Proactive Resource Management for LTE in Unlicensed Spectrum: A Deep Learning Perspective

Ursula Challita, Li Dong, Walid SaadIEEE Trans. Wireless Communications 2018 · OpenReview

TL;DR

This paper proposes a novel deep learning approach using LSTMs for proactive resource management in LTE-LAA networks operating in unlicensed spectrum. By predicting future traffic demands and channel availability, the system can dynamically allocate resources to maximize LTE-LAA throughput while ensuring fair coexistence with existing WiFi networks, outperforming reactive approaches.

What Problem Does This Paper Address?

The increasing demand for wireless data has led to spectrum scarcity. LTE-LAA offers a solution by allowing LTE to utilize unlicensed spectrum bands, promising improved performance. However, this introduces a critical challenge: ensuring fair coexistence with existing WiFi deployments. Without proper management, LTE-LAA can significantly degrade WiFi performance. Existing solutions often rely on reactive "sense-and-avoid" mechanisms, which can lead to underutilization of the spectrum and suboptimal performance. This paper addresses the need for a proactive approach that can intelligently manage resources, predict future network conditions, and ensure fairness between LTE-LAA and WiFi.

Methods

The authors formulate the resource allocation problem as a noncooperative game among LTE-LAA small base stations (SBSs). These SBSs are modeled as "Homo Egualis" agents, meaning they are inequity-averse and consider fairness in their decision-making. To solve this game, they propose a novel deep learning algorithm based on Long Short-Term Memory (LSTM) networks. LSTMs are chosen for their ability to learn long-term dependencies in sequential data, which is crucial for predicting future traffic patterns and channel availability. The approach involves:

  • Proactive Game Formulation: SBSs aim to maximize their throughput over a prediction window (T) while adhering to long-term fairness constraints with WiFi and other LTE-LAA operators.
  • RL-LSTM Algorithm: An end-to-end reinforcement learning framework using LSTMs is developed. This allows SBSs to learn a sequence of actions (channel selection and access probability) over the prediction window without explicit knowledge of future network states.
  • Encoder-Decoder Architecture: A traffic encoder learns a representation of historical traffic data, and an action decoder uses this representation to predict the optimal action sequence.
  • Mixed-Strategy Nash Equilibrium (NE): The algorithm aims to find a mixed-strategy NE, where players probabilistically choose their actions, reflecting the dynamic and uncertain nature of the wireless environment.
  • Incremental Penalty Algorithm: This method is used to handle the coupled constraints of the game, ensuring feasibility.

Datasets

The paper utilizes real-world traffic data traces for training and evaluation. The specific dataset is referenced as being from [44]. The data was divided into 80% for training and 20% for testing.

Key Insights and Ideas

  • Proactive vs. Reactive: Proactive resource management, by predicting future conditions, offers significant gains over reactive approaches that only respond to current states. This is especially true when traffic is not uniform.
  • LSTM's Power for Temporal Dependencies: LSTMs are well-suited for this problem due to their ability to capture long-term temporal dependencies in traffic patterns, which is essential for effective prediction and proactive decision-making.
  • Homo Egualis Agents for Fairness: Modeling SBSs as inequity-averse agents naturally incorporates fairness into the resource allocation, ensuring a better tradeoff between efficiency and coexistence.
  • Time Window (T) Optimization: The length of the prediction window (T) is a critical parameter. While larger T generally offers more opportunities for proactive gains, there's an optimal point beyond which gains diminish, suggesting a balance between prediction horizon and computational complexity.
  • Carrier Aggregation Benefits: Enabling carrier aggregation can improve spectrum access time, but its benefit diminishes with a higher density of SBSs.

Key Results

  • Throughput Gains: The proposed proactive scheme achieves up to 28% throughput gain compared to a conventional reactive approach and 11% gain over a proportional fair coexistence mechanism.
  • WiFi Performance Preservation: The framework effectively prevents performance degradation for WiFi networks, even when a large number of LTE operators deploy LTE-LAA.
  • Fairness: The scheme ensures long-term equal weighted airtime fairness between LTE-LAA and WiFi, as demonstrated by Jain's fairness index.
  • Optimal Time Window: For uniform traffic distributions, an optimal time window of T=1 is identified. However, for fluctuating traffic, larger T values yield greater benefits.
  • Comparison to Centralized Schemes: The proposed distributed scheme offers a better tradeoff between efficiency and fairness compared to centralized proportional fairness (PF) and total network throughput (TNT) maximization schemes.

Limitations and Future Work

  • Hyperparameter Sensitivity: The convergence of the proposed algorithm is highly dependent on hyperparameters, particularly the learning rate and hidden layer size of the LSTMs. Experimental studies are needed to find optimal problem-specific learning rates.
  • Synchronization Challenges: A practical challenge for real-world deployment is the synchronization between SBSs and WAPs, which may require inter-operator cooperation.
  • Future Work: The authors suggest extending the work to multi-mode SBSs that operate across licensed, unlicensed, and mmWave bands, addressing coexistence with WiFi and WiGig simultaneously.

Summary updated

Abstract

Performing cellular long term evolution (LTE) communications in unlicensed spectrum using licensed assisted access LTE (LTE-LAA) is a promising approach to overcome wireless spectrum scarcity. However, to reap the benefits of LTE-LAA, a fair coexistence mechanism with other incumbent WiFi deployments is required. In this paper, a novel deep learning approach is proposed for modeling the resource allocation problem of LTE-LAA small base stations (SBSs). The proposed approach enables multiple SBSs to proactively perform dynamic channel selection, carrier aggregation, and fractional spectrum access while guaranteeing fairness with existing WiFi networks and other LTE-LAA operators. Adopting a proactive coexistence mechanism enables future delay-tolerant LTE-LAA data demands to be served within a given prediction window ahead of their actual arrival time thus avoiding the underutilization of the unlicensed spectrum during off-peak hours while maximizing the total served LTE-LAA traffic load. To this end, a noncooperative game model is formulated in which SBSs are modeled as homo egualis agents that aim at predicting a sequence of future actions and thus achieving long-term equal weighted fairness with wireless local area network and other LTE-LAA operators over a given time horizon. The proposed deep learning algorithm is then shown to reach a mixed-strategy Nash equilibrium, when it converges. Simulation results using real data traces show that the proposed scheme can yield up to 28% and 11% gains over a conventional reactive approach and a proportional fair coexistence mechanism, respectively. The results also show that the proposed framework prevents WiFi performance degradation for a densely deployed LTE-LAA network.
BibTeX
@article{challita2018proactive,
  author = {Ursula Challita and Li Dong and Walid Saad},
  title = {{Proactive Resource Management for LTE in Unlicensed Spectrum: A Deep Learning Perspective}},
  journal = {IEEE Trans. Wireless Communications},
  year = {2018},
  url = {https://openreview.net/forum?id=GzRG36XYyYD},
}
Figure