Hierarchical Optimization for Energy-Efficient Semantic Communications with Heterogeneous Users

Semantic communication is a promising paradigm for future wireless networks, yet its practical implementation faces significant challenges in energy-efficient resource allocation, especially in systems with heterogeneous users. The joint optimization of semantic compression, computation, and communication resources formulates a large-scale, non-convex problem, which is computationally complex to solve directly. In this paper, we address this challenge by proposing a novel hierarchical optimization framework. Our main contribution is a decomposition of the original problem into a two-layer structure: an upper-level problem that determines strategic semantic compression ratios and a lower-level problem that allocates tactical computation and communication resources. For the non-convex lower-level problem, we propose an algorithm that integrates Successive Convex Approximation (SCA) and the Alternating Direction Method of Multipliers (ADMM). For the upper-level problem, we transform it into a well-structured monotonic optimization problem and solve it efficiently using a Block Coordinate Descent (BCD) procedure. Our decomposition-based approach yields a computationally efficient and high-performance solution, offering a practical framework for designing energy-efficient multi-user semantic communication systems.