DRL based partial offloading for maximizing sum computation rate of FDMA-based wireless powered mobile edge computing

Abstract

Since most Internet of Things (IoT) nodes are powered by small-capacity batteries and equipped with small computing units, it is difficult for them to process computing-intensive and delay-sensitive tasks. To address these two limitations, a feasible approach is to use both wireless power transmission (WPT) and mobile edge computing (MEC) technologies. Wireless edge devices (EDs) have stable energy supply by capturing radio frequency (RF) energy transmitted by the energy source and offloading computation workload to edge computing servers (ECSs) to enhance the computation rate. In this paper, an FDMA-based wireless-powered MEC network is considered, which consists of an ECS, an RF energy source and multiple EDs. The research goal is to maximize the sum computation rate (SCR) by jointly optimizing the WPT duration, energy allocation for local computation, and computation offloading, and bandwidth allocation among EDs. We first formulate this as a non-convex optimization problem that is hard to address. Then, we decompose it into a top-problem for optimizing the WPT duration and a sub-problem for optimizing the energy allocation and bandwidth allocation under a given WPT duration. An online offloading algorithm based on deep reinforcement learning (DRL) is proposed to quickly obtain the near-optimal WPT duration. For the sub-problem, we design an efficient two-stage algorithm to obtain the optimal solution. The simulation results show that the proposed algorithm can obtain the near-optimal solutions with low computation complexity.