A robust routing strategy based on deep reinforcement learning for mega satellite constellations

Abstract

AbstractFor mega satellite constellations, it has been a great challenge to achieve global routing and guarantee the performance of inter‐satellite transmission. To address the problem, a robust routing strategy based on deep reinforcement learning is proposed in this letter. The proposed method is applicable to degraded transmission performance, which exhibits better conformity to real‐world scenarios. Moreover, the age of information (AoI) of packets are utilized as one of the multi‐optimization targets to ensure the effectiveness of message transmission throughout the network. Numerical simulations show the outstanding average AoI performance of the proposed method and its greater robustness to jamming compared to existing methods. Meanwhile, it is also more effective for utilizing resources.