Intelligent Channel Prediction and Power Adaptation in LEO Constellation for 6G

Abstract

Integrated satellite and terrestrial networks (ISTN) are rapidly evolving to meet the ever-increasing demands of higher throughput, lower latency, and wider coverage for future communications. Meanwhile, the ultra-dense deployment of low earth orbit (LEO) satellites has emerged as a promising architecture of ISTN, which will enable seamless broadband coverage while posing huge challenges to wireless resource allocation. Artificial intelligence can be applied to the ISTN based on the LEO constellations scenarios, endowing the resource allocation schemes with intelligence and timeliness. In this article, we have carried out a detailed analysis of the ISTN based on the LEO constellations from the network model and interference scenarios. Subsequently, considering the strict requirements of intelligent resource allocation for network state perception, a channel state information (CSI) prediction approach based on deep learning algorithms is developed, which contains the prediction of interference period and atmospheric attenuation. With the premise of perceiving CSI comprehensively, a deep reinforcement learning (DRL)-driven resource allocation scheme is developed. Simulation results demonstrate the effectiveness of the proposed DRL-driven resource allocation scheme in terms of interference mitigation and constellation capacity improvement.