Learning-Based Energy-Aware Channel Selection for Machine Type Communications

Abstract

Ordered channel selection plays a key role in machine type communications (MTC) to realize Industry 4.0. The conventional MTC approaches lead to poor performance mainly due to ignoring channel state information (CSI) uncertainty, channel selection coupling and the long-term constraint of energy consumption. In this paper, we propose two kinds of energy-aware channel selection algorithms. With global state information (GSI), we leverage Lyapunov optimization and matching theory to propose an energy-aware matching-based channel selection algorithm named E-MGSI. Then, we extend it into the scenario that the CSI is unavailable for machine type devices (MTDs). To augment E-MGSI with leaning capability, a matching-learning-based algorithm named E-MUCB is proposed to overcome information uncertainty in a low-complexity manner. Theoretical analysis explicitly proves that the proposed E-MUCB provides a bounded deviation from the throughput performance derived with GSI. Finally, simulations are carried out to verify that the proposed algorithms can maximize network throughput while satisfying the long-term energy consumption constraint.