Semantic Communication-Enhanced UAV-Assisted Task Offloading Mechanism in Interference Environments

Abstract

Semantic communication has emerged as a promising paradigm for reducing communication overhead by transmitting task-relevant semantic information, thereby enabling low-latency transmission in mobile edge computing (MEC) systems assisted by multiple unmanned aerial vehicles (UAVs) for urban digital twins.This paper proposes a collaborative task offloading framework for multi-UAV-assisted MEC systems by integrating semantic communication and semantic caching. The framework adopts a functionally decoupled heterogeneous multi-UAV architecture, in which user access, semantic encoding, and semantic decoding are assigned to different UAV nodes to improve collaborative efficiency. Moreover, a semantic fidelity model is developed to characterize the reliability of received semantic information under semantic compression and channel interference. Furthermore, a semantic similarity-based caching mechanism is designed to reduce redundant transmissions in remote sensing scenarios. Based on these designs, a joint optimization problem is formulated to optimize UAV trajectories, task offloading ratios, semantic compression ratios, transmit powers, and channel selections, with the goal of reducing system latency while maintaining semantic transmission reliability.Simulation results show that the proposed semantic communication and caching-aware multi-agent soft actor–critic (SCCA-MASAC) algorithm outperforms baseline methods such as MAD-DPG and MATD3 in terms of convergence performance and system utility. It also effectively reduces system latency while maintaining high semantic fidelity. Ablation studies further validate the effectiveness of the proposed semantic communication and semantic caching mechanisms in reducing redundant transmissions and enhancing overall system performance.