Implementation of DNN-Based Physical-Layer Network Coding

Physical-layer network coding (PNC) has been proposed to double the throughput of a wireless two-way relay network (TWRN) and has been adapted so that it can be applied in vehicular ad-hoc networks (VANETs). Deep neural network (DNN)-based physical-layer network coding (PNC) for the TWRN system has been studied in recent years, under the assumption that the channel state information (CSI) is available to the end nodes and the relay node. However, the state-of-the-art DNN-based PNC system was only feasible for theoretical simulations, and no practical system design has been studied. When moving from theory to practical wireless systems, there are two critical issues to tackle: 1) unlike conventional regularly shaped quadrature amplitude modulations (QAMs), the constellation points of a DNN-based PNC system's end nodes are irregular and unpredictable to the relay node; and 2) other than in high-precision simulations, the practical DNN-based PNC system's constellation points are not fixed, that is, there is usually a cluster of points that must be grouped together so that the transceiver's digital-to-analog converter (DAC) and analog-to-digital converter (ADC) can work properly. To solve the above issues, the authors proposed the DNN-PNC implementation that can be realized using the Universal Software Radio Peripheral (USRP) platform. The computational complexity of the authors' simulated DNN-PNC is one order of magnitude smaller than that given in the literature, while maintaining a good bit error rate (BER). In the implementation phase, the authors find that the naturally-happened power imbalance between the two end nodes' signals in the authors' implementation can significantly boost the PNC constellation recovery, and hence reduce the system BER. The authors' experimental results showed that the BER of DNN-PNC can be lower than $10^{-4}$ when the signal-to-noise ratio (SNR) is 16 dB for the 2-QAM case and 20 dB for the 4-QAM case, respectively. These results indicate that DNN-based PNC is feasible for VANET applications, where the CSI changes and is difficult to obtain.

• Record URL:
  http://dx.doi.org/10.1109/TVT.2023.3237589
• Availability:
  • Find a library where document is available. Order URL: http://worldcat.org/issn/00189545
• Supplemental Notes:
  • Copyright © 2023, IEEE.
• Authors:
  • Wang, Xuesong
  • Lu, Lu
• Publication Date: 2023-6

Language

• English

Media Info

• Media Type: Web
• Features: References;
• Pagination: pp 7380-7394
• Serial:
  • IEEE Transactions on Vehicular Technology
  • Volume: 72
  • Issue Number: 6
  • Publisher: Institute of Electrical and Electronics Engineers (IEEE)
  • ISSN: 0018-9545
  • Serial URL: http://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=25

Subject/Index Terms

• TRT Terms: Coding systems; Motor vehicles; Neural networks; Vehicular ad hoc networks; Wireless communication systems
• Subject Areas: Data and Information Technology; Highways; Vehicles and Equipment;

Filing Info

• Accession Number: 01886092
• Record Type: Publication
• Files: TRIS
• Created Date: Jun 28 2023 9:09AM