Multi-Agent Deep Reinforcement Learning Based Channel Allocation for Networked Satellite Telemetry System

Numerous mega low earth orbit (LEO) satellite constellation plans have recently emerged as an indispensable part of the future satellite communication era. Since the traditional ground-based and geostationary earth orbit (GEO)-based telemetry systems are unsuitable for monitoring the operation status of mega constellations, a networked telemetry system is adopted to achieve full time, low delay telemetry in this paper. In order to satisfy the data transmission requirements of extensive satellites, this paper formulates the channel allocation problem, which aims at maximizing the overall transmitted data value by allocating different medium earth orbit (MEO) beams in different time slots to different LEO satellites. Since the data generation states of LEO satellites are hybrid constant and stochastic, the MEO satellites could allocate channels more timely than the ground mission center, and the action space for channel allocation is too large, the multi-agent deep reinforcement learning based algorithm is consequently adopted to solve the channel allocation problem. This paper verifies the effectiveness of our proposed channel allocation algorithm by numerical simulation.