Data-Driven Energy Profiling for Resource-Efficient 5G Vertical Services.

Energy efficiency is a major concern in the development of future mobile networks. Besides the infrastructure, a significant challenge is the power consumption of the user equipment, as it directly affects the quality of experience. We have conducted comprehensive laboratory measurements on the latest commercial 5G devices to assess power consumption, considering different frequencies, bandwidths and duplex patterns. A key result is the non-linear increase in power consumption with uplink transmit power. An empirical model with machine learning methods is proposed to enable quantitative analysis of the power consumption based on the communication behavior. By applying this model to an extensive data set, spatiotemporal predictions of real-world user equipment power consumption were performed. Depending on the deployment location and communication behavior, battery life can be as much as five times lower. These results can be utilized to perform energy-aware scheduling and deployment site selection to enhance the energy footprint of mobile platforms and Internet of Things devices.