Exploration of Adaptive Energy Optimization for 5G Roadside Unit

Abstract

High energy consumption of roadside unit (RSU) is a great challenge to the deployment of 5G Vehicle-to-Infrastructure(V2I) communication network in large scale. Compared with letting all RSUs work all day, scheduling RSU according to actual V2I traffic load is no doubt an intuitive and ideal energy optimization solution. In this paper, an energy optimization model is proposed to maximize the working efficiency per unit energy consumption of RSU under the premise of meeting V2I communication needs. The periodic statistical features of the travel time and the amount of V2I communication traffic when vehicles passing through RSUs are introduced as the model inputs to endow the model with the adaptability to real-time traffic flow. Afterwards, the method of deep reinforcement learning is applied to solve the approximate optimal solutions. Experiment results demonstrate that the proposed method is feasible and effective, as it can adaptively adjust the duty RSUs with the change of traffic flow to reduce the overall energy consumption compared with other simplistic energy-saving methods that are usually used in practice.