Image Processing Task Offloading in UAV-Assisted MEC System

In the Internet of Things (IoT), given requirements on Quality of Service (QoS), Unmanned Aerial Vehicles (UAVs) have been considered to become aerial servers for providing additional computing resources for Mobile Devices (MDs) with limited computation power. Considering the issue of the abstract scenarios and idealized computation models of the existing research, an image processing-oriented UAV-assisted computation offloading framework and computation model is first proposed. MDs offload the raw image data to the UAV for being processed, and the UAV returns the processed image data to MDs for further utilization. In this paper, the challenge of minimizing system processing delay is modeled as a Markov Decision Process (MDP) by jointly considering the offloading decision, UAV movement trajectory, dynamic channel state, and dynamic hardware processing frequency. Considering the high-dimensional state space and the continuous action space, the Deep Deterministic Policy Gradient (DDPG)-based algorithm scheme is designed to solve the UAV offloading strategy. Numerical simulation results show that the offloading scheme in the proposed solution model is significantly effective.