Down-clocking Scheme using Deep Learning for Minimizing Energy Consumption in Wireless Networks
Jae-Hyeon Park, Seung Hyun Jeong, Young-Joo Suh · 2020
Wi-Fi interface is known to consume a lot of energy in mobile devices, and Idle Listening (IL) dominates clients' energy consumption in Wi-Fi. In this paper, we propose IL down-clocking schemes using deep learning model to reduce the energy consumption in IL time. We exploit the orthogonal frequency-division multiplexing (OFDM) subcarrier addressing for the preamble design. To minimize preamble length for energy efficiency, we use a deep learning model with the recurrent neural network (RNN). Our experimental evaluation using OPNET network simulator and USRP/GNU Radio implementation shows that our scheme outperforms the state-of-the-art down-clocking scheme in both energy consumption and network throughput.