UAV-Enabled Secure Data Collection and Energy Transfer in IoT via Diffusion Model-Enhanced Deep Reinforcement Learning

Abstract

The Internet of Things (IoT) serves a vital function in supporting real-time decision-making across various applications by facilitating seamless data exchange between devices. However, as the IoT networks typically exchange data over wireless channels, the data transmission process is highly susceptible to malicious interference from jammers in the environment. Moreover, ensuring the freshness of the collected data of the decision center and managing the limited energy resources of IoT devices present significant challenges in IoT networks. In this paper, we consider a unmanned aerial vehicle (UAV)-assisted IoT network in the presence of a jammer, where the UAV is deployed to charge IoT devices through radio frequency (RF) energy transfer, and the IoT devices subsequently use the harvested energy to upload sensing data to the UAV using time division multiple access (TDMA). We aim to minimize both the secure age of information (AoI) of IoT devices and the energy consumption of the UAV by optimizing the UAV trajectory, IoT device scheduling, and the proportion of data transmission duration. Given the non-convex and dynamic nature of this optimization problem, we propose a diffusion model-enhanced twin delayed deep deterministic policy gradient (DMTD3) algorithm to solve the problem. Specifically, considering the analytical and reasoning capabilities of the diffusion model, we integrate it into the actor network of TD3 to generate rational actions based on the observed state. Simulation results demonstrate the effectiveness of the proposed DM-TD3 algorithm compared to five benchmark approaches.