Unsupervised Learning Approaches for User Clustering in NOMA enabled Aerial SWIPT Networks

摘要

This paper studies the application of simultaneous wireless information and power transfer (SWIPT) to millimeterwave non-orthogonal multiple access (mmWave-NOMA) enabled aerial networks, where an aerial base station (ABS) sends wireless information and energy simultaneously via NOMA schemes to multiple single-antenna information decoding (ID) devices and energy harvesting (EH) devices. This paper aims to maximize the harvested sum-power of all EH devices subject to given minimum rate constraints at different ID devices. Furthermore, we develop two machine learning based clustering algorithms, namely, K-means and K-medoids, where devices' locations are extracted to model the features for clustering. Our simulation results demonstrate: 1) the impact of different clustering approaches on the sum EH power under different spatial distributions of devices; 2) the proposed machine learning based clustering framework for mmWave-NOMA enabled aerial SWIPT networks is capable of achieving considerate improvements in terms of the harvested energy compared to conventional aerial SWIPT networks.