Semantic Communication-aided Computation Offloading for Latency-aware Applications in Edge Computing Networks

Semantic communications represent an emerging communication paradigm based on its novel concept of “semantic meaning transmission”. Semantic communications can be effectively harnessed for computation offloading, thereby offering an innovative approach to enhance transmission efficiency by conveying semantic information. However, minimizing transmission delay tends to increase computational loads in semantic communication-aided edge computing networks. In this work, we propose a semantic communication-aided computation offloading framework. By introducing a semantic extraction factor, the proposed framework can achieve a balance among transmission time, computation time, and task execution performance. Then, we formulate a task execution time minimization problem for latency-aware applications with semantic communications. To obtain the optimal resource allocation decisions with low time complexity, we propose a deep reinforcement learning-based semantic offloading (DRL-SO) algorithm. Simulation results are provided to demonstrate that the proposed DRL-SO algorithm is capable of enhancing resource utilization and diminishing task response time.