A DRL-Based Partial Offloading Strategy for WP-MEC With Multiple Access Points

The integration of wireless power transfer (WPT) and mobile edge computing (MEC) provides an effective solution for overcoming the energy and computational limitations of Internet of Things (IoT) devices by enabling them to harvest energy from radio frequency (RF) signals and offload data to edge servers. A crucial challenge in wireless-powered MEC (WP-MEC) networks is how to efficiently optimize offloading decisions and resource allocation to enhance overall system performance. In this article, we investigate the partial offloading strategy within a WP-MEC network consisting of multiple hybrid access points (HAPs). The optimization problem is formulated as a mixed-integer nonlinear programming (MINLP) problem with variables of WPT duration, offloading decisions, and energy allocation. To solve this problem, we propose a deep reinforcement learning (DRL)-based framework, which employs a neural network architecture combining convolutional and fully connected layers to output offloading decisions. In addition, we design an optimization algorithm for the joint optimization of WPT duration and offloading proportions. Numerical results demonstrate that the proposed method achieves better performance than the existing DRL methods, which demonstrates the efficiency of the proposed method.