A DRL-Framework for Full-Duplex WPCN Enabled Mobile Edge Computing

As the Internet of Things (IoT) continues to expand in both daily life and industrial production, the data generated by IoT devices is increasing rapidly. To address the challenges of limited battery life and computational capacity in these devices, two promising solutions have emerged: wireless power transfer (WPT) and mobile edge computing (MEC). By receiving radio frequency (RF) energy from a hybrid access point (HAP), wireless devices (WDs) can either offload data to the HAP or process it locally. In this paper, we explore the full-duplex WPT-MEC scenario and maximize the sum computation rate (SCR) by optimizing the offloading decision and resource allocation of WDs in different access modes and energy harvest models. This problem is formulated as a mixed-integer nonlinear programming (MINLP) problem. To address this, we decompose the problem into two sub-problems: one focuses on optimizing offloading decisions, and the other focuses on energy power and bandwidth allocation under given the offloading decisions. We propose a deep reinforcement learning approach based on policy gradients to make offloading decisions, and we design an optimization algorithm for efficient bandwidth and energy resource allocation. Simulation results show that our proposed algorithm achieves approximately 96% of the maximum possible computation rate while maintaining low computational complexity