Generative Semantic Communication for Efficient Utilization of Satellite Resources in 6G

ABSTRACT In this paper, a two‐stage resource allocation framework is proposed for generative semantic communication (GSC)–enabled multibeam satellite systems. By integrating semantic compression and generative artificial intelligence, GSC enables efficient downlink transmission by allowing users to reconstruct content from compressed semantic prompts. The proposed framework jointly optimizes the selection of generative cells and generative user ratios as well as bandwidth and power allocation through a multiobjective optimization problem. To reduce complexity, a constrained subset search method and a pruning‐based strategy are introduced, achieving near‐optimal performance with significantly reduced computation compared with exhaustive search. Simulation results show that the proposed method outperforms conventional ones without GSC in terms of success rate, and goodput under moderate‐to‐heavy traffic demand, underscoring GSC's potential for a wide range of satellite‐based applications, including disaster relief, defense, autonomous driving, and large‐scale extended reality.