Joint Extraction Ratio, Mode Selection, and Channel Allocation for Semantic-based D2D Communications

Device-to-device (D2D) communication is considered a key technology of 5G. Integrating semantic communication (SemCom) into D2D communication has the potential to optimize information content based on user intent and context, thereby enhancing efficiency in utilizing limited wireless resources and improving user experience. However, SemCom introduces additional computational costs. Thus, addressing the practical considerations of whether and how to integrate SemCom and D2D communication remains a pressing issue. This paper proposes a transmission framework in which the traditional mode (BitCom) and the SemCom mode coexist, and then an efficient distributed dynamic resource management problem is proposed to minimize the system’s cost. Firstly, we construct cost models for the two communication modes based on latency and energy consumption. Next, we propose an optimization problem for joint semantic extraction ratio, transmission mode selection, and channel allocation. To efficiently address the problem, we decompose the original problem into two parts. With the given D2D mode selection and channel allocation, we optimize the semantic extraction ratio by leveraging convex optimization methods. Subsequently, with the given semantic extraction ratio, we formulate the mode selection and channel allocation problem as a stochastic game and demonstrate the existence of a Nash equilibrium (NE). Finally, we propose a stochastic learning algorithm with multi-agent to converge to the NE solution. Simulation results verify that under conditions of severe mutual interference, the system cost can even be reduced by 39% compared to the single BitCom mode.