Energy Efficiency Optimization for IoT Systems With Reconfigurable Intelligent Surfaces: A Self-Supervised Reinforcement Learning Approach

In this paper, we present an optimal uplink transmission strategy to maximize the energy efficiency for Internet of Things (IoT) assisted by reconfigurable intelligent surface (RIS). To quickly and accurately obtain the optimal transmission power of the IoT devices and the phase shift vector of the RIS, we develop a self-supervised double deep Q network (SSDDQN) learning algorithm. We demonstrate that the SSDDQN-based method achieves energy efficiency comparable to global-optimal mathematical methods while significantly reducing execution time during inference. Additionally, numerical results show that SSDDQN reduces training time costs by 33.3% compared to traditional DDQN.