Joint Edge Computing and Semantic Communication in UAV-Enabled Networks

In this letter, we investigate a joint edge computing and semantic communication in the UAV-enabled network. Therein, a UAV executes tasks offloaded from ground user equipments (UEs). Meanwhile, it acts as an IoT device to provide image data to a Metaverse platform through a ground base station (BS). A semantic communication (SemCom) technique is implemented at the UAV to extract scene graphs from captured images. The UAV then transmits the scene graphs (rather than the orignal images) to the BS. The small size of the scene graphs allows the UAV to transmit multiple images to the BS within a short duration. However, the scene graph extraction consumes computing resource, which may reduce the performance of the task offloading of the UEs. Therefore, we aim to optimize the UAV's computing resource allocation and the fractions of the tasks offloaded from the UEs to achieve the min-max between i) the total latency of image collection, scene graph extraction, and scene graph communication, and ii) the computation offloading latency over the ground UEs. Given the dynamics and uncertainty of the wireless channels, the distances between the UAV and UEs, and the computing resources of the UEs, we leverage the Advantage Actor-Critic (A2C) and Proximal Policy Optimization (PPO) to solve it.