Deep Reinforcement Learning Based Task Offloading and Resource Allocation for MEC-Enabled IoT Networks

Mobile Edge Computing (MEC) is emerging as a paradigm for meeting the ever-increasing demands of Internet of Things (IoT) applications in real time. Furthermore, the incorporation of renewable energy harvesting capabilities into base stations or IoT devices has the potential to reduce grid energy consumption. However, it is critical to make an efficient decision for task offloading and resource allocation in order to fully utilize system potential and reduce carbon emissions. In this paper, we propose a carbon-aware MEC framework for a hybrid renewable and grid-energy MEC system. In particular, we aim to optimize the system’s task queue length and carbon emissions in an unpredictable environment with stochastic tasks and renewable energy arrivals. We first formulate the joint task offloading and resource allocation problem by optimizing the total system cost (including task queue length and carbon emissions) and then propose a DDPG-based joint optimization strategy, eventually obtaining an effective resolution through continuous action space learning in the changing environment. Numerical results confirm that our proposal can yield efficient offloading and reduce carbon emissions for the proposed MEC network.