Abstract
Advancements in satellite technology have enabled direct device-to-satellite communications as a viable solution for providing global internet access, particularly in remote, rural, and underserved regions where terrestrial infrastructure is unavailable or inadequate. However, conventional device-to-single-satellite systems suffer from severe free-space path loss (FSPL) and high channel correlation among satellite antenna elements due to the large user-to-satellite distance, which fundamentally limits spatial multiplexing gains. While high-power terminals can partially mitigate these challenges, power-limited devices such as mobile handsets cannot, and compensating solely through large satellite antennas is impractical due to mass and launch cost constraints. To address these limitations, this thesis investigates multi-satellite multiple-input multiple-output (Space-MIMO) techniques, with and without inter-satellite cooperation, to enhance uplink performance in direct device-to-satellite communications. After reviewing the fundamental aspects of satellite communications, the thesis focuses on uplink transmission from a single-antenna handheld user to a cluster of satellites, exploiting inter-satellite links (ISLs) to enable cooperative signal detection. Analytical and simulation results demonstrate that a large-scale constellation comprising 3,168 satellites can achieve an approximately eightfold capacity improvement when cooperation among 12 satellites is enabled, compared to conventional single-satellite access. The impact of non-ideal ISLs, including terahertz (THz) and free-space optical (FSO) links, is also examined, for which accurate closed-form capacity approximations and upper bounds are derived and validated via Monte Carlo simulations. Finally, the thesis addresses cooperative channel estimation for Space-MIMO orthogonal frequency-division multiplexing (OFDM) uplink transmission, effectively mitigating pilot contamination in multi-user scenarios by exploiting spatial orthogonality across cooperating satellites. A low-overhead sensing-matrix-based compression scheme is proposed to reduce ISL feedback requirements, achieving up to 81% improvement in spatial decorrelation and 86% reduction in normalized mean square error (NMSE) while maintaining reliable channel estimation performance.