Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

CNN Autoencoder Resizer: A Power-Efficient LoS/NLoS Detector in MIMO-enabled UAV Networks

Azim Akhtarshenas, Navid Ayoobi, David Lopez-Perez, Ramin Toosi, Matin Amoozadeh

TL;DR

The paper addresses LoS/NLoS identification in UAV-enabled wireless networks, a problem intensified by power constraints on sensing hardware. It introduces the CNN Autoencoder Resizer (CAR), a dual-autoencoder framework with a shared latent vector that maps low-dimensional channel outputs to high-dimensional representations, serving as a preprocessing step to boost discrimination without increasing power consumption. CAR achieves a substantial accuracy gain, elevating LoS/NLoS detection from $66\%$ to $86\%$ while maintaining power efficiency, and is designed to work with smaller antenna arrays by transferring learned high-resolution representations from larger arrays. The approach offers a practical, scalable enhancement for UAV-based NTN deployments, enabling more reliable connectivity with reduced hardware burden.

Abstract

Optimizing the design, performance, and resource efficiency of wireless networks (WNs) necessitates the ability to discern Line of Sight (LoS) and Non-Line of Sight (NLoS) scenarios across diverse applications and environments. Unmanned Aerial Vehicles (UAVs) exhibit significant potential in this regard due to their rapid mobility, aerial capabilities, and payload characteristics. Particularly, UAVs can serve as vital non-terrestrial base stations (NTBS) in the event of terrestrial base station (TBS) failures or downtime. In this paper, we propose CNN autoencoder resizer (CAR) as a framework that improves the accuracy of LoS/NLoS detection without demanding extra power consumption. Our proposed method increases the mean accuracy of detecting LoS/NLoS signals from 66% to 86%, while maintaining consistent power consumption levels. In addition, the resolution provided by CAR shows that it can be employed as a preprocessing tool in other methods to enhance the quality of signals.

CNN Autoencoder Resizer: A Power-Efficient LoS/NLoS Detector in MIMO-enabled UAV Networks

TL;DR

The paper addresses LoS/NLoS identification in UAV-enabled wireless networks, a problem intensified by power constraints on sensing hardware. It introduces the CNN Autoencoder Resizer (CAR), a dual-autoencoder framework with a shared latent vector that maps low-dimensional channel outputs to high-dimensional representations, serving as a preprocessing step to boost discrimination without increasing power consumption. CAR achieves a substantial accuracy gain, elevating LoS/NLoS detection from to while maintaining power efficiency, and is designed to work with smaller antenna arrays by transferring learned high-resolution representations from larger arrays. The approach offers a practical, scalable enhancement for UAV-based NTN deployments, enabling more reliable connectivity with reduced hardware burden.

Abstract

Optimizing the design, performance, and resource efficiency of wireless networks (WNs) necessitates the ability to discern Line of Sight (LoS) and Non-Line of Sight (NLoS) scenarios across diverse applications and environments. Unmanned Aerial Vehicles (UAVs) exhibit significant potential in this regard due to their rapid mobility, aerial capabilities, and payload characteristics. Particularly, UAVs can serve as vital non-terrestrial base stations (NTBS) in the event of terrestrial base station (TBS) failures or downtime. In this paper, we propose CNN autoencoder resizer (CAR) as a framework that improves the accuracy of LoS/NLoS detection without demanding extra power consumption. Our proposed method increases the mean accuracy of detecting LoS/NLoS signals from 66% to 86%, while maintaining consistent power consumption levels. In addition, the resolution provided by CAR shows that it can be employed as a preprocessing tool in other methods to enhance the quality of signals.
Paper Structure (13 sections, 4 equations, 4 figures)

This paper contains 13 sections, 4 equations, 4 figures.

Table of Contents

  1. Introduction
  2. Related Works
  3. Motivation and Contribution
  4. System model
  5. Proposed Method and Architecture
  6. Dataset
  7. CNN Autoencoder Resizer (CAR)
  8. Binary CNN Classifier
  9. Result and discussion
  10. Experimental setup
  11. Comparison of mapped channel output with the original and equivalent higher channel output
  12. Evaluating the effect of the length of the filters along time access
  13. Summary and Conclusion

Figures (4)

  • Figure 1: The architecture of the proposed CNN autoencoder resizer (CAR). CAR includes two autoencoders that share a common latent vector. The first autoencoder processes the lower dimensional channel outputs while the second one processes the higher dimensional channel outputs. After training stage, CAR is able to map a low-dimensional channel output to a latent vector using the encoder in the first autoencoder, and then decodes it to a higher dimensional channel output using the decoder of the second autoencoder. The encoders and decoders are highlighted by gray and yellow backgrounds, respectively.
  • Figure 2: The pipeline of labeling channel outputs. The low-dimensional channel output is fed to the first encoder. The second decoder uses encoded vector and generates its high-dimensional counterpart. The binary CNN classifiers then processes this high-dimensional channel output to predict its label.
  • Figure 3: The performance of CAR on the accuracy of detection of LoS/NLoS by mapping the low-dimensional channel output to high-dimensional channel output.
  • Figure 4: The effect of changing the length of filters along time axis in terms of accuracy.