Constellation Optimization for GEO Satellite Communication System via Heuristic and Autoencoder Frameworks = 정지궤도 위성 통신 시스템에서 휴리스틱 및 오토인코더 기반의 성상도 최적화 연구

다국어 초록 (Multilingual Abstract)

With the advancement of 6G and Non-Terrestrial Networks (NTN), research in satellite communications is actively progressing. Although low Earth orbit (LEO) constellations are gaining significant attention, geostationary Earth orbit (GEO) satellites remain crucial for key applications such as meteorology, space weather, defense, and surveillance, necessitating continued research on GEO systems. A primary challenge in satellite communication is the severe scarcity of frequency resources, making efficient spectrum utilization a paramount task. Power-domain based non-orthogonal multiple access (NOMA) frameworks have been proposed to enhance spectral efficiency. However, in multi-satellite environments, these methods inevitably introduce significant multi-user interference (MUI) and inter-satellite interference (ISI). Furthermore, conventional receivers, such as successive interference cancellation (SIC), exhibit clear performance limitations, including error floors, in these complex interference scenarios.

This paper proposes an End-to-End (E2E) transmission framework based on an Autoencoder (AE) model to address these challenges. Instead of using fixed geometric constellations, this framework reconfigures the constellation by directly learning from the channel and interference environment. This involves a joint training process where the satellite (Encoder) and the ground base station (GBS) (Decoder) are optimized together under complex interference channel conditions. Through this process, the autoencoder autonomously designs a new, optimized constellation that minimizes the impact of interference and maximizes symbol distinguishability, tailored specifically to the current channel environment.

The proposed AE-based framework can manage interference and decode signals far more efficiently than traditional detection methods used in multi-satellite NOMA environments. It is expected to make a substantial contribution to enhancing the reliability and efficiency of future 6G satellite communication systems, which will face increasingly complex interference scenarios.

목차 (Table of Contents)

• List of Figures iii
• List of Tables iv
• List of Abbreviations v
• Abstract vii
• 1 Introduction 1
• 1.1 Background and Motivation 1
• 1.2 Outline and Contributions 8
• 2 Literature Review 10
• 2.1 Satellite Communication Systems 10
• 2.2 Non-Orthogonal Multiple Access in Satellite Networks 13
• 2.3 Autoencoders-based Communication Systems 16
• 3 Constellation Reconfiguration for Data Rate Improvement in GEO Satellite Networks 20
• 3.1 System Model (Single GEO Satellite and Single GBS link) 21
• 3.1.1 Link Budget Analysis 22
• 3.1.2 Channel model: Shadowed-Rician fading channel 23
• 3.2 Heuristic Constellation Reconfiguration 25
• 3.3 Numerical Results 26
• 3.4 Summary 28
• 4 Autoencoder-Based Constellation Redesign for Joint Operation of Dual GEO Satellites 30
• 4.1 System Model (Dual GEO Satellite and Single GBS Link) 31
• 4.2 Proposed Autoencoder-Based Framework 32
• 4.2.1 Neural Network-based Encoding at GEO Satellites 33
• 4.2.2 Neural Network-based Decoding at the GBS 35
• 4.2.3 Loss Function and Training 36
• 4.3 Numerical Results 37
• 4.3.1 Conventional Detection Methods 38
• SIC detection 38
• JML detection 39
• 4.3.2 Performance Comparison: AE vs. Traditional Methods 43
• 4.3.3 Constellation Reconfiguration 46
• 4.3.4 AE Performance with Different Modulation Schemes 48
• 4.3.5 Impact of Channel Coding on AE Performance 51
• 4.4 Summary 54
• 5 Conclusion 56
• 5.1 Concluding Remarks 56
• 5.2 Future Research Directions 58
• Bibliography 61
• Korean Abstract 71
• Acknowledgements 73