manage the complexities of space operations. The final objective is rigorously testing and validating the model in high-fidelity simulation environments to confirm its reliability and effectiveness.Neoskye's approach is built on a solid foundation of extensive experience in autonomous systems and control technologies. Notably, Neoskye has successfully implemented the Tactical Vision System (TVS) for the Army's Datalink project and developed an autonomous maneuvering system for NASA's space debris capture missions. By integrating advanced AI techniques and drawing on the insights gained from these past projects, Neoskye is poised to create a cutting-edge control system that will significantly enhance satellite operations' operational readiness and efficiency for the U.S. Space Force and other federal agencies.