Machine Learning enabled Spectrum Sharing in Dense LTE-U/Wi-Fi Coexistence Scenarios

Adam Dziedzic, Vanlin Sathya, Muhammad Iqbal Rochman, Monisha Ghosh, Sanjay Krishnan

Abstract: The application of Machine Learning (ML) techniques to complex engineering problems has proved to be an attractive and efficient solution. ML has been successfully applied to several practical tasks like image recognition, automating industrial operations, etc. The promise of ML techniques in solving non-linear problems influenced this work which aims to apply known ML techniques and develop new ones for wireless spectrum sharing between Wi-Fi and LTE in the unlicensed spectrum. In this work, we focus on the LTE-Unlicensed (LTE-U) specification developed by the LTE-U Forum, which uses the duty-cycle approach for fair coexistence. The specification suggests reducing the duty cycle at the LTE-U base-station (BS) when the number of co-channel Wi-Fi basic service sets (BSSs) increases from one to two or more. However, without decoding the Wi-Fi packets, detecting the number of Wi-Fi BSSs operating on the channel in real-time is a challenging problem. In this work, we demonstrate a novel ML-based approach which solves this problem by using energy values observed during the LTE-U OFF duration. It is relatively straightforward to observe only the energy values during the LTE-U BS OFF time compared to decoding the entire Wi-Fi packet, which would require a full Wi-Fi receiver at the LTE-U base-station. We implement and validate the proposed ML-based approach by real-time experiments and demonstrate that there exist distinct patterns between the energy distributions between one and many Wi-Fi AP transmissions. The proposed ML-based approach results in a higher accuracy (close to 99% in all cases) as compared to the existing auto-correlation (AC) and energy detection (ED) approaches.

Introduction

As mobile data demand continues to skyrocket, cellular providers are increasingly looking toward the unlicensed 5 GHz spectrum—traditionally the domain of Wi-Fi—to augment their capacity. This move introduces a significant challenge: how to ensure that LTE-Unlicensed (LTE-U) can coexist fairly with existing Wi-Fi networks without causing prohibitive interference.

Heterogeneous network environment featuring LTE-U and Wi-Fi coexistence across various applications like Smart Cities and IoT.

The Coexistence Challenge

LTE-U coexists with Wi-Fi by using a "duty cycle" approach: the base station transmits for a set period (ON) and then remains silent (OFF) to allow Wi-Fi users to access the channel. To be a "fair neighbor," an LTE-U base station must dynamically adjust its duty cycle based on the density of nearby Wi-Fi networks (Basic Service Sets or BSSs).

However, identifying the exact number of active Wi-Fi networks in real-time is technically difficult. Decoding Wi-Fi packets requires expensive, specialized hardware (a full Wi-Fi receiver) integrated into the LTE base station. Simpler traditional methods, such as Energy Detection (ED) or Autocorrelation (AC), often fail to provide the necessary accuracy in dense, noisy environments.

A Machine Learning Approach to Energy Fingerprinting

Researchers at the University of Chicago have developed a novel solution that treats network detection as a time-series classification task. Instead of decoding packets, the system observes raw energy values on the channel specifically during the LTE-U "OFF" duration.

The LTE-U duty cycle approach, showing ON periods for LTE-U transmission and OFF periods where Wi-Fi and other LAA transmissions can occur.

The core insight is that the distribution of energy over time contains a distinct "fingerprint" for the number of active transmitters. Even without reading the data within the packets, the timing and overlap of energy bursts allow a trained ML model to distinguish between one, two, or five active networks.

Methodology and Experimental Setup

To validate this approach, the researchers built a custom experimental testbed using National Instruments (NI) USRP RIO boards to emulate LTE-U and Wi-Fi hardware. They collected 3.4 GB of real-world data covering scenarios with 0 to 5 active Wi-Fi access points (APs) across varying distances and channel conditions, including both Line-of-Sight (LOS) and Non-Line-of-Sight (NLOS) environments.

The team evaluated several neural network architectures, including Fully Connected (FC) networks and modified VGG models. Ultimately, the Fully Convolutional Network (FCN) proved to be the most robust, as it excels at finding patterns in raw time-series data without requiring manual feature engineering.

Key Results

  • Superior Accuracy: The proposed ML approach achieved approximately 99% accuracy in identifying the number of Wi-Fi APs across nearly all tested scenarios.
  • Performance in Density: In dense scenarios (3+ APs), the accuracy of traditional Energy Detection dropped as low as 56%, while the ML model maintained its 99% performance.
  • Reduced Latency: The ML-based system could detect and adapt the duty cycle in 3.1 seconds, significantly faster than Energy Detection (5.9s) or Autocorrelation (4.8s).
  • Computational Efficiency: By implementing FFT-based convolution with compression, the researchers reduced the computational load by 60% while maintaining high accuracy.

Conclusion and Future Directions

This research demonstrates that deep learning can provide a high-accuracy, low-latency solution for spectrum sharing in complex wireless environments. By leveraging energy patterns during silent periods, LTE-U can effectively manage its duty cycle for fair coexistence without the need for complex packet decoding hardware. Future work will explore distinguishing between different types of Wi-Fi traffic (e.g., video streaming vs. voice) and extending these techniques to the newly opened 6 GHz bands.

Summary updated

Abstract

The application of Machine Learning (ML) techniques to complex engineering problems has proved to be an attractive and efficient solution. ML has been successfully applied to several practical tasks like image recognition, automating industrial operations, etc. The promise of ML techniques in solving non-linear problems influenced this work which aims to apply known ML techniques and develop new ones for wireless spectrum sharing between Wi-Fi and LTE in the unlicensed spectrum. In this work, we focus on the LTE-Unlicensed (LTE-U) specification developed by the LTE-U Forum, which uses the duty-cycle approach for fair coexistence. The specification suggests reducing the duty cycle at the LTE-U base-station (BS) when the number of co-channel Wi-Fi basic service sets (BSSs) increases from one to two or more. However, without decoding the Wi-Fi packets, detecting the number of Wi-Fi BSSs operating on the channel in real-time is a challenging problem. In this work, we demonstrate a novel ML-based approach which solves this problem by using energy values observed during the LTE-U OFF duration. It is relatively straightforward to observe only the energy values during the LTE-U BS OFF time compared to decoding the entire Wi-Fi packet, which would require a full Wi-Fi receiver at the LTE-U base-station. We implement and validate the proposed ML-based approach by real-time experiments and demonstrate that there exist distinct patterns between the energy distributions between one and many Wi-Fi AP transmissions. The proposed ML-based approach results in a higher accuracy (close to 99\% in all cases) as compared to the existing auto-correlation (AC) and energy detection (ED) approaches.
BibTeX
@article{dziedzic2020machine,
  author = {Adam Dziedzic and Vanlin Sathya and Muhammad Iqbal Rochman and Monisha Ghosh and Sanjay Krishnan},
  title = {{Machine Learning enabled Spectrum Sharing in Dense LTE-U/Wi-Fi Coexistence Scenarios}},
  journal = {IEEE Open Journal of Vehicular Technology},
  year = {2020},
  doi = {10.1109/ojvt.2020.2981519},
}
Figure