A novel adaptive code block group hybrid automatic repeat request scheme for low earth orbit satellite networks

• Published in: International journal of satellite communications and networking 2021-09, Vol.39 (5), p.570-589
• Main Authors: Deng, Zhongliang, Liu, Hao, Wang, Ke, Lin, Wenliang, Yu, Xiaoyi, Zheng, Shixuan
• Format: Article
• Language: English
• Subjects: 5G‐and‐beyond networks; Algorithms; Automatic repeat request; HARQ; LEO; Low earth orbit satellites; Low earth orbits; Machine learning; Markov processes; MDP; Memory management; Satellite communications; Satellite networks
• ISSN: 1542-0973; 1542-0981
• DOI: 10.1002/sat.1402

Summary Low earth orbit (LEO) satellite networks provide important support to expand 5G‐and‐beyond networks to realize global coverage. Hybrid automatic repeat requests (HARQs) can greatly improve the reliability of transmission through dynamic satellite channels. Due to the large propagation delay in LEO satellite channels, the biggest challenges of HARQ in satellite networks are the delayed channel state information (CSI) and overflow of receiver memory. In this paper, we propose a novel adaptive code block group HARQ (ACBG‐HARQ) scheme that achieves lower retransmission rates with reliable data receiving. First, we build a new channel state model (CSM) to describe the relationship between the packet error rate (PER) and global channel degradation. Three states of receiver memory allocation for HARQ are designed with respect to the channel state and memory occupation: idle state, non‐full state, and full state. Then, we propose an optimal HARQ scheme for LEO satellite networks with the Markov decision process (MDP) based on the fuzzy Q‐learning (FQL) algorithm to improve the accuracy of ACBG‐HARQ state transfer, which contributes to adaptive selection of the HARQ code block group size. Simulation results show that the ACBG‐HARQ scheme can achieve a maximum spectrum efficiency gain of 10% and that the probability of receiver memory overflow is reduced by 1%∼6%.