Load Balancing for 5G Ultra-Dense Networks Using Device-to-Device Communications

Hongliang Zhang, Lingyang Song, Ying Jun ZhangIEEE Trans. Wireless Communications 2018 · DOI

TL;DR

This paper tackles the challenge of uneven data traffic distribution in 5G ultra-dense small cell networks. They propose an iterative algorithm that leverages Device-to-Device (D2D) communications to offload traffic from congested cells to underutilized ones, thereby improving overall system sum-rate.

What Problem Does This Paper Address?

5G networks are moving towards ultra-dense deployments of small cells to meet the ever-increasing demand for mobile data. However, this density leads to significant spatial and temporal fluctuations in traffic load across these small cells. Traditional load balancing techniques, which rely on borrowing resources from neighboring cells, are not suitable for 5G because these cells will share the same spectrum band, leading to interference. This paper addresses the problem of efficiently balancing this traffic load in such an environment without requiring additional spectrum.

Methods

The authors propose an iterative resource allocation and D2D routing (I-RA-DR) algorithm. The core idea is to decouple the complex optimization problem into two subproblems:

  • Resource Allocation: This subproblem is reformulated as a Monotonic Optimization (MO) problem and solved using an accelerated polyblock outer approximation method. This approach aims to find the optimal allocation of radio resources (like subchannels and power) to links.
  • D2D Routing: This subproblem involves two steps: SBS (Small Base Station) selection and routing. It's transformed into a Complementary Geometric Programming (CGP) problem, which is then solved using an iterative convex relaxation approach. This part determines how data flows are routed through D2D links to underutilized cells.

The algorithm iteratively optimizes these two subproblems until a stopping criterion is met.

Datasets

Not discussed in the paper. The paper focuses on theoretical modeling and simulation-based evaluation.

Key Insights and Ideas

  • D2D for Load Balancing in 5G: The paper demonstrates that D2D communications are a viable and effective mechanism for load balancing in 5G ultra-dense small cell networks, especially given the spectrum sharing constraints.
  • Decoupling for Tractability: Breaking down the complex joint optimization problem into resource allocation and D2D routing subproblems significantly simplifies the solution process.
  • Monotonic Optimization and CGP: The successful application of MO for resource allocation and CGP for D2D routing highlights the power of these optimization frameworks for wireless network problems.
  • Iterative Improvement: The iterative nature of the I-RA-DR algorithm allows for gradual improvement of the system sum-rate by refining both resource allocation and routing decisions.

Key Results

  • Average Sum-Rate Increase: Simulation results show that the proposed I-RA-DR algorithm achieves an average increase of 20% in the data sum-rate in neighboring small cells by offloading traffic from congested cells.
  • Performance Improvement: The I-RA-DR algorithm consistently outperforms schemes without load balancing and a greedy routing algorithm in terms of system sum-rate. For instance, in one scenario, the system sum-rate increased by 12% compared to the scheme without load balancing.
  • Improved Data Rate for Edge Users: The D2D load balancing effectively improves the data rate for cell-edge users by allowing their traffic to be offloaded.
  • Convergence: The algorithm is shown to converge, with around 10 iterations typically needed for convergence, and the number of iterations does not significantly increase with the number of users.

Limitations and Future Work

  • Computational Complexity: While the proposed algorithm is efficient, the paper acknowledges that the computational complexity of MO is still an open problem. The accelerated method helps, but for very large networks, complexity might remain a concern.
  • Sub-optimal Solutions: The accelerated method for resource allocation is noted to provide sub-optimal solutions, trading off optimality for faster convergence.
  • Future Work: The paper suggests exploring more advanced D2D routing strategies and investigating the impact of different D2D communication models (e.g., multi-hop D2D) on load balancing performance. Further research could also focus on dynamic traffic scenarios and real-time adaptation of the load balancing strategy.

Summary updated

BibTeX
@article{zhang2018load,
  author = {Hongliang Zhang and Lingyang Song and Ying Jun Zhang},
  title = {{Load Balancing for 5G Ultra-Dense Networks Using Device-to-Device Communications}},
  journal = {IEEE Trans. Wireless Communications},
  year = {2018},
  doi = {10.1109/twc.2018.2819648},
}
Figure