MAGIC: Matching Game-Based Resource Allocation With Incomplete Information in Space Communication Network

Collaboration between low Earth orbit (LEO) and geostationary Earth orbit (GEO) satellites in space communication networks has the advantages of wider coverage and higher communication capacity. However, effective resource allocation in the space communication network faces significant challenges due to incomplete information introduced by the highly dynamic communication environment. In this work, we focus on Ma tching
${G}$
ame-based resource allocation strategy with
${I}$
ncomplete information in the space
${C}$
ommunication network, called MAGIC. Specifically, we formulate the multi-dimensional resource allocation with incomplete information as the revenue maximization problem of access satellite, which is the sum priorities of the successfully accessed users. The revenue maximization problem is a mixed integer nonlinear programming problem, and a three-sided matching game is employed to solve it. Meanwhile, we apply a model-free reinforcement learning framework to pre-train the historical network data to compensate for the shortcomings caused by incomplete information. Furthermore, user-optimal and access satellite-optimal resource allocation algorithms are designed to achieve optimal resource scheduling. Simulation results demonstrate the effectiveness and convergence of proposed algorithms from the single time slot and multiple time slot perspectives of different network parameters.