[Paper Review] Joint In-Band Backhauling and Interference Mitigation in 5G Heterogeneous Networks

Trung Kien Vu, Mehdi Bennis+3 more|arXiv (Cornell University)|Apr 10, 2016

Advanced MIMO Systems Optimization13 references22 citations

TL;DR

This paper proposes a joint in-band backhauling and interference mitigation framework for 5G heterogeneous networks (HetNets) using massive MIMO macro cells and self-backhauled small cells. By leveraging random matrix theory and stochastic optimization, it achieves 1.7 Gbps average user equipment (UE) throughput and 1 Gbps cell-edge throughput at 28 GHz with 200 UEs/km², demonstrating a 62× gain over 2.4 GHz in ultra-dense deployments.

ABSTRACT

In this paper, we study the problem of joint inband backhauling and interference mitigation in 5G heterogeneous networks (HetNets) in which a massive multiple-input multipleoutput (MIMO) macro cell base station equipped with a large number of antennas, overlaid with self-backhauled small cells is assumed. This problem is cast as a network utility maximization subject to wireless backhaul constraints. Due to the non-tractability of the problem, we first resort to random matrix theory to get a closed-form expression of the achievable rate and transmit power in the asymptotic regime, i.e., as the number of antennas and users grows large. Subsequently, leveraging the framework of stochastic optimization, the problem is decoupled into dynamic scheduling of macro cell users and backhaul provisioning of small cells as a function of interference and backhaul links. Via simulations, we evaluate the performance gains of our proposed framework under different network architectures and low/high frequency bands. Our proposed HetNet method achieves the achievable average UE throughput of 1.7 Gbps as well as ensures 1 Gbps cell-edge UE throughput when serving 200 UEs per km2 at 28 GHz with 1 GHz bandwidth. In ultra-dense network, the UE throughput at 28 GHz achieves 62x gain as compared to 2.4 GHz.

Motivation & Objective

• To address the challenge of dynamically optimizing network performance in 5G HetNets with in-band wireless backhaul and interference.
• To jointly schedule macro cell users and provision backhaul capacity to small cells under varying wireless backhaul constraints.
• To maximize network utility while ensuring stability and minimizing interference in massive MIMO-enabled HetNets.
• To evaluate performance gains in ultra-dense mmWave deployments across low- and high-frequency bands.
• To establish a theoretical utility-queue backlog tradeoff using Lyapunov optimization.

Proposed method

• Formulates the problem as a network utility maximization (NUM) with wireless backhaul and stability constraints.
• Applies random matrix theory (RMT) to derive a closed-form deterministic equivalent for the asymptotic SINR in the large-antenna regime.
• Uses the Lyapunov optimization framework to decouple the joint problem into dynamic scheduling and backhaul provisioning subproblems.
• Employs successive convex approximation (SCA) to solve the resulting mixed-integer non-convex optimization problem.
• Derives deterministic equivalents for SINR expressions of both macro and small cell users using Stieltjes transforms and matrix trace approximations.
• Introduces a dynamic operation mode to control interference among adjacent small cells via precoder design.

Experimental results

Research questions

• RQ1 How can in-band backhauling and interference mitigation be jointly optimized in massive MIMO-based 5G HetNets?
• RQ2 What is the achievable spectral efficiency and throughput gain when deploying mmWave bands (e.g., 28 GHz) in ultra-dense HetNets?
• RQ3 How does the Lyapunov parameter ν affect the tradeoff between network utility and queue backlog in dynamic scheduling?
• RQ4 What performance gains are achievable in terms of average and cell-edge user throughput when using 28 GHz vs. 2.4 GHz bands?
• RQ5 How does the number of antennas and user density impact the feasibility and performance of the proposed framework?

Key findings

• At 28 GHz with 1 GHz bandwidth and 200 UEs/km², the proposed HetNet achieves an average UE throughput of 1.7 Gbps and a cell-edge throughput of 1 Gbps.
• In ultra-dense deployments with an inter-site distance (ISD) of 80 m, the UE throughput at 28 GHz achieves a 62× gain over 2.4 GHz due to wider bandwidth and beamforming.
• For ISD of 250 m, the throughput gain at 28 GHz is 56× compared to 2.4 GHz, primarily due to 50× larger bandwidth and directional beamforming.
• When ISD drops below 50 m, network queue size increases significantly, indicating congestion that requires power or admission control adjustments.
• The framework exhibits an [O(1/ν), O(ν)] utility-queue backlog tradeoff, confirming a theoretical utility-delay tradeoff with increasing ν.
• The proposed method achieves 473 Mbps per UE in ultra-dense 28 GHz deployments with ISD of 33 m, while 2.4 GHz throughput drops below 10 Mbps under the same conditions.