Abstract
This paper investigates the energy efficiency maximization problem in a resource-constrained visible light communication (VLC)-based probabilistic semantic communication (PSCom) system. In the considered model, light-emitting diode (LED) transmitters perform semantic compression, reducing data size at the cost of computation overhead. The compressed semantic information is transmitted to the users for semantic inference based on a shared knowledge base, which requires regular updates to maintain synchronization. Rate splitting multiple access (RSMA) is used to transmit both knowledge base and information data simultaneously. The goal is to maximize the energy efficiency of the system through optimizing transmit beamforming, direct current (DC) bias, rate allocation, and semantic compression ratio, considering both communication and computation costs. An alternating optimization algorithm, utilizing successive convex approximation and Dinkelbach method, is proposed to solve the problem. Simulation results validate the effectiveness of the proposed approach.