Joint Scheduling Design in Wireless Powered MEC IoT Networks Aided by Reconfigurable Intelligent Surface

Internet of things (IoT) technology is critical to realize universal connections of everything and pervasive intelligence for the future world. The forthcoming IoT technology will be characterized by two predominant features: energy self-sustainability, which is fueled by the recent thrilling wireless power transfer (WPT) technology, and sufficient computation power capability, which will be empowered by the mobile edge computing (MEC) networking. Very recently a promising technology named reconfigurable intelligent surfaces (RIS) has attracted much attention due to its effective beamforming capability and viable potentials to enhance wireless communication system. In this paper we consider exploiting RIS to enhance the WPT-based MEC IoT networks via boosting its energy transferring and communication efficiency. Specifically, we consider the scheduling design through jointly optimizing the WPT-time allocation, dynamic RIS phase control and all IoT mobile devices’ offloading decisions to improve the entire MEC network’s computation capability. This problem is very challenging due to its high dimension discrete variable space. Here we adopt a reinforcement learning (RL) based online method, which utilizes a novel double deep Q-network (DDQN) structure to effectively overcome the overestimation issue and outperforms the conventional deep Q-network (DQN) learning methods. Numerical results verify the effectiveness of our proposed algorithm and demonstrate the benefits of introducing RIS to assist the WPT-based MEC network.