Uplink Capacity Enhancement for Workload-weighted FD-SWIPT-NOMA Networks Based on Deep Learning

Zhipeng Feng, Changhao Du, Han Liu, Jianping An, Zhongshan Zhang

Research output: Contribution to journalArticlepeer-review

Abstract

The next generation mobile communication system has put forward higher requirements for uplink capacity. A system integrating full-duplex (FD), non-orthogonal multiple access (NOMA), and simultaneous wireless information and power transfer (SWIPT) techniques (named as FS-NOMA systems) has exhibited considerable technical advantages in terms of uplink spectral efficiency (SE). With the exponential growth of mobile data traffic, the devices’ workload will also seriously affect the system performance. Thus, to further enhance the uplink capacity, the workload-weighted FS-NOMA (WFS-NOMA) system must be investigated. To reflect the impact caused by workload, true workload-weighted sum-rate (TWSR) is employed to measure the performance of WFS-NOMA. In this paper, an uplink enhanced scheme (UES) based on deep learning is proposed to address the uplink capacity enhancement problem. Unlike the conventional iterative schemes which usually need a relatively longer computational time to perform an iterative process, the optimal uplink powers of WFS-NOMA users can be automatically determined depending on a deep neural network (DNN) with a shorter computational time compared with the iterative schemes. In addition, a typical transmit power allocation (TPA) scheme called power back-off allocation (PBA) is evaluated as the TWSR benchmark of the uplink unenhanced scheme (UUS), which also constitutes the skip connection of the proposed DNN for accelerating convergence. The closed-form expression of TWSR with PBA is derived and clarified. Moreover, an iterative scheme is proposed as the TWSR performance benchmark of the UES. Numerical results show that the implementation of the proposed UESs substantially improve the uplink capacity compared with that of the UUS. The DNN-based UES is capable of offering comparable capacity and less computational complexity compared with that of the iteration-based UES even under heavy work load scenarios.

Original languageEnglish
Pages (from-to)1-14
Number of pages14
JournalIEEE Transactions on Vehicular Technology
DOIs
Publication statusAccepted/In press - 2024

Keywords

  • NOMA
  • SWIPT
  • deep neural network
  • full-duplex
  • uplink enhancement

Fingerprint

Dive into the research topics of 'Uplink Capacity Enhancement for Workload-weighted FD-SWIPT-NOMA Networks Based on Deep Learning'. Together they form a unique fingerprint.

Cite this