RL-Driven Semantic Compression Model Selection and Resource Allocation in Semantic Communication Systems

Semantic communication (SemCom) is an emerging paradigm that leverages semantic-level understanding to improve communication efficiency, particularly in resource-constrained scenarios. However, existing SemCom systems often overlook diverse computational and communication capabilities and requirements among different users. Motivated by the need to adaptively balance semantic accuracy, latency, and energy consumption, this paper presents a reinforcement learning (RL)-driven framework for semantic compression model (SCM) selection and resource allocation in multi-user SemCom systems. To address the challenges of balancing image reconstruction quality and communication performance, a system-level optimization metric called Rate-Distortion Efficiency (RDE) has been defined. The framework considers multiple SCMs with varying complexity and resource requirements. A proximal policy optimization (PPO)-based RL approach is developed to dynamically select SCMs and allocate bandwidth and power under non-convex constraints. Simulations demonstrate that the proposed method outperforms several baseline strategies. This paper also discusses the generalization ability, computational complexity, scalability, and practical implications of the framework for real-world SemCom systems.