Reliable Semantic Communication With QoE-Driven Resource Scheduling for UAV-Assisted MEC

Semantic communication is a great potential technology to improve the reliability and effectiveness of communication systems, but the limited computing capacity of users, the potential jamming attacks, and the inflexibility of fixed base stations pose significant challenges to communication efficiency. To overcome those challenges, this paper proposes a joint resource scheduling approach for unpiloted aerial vehicle (UAV)-assisted mobile edge computing (MEC) system based on reliable semantic communications against malicious jamming attacks. Specifically, we maximize the system's quality-of-experience (QoE) by jointly optimizing UAV trajectory, task offloading decision, and user association. Since the problem is non-convex and the jamming behavior is dynamic, an advanced softmax deep double deterministic policy gradients (SD3) based on the prioritized experience replay (PER) algorithm (called PER-SD3) is designed to realize an efficient communication strategy against jamming. Simulation results demonstrate that the proposed PER-SD3 approach can significantly improve the system delay and semantic communication performances compared with baseline approaches in the literature.