Adaptive Resource Allocation Considering Power-Consumption Outage: A Deep Reinforcement Learning Approach

Abstract

A major feature of mobile devices in the future wireless network is the high data rate. However, the newly proposed power-consumption outage indicates that the heat generated by high data rate devices has an influence on the performance. This correspondence investigates a novel resource allocation scheme considering power-consumption outage. Specifically, based on the analysis of the heat transfer model in the smartphone, we formulate the problem to jointly allocate the downlink power and bandwidth while considering the impact of power-consumption outage. According to the dynamic feature of the problem, the Markov decision process (MDP) is utilized to model it. Furthermore, dut to the continuity of action space, the problem is handled by normalized advantage function (NAF), a deep reinforcement learning (DRL) algorithm. The effectiveness of the proposed scheme is corroborated in the simulation results on the basis of which its applicability for services with different features is further discussed.