Energy-aware Multiple Access Using Deep Reinforcement Learning

Abstract

Deep Reinforcement Learning (DRL), as an emerging trend in the reinforcement learning paradigm, has recently been used for multiple access of wireless nodes to frequency spectrum. Although existing research works are promising in terms of frequency spectrum utilization, the concept of energy-awareness is missing. Nevertheless, the high energy-consumption of DRL algorithms is a serious concern, especially in battery-constrained Internet of Things (IoT) nodes. In this paper, a simple yet effective mechanism is introduced to reduce state size of the DRL algorithm, which results in reduction of energy consumption for IoT nodes. Our simulations indicate that state size can be reduced, without significant change in the system performance.