Date of Award
12-2023
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Department of Electrical and Computer Engineering
First Advisor
Stephen C. Cain, PhD
Abstract
Multi-agent systems and swarms in spacecraft formation flying are of ever-increasing importance in a contested space environment—use of multiple spacecraft to contribute to a cooperative mission potentially increases positive outcomes on orbit, while autonomy becomes an ever more important requirement to reduce reaction time in dynamic situations and lower the burden on space operators. This research explores difficult swarm Guidance Navigation and Control (GNC) scenarios using Deep Reinforcement Learning (DRL). DRL polices are trained to provide guidance inputs to agents in multi-agent swarm environments for completing complex, teamwork focused objectives in geosynchronous orbit. An example scenario is explored for a group of satellite agents moving to triangulate an object in a relative orbit space that has maneuvered and is moving in the relative frame. Complex reward structures are used to encourage guidance that positions swarm members for minimizing triangulated position error, using angles-only information for navigation relative to the target. A multi-view triangulation method is proposed for the on-orbit satellite swarm scenario, with a unique weighting system to improve performance with low powered sensors. A framework is then created for scaling a cooperative swarm to large numbers of agents with fast RL model implementation. Finally, a framework is explored for limiting observation sharing between agents while maximizing performance with low powered sensors and increasing scalability of the swarm. The framework and experimental results in this research may be used in future engineering efforts, simulation and training events, or as planning tools for operators.
AFIT Designator
AFIT-ENG-DS-23-D-003
Recommended Citation
Yielding, Nicholas J., "MARSS: Multi-Agent Reinforcement learning for Satellite Swarms" (2023). Theses and Dissertations. 7675.
https://scholar.afit.edu/etd/7675
Included in
Artificial Intelligence and Robotics Commons, Navigation, Guidance, Control and Dynamics Commons
Comments
A 12-month embargo was observed for posting this dissertation on AFIT Scholar.
Approved for public release ; PA case number on file.