Enabling Uncoordinated Dynamic Spectrum Sharing Between LTE and NR Networks

Merkebu Girmay, Vasilis Maglogiannis, Dries Naudts, Timo De Waele, Eli De Poorter, Adnan Shahid, H. Vincent Poor, Ingrid Moerman

This paper, titled "Enabling Uncoordinated Dynamic Spectrum Sharing Between LTE and NR Networks," presents a novel way for 4G and 5G networks to share the same frequency bands without needing a direct communication link between them.

TL;DR

The authors propose a "cross-band" Dynamic Spectrum Sharing (DSS) scheme that allows LTE and 5G New Radio (NR) networks to intelligently borrow each other's unused subframes. By using a CNN-based sensing system instead of a dedicated coordination channel, the approach increases spectrum efficiency by up to 20.7% and throughput by up to 13.5% compared to static allocations.

What Problem Does This Paper Address?

As the world transitions to 5G, mobile operators face a "spectrum crunch." 5G needs the lower frequency bands currently occupied by 4G LTE to provide wide-area coverage. Static "re-farming"—where a chunk of spectrum is permanently taken from 4G and given to 5G—is inefficient because it doesn't account for real-time traffic fluctuations.

Existing Dynamic Spectrum Sharing (DSS) solutions solve this but usually require a coordination signaling channel (like the X2 interface) between the LTE eNodeB and the NR gNodeB. This is a major hurdle for uncoordinated networks, such as those run by different operators or private 5G deployments. The authors tackle the challenge of enabling DSS without this direct coordination, allowing networks to coexist simply by "listening" to the airwaves.

Methods

The proposed solution relies on two main pillars: leveraging existing broadcast features and a machine learning-based sensing engine.

  1. MBSFN and MBS Leveraging: The system uses the Multimedia Broadcast Multicast Service (MBSFN) in LTE and the Multicast Broadcast Service (MBS) in NR. These features allow a network to "mute" certain subframes (normally reserved for broadcast). The authors' scheme uses these muted windows as opportunities for the other network to transmit data.
  2. TRTC (Technology Recognition and Traffic Characterization): Since the networks don't talk to each other, they need to sense when a subframe is available. The authors developed a Convolutional Neural Network (CNN) that analyzes In-phase and Quadrature (IQ) samples. By performing a Fast Fourier Transform (FFT) on these samples, the CNN classifies the signal as LTE, NR, or an "Overlap" (collision).
  3. Adaptive Resource Allocation: Both networks use a resource controller that monitors their own traffic queues. If one network has low demand, it increases the number of muted subframes, which the other network detects via the TRTC and uses to offload its own traffic. This is managed via a binary search algorithm against a pre-computed capacity lookup table.

Datasets

The study utilized synthetic datasets generated through high-fidelity simulation tools:

  • Vienna LTE System-Level Simulator: Used to model the LTE environment.
  • 5G MATLAB Toolbox: Used to model the NR environment.
  • IQ Sample Dataset: The authors generated 7,500 examples per class (LTE, NR, Overlap) across a wide range of Signal-to-Noise Ratios (SNR) from -15 dB to 30 dB to train and validate the CNN.

Key Insights and Ideas

  • The "Silent" Coordinator: The most significant insight is that physical layer features (MBSFN/MBS) can act as a signaling mechanism. By muting subframes, a network "signals" availability to its neighbor without a single bit of backhaul coordination.
  • Avoiding the "Reference Signal" Trap: One of the biggest headaches in DSS is that LTE always transmits Cell-specific Reference Signals (CRS), which interfere with 5G. By using MBSFN subframes, the LTE network completely mutes these symbols, creating a "clean" window for 5G NR.
  • The Batch Size Sweet Spot: The authors found that a batch size of 16 Time Resolution Windows (TRWs) is optimal. It provides enough data for the CNN to be accurate while remaining fast enough to meet the 1ms subframe timing requirements of cellular networks.

Key Results

The proposed uncoordinated DSS scheme was compared against a static band configuration:

  • Throughput Gains: LTE throughput improved by 13.5%, and NR throughput improved by 8.3%.
  • Spectrum Efficiency: Utilization efficiency increased by 11.8% on the LTE band and a staggering 20.7% on the NR band.
  • ML Accuracy: The CNN-based TRTC achieved over 98% accuracy for signal recognition at SNR levels of 5 dB or higher.
  • Reliability: The probability of buffer overflow (data packets being dropped) decreased from 14.6% in standard setups to roughly 9% with the proposed scheme.

Limitations and Future Work

  • Synchronization Requirements: The system still assumes subframe-level synchronization (likely via GPS/PPS), which is common but necessary for the "muted" windows to align perfectly.
  • Reactive vs. Proactive: The current sensing is somewhat reactive. The authors suggest that future work could integrate proactive traffic prediction (using LSTMs or Transformers) to anticipate traffic bursts and adjust the spectrum sharing before the queues fill up.
  • Monitoring Overhead: There is a slight "monitoring phase" penalty where the system must sense the environment, leading to a marginal (<2%) drop in efficiency compared to a perfectly coordinated (wired) system.

Summary updated

Abstract

Dynamic Spectrum Sharing (DSS) is an enabler for a seamless transition from 4G Long Term Evolution (LTE) to 5G New Radio (NR) by utilizing existing LTE bands without static spectrum re-farming. In this paper, we propose a cross-band DSS scheme that utilizes the Multimedia Broadcast Multicast Service over a Single Frequency Network (MBSFN) feature of an LTE network and the Multicast Broadcast Service (MBS) feature of an NR network. The proposed DSS scheme utilizes LTE and NR resource controllers to assign muted MBSFN subframes on the LTE band and muted MBS subframes on the NR band based on traffic needs. In contrast to the state-of-the-art, the proposed DSS scheme does not require a coordination signaling channel between the LTE and NR networks. Instead, a machine learning-based Technology Recognition and Traffic Characterization (TRTC) system is used to identify and characterize traffic patterns. The LTE and NR resource controllers use the TRTC to sense the muted subframes and offload traffic accordingly. On average, the proposed DSS, as compared to static band configuration, improves the LTE throughput, NR throughput, LTE band spectrum utilization efficiency, and NR band spectrum utilization efficiency by 13.5%, 8.3%, 11.8%, and 20.7%, respectively.
BibTeX
@article{girmay2023enabling,
  author = {Merkebu Girmay and Vasilis Maglogiannis and Dries Naudts and Timo De Waele and Eli De Poorter and Adnan Shahid and H. Vincent Poor and Ingrid Moerman},
  title = {{Enabling Uncoordinated Dynamic Spectrum Sharing Between LTE and NR Networks}},
  journal = {IEEE Transactions on Wireless Communications},
  year = {2023},
  doi = {10.1109/twc.2023.3329385},
}
Figure