Dynamic Interference Prediction and Receive Beamforming for Dense LEO Satellite Networks

Abstract

Dense low earth orbit (LEO) satellite networks with full frequency reuse can offer seamless global coverage and high spectrum efficiency. However, multiple satellites have overlapping coverage areas, leading to co-channel interference that degrades communication system performance. Moreover, the high dynamic nature of LEO satellites makes the interference varies over time. In this paper, we analyze the receive beamforming to mitigate the complex and time-varying interference in dense LEO satellite networks, and the interference mitigation is formulated as a long-term data rate maximizing problem. To address this problem, a joint intelligent interference prediction and receive beamforming design algorithm is proposed. First, an interference prediction algorithm based on long short-term memory (LSTM) is employed to predict the direction of arrival (DOA) information. Then, a hybrid beamforming algorithm based on deep reinforcement learning (DRL) is proposed to mitigate interference. Simulation results show that the proposed algorithm effectively improves long-term data rate for users and outperforms other benchmark algorithms.