Satellite Navigation and Coordination with Limited Information Sharing

TL;DR

This paper investigates transfer learning for collision avoidance in space traffic management, demonstrating improved performance with limited information sharing and robustness to satellite dynamics perturbations.

Contribution

It introduces a transfer learning approach for multi-agent collision avoidance in space, outperforming models trained directly on space environments and assessing information-sharing benefits.

Findings

Transfer learning outperforms direct training in space environments.

The approach remains effective despite satellite dynamics perturbations.

Methods can evaluate benefits of information-sharing among satellite operators.

Abstract

We explore space traffic management as an application of collision-free navigation in multi-agent systems where vehicles have limited observation and communication ranges. We investigate the effectiveness of transferring a collision avoidance multi-agent reinforcement (MARL) model trained on a ground environment to a space one. We demonstrate that the transfer learning model outperforms a model that is trained directly on the space environment. Furthermore, we find that our approach works well even when we consider the perturbations to satellite dynamics caused by the Earth's oblateness. Finally, we show how our methods can be used to evaluate the benefits of information-sharing between satellite operators in order to improve coordination.