User Clustering and Power Allocation for Hybrid Non-Orthogonal Multiple Access Systems

Kaidi Wang, Wei Liang, Yi Yuan, Yuanwei Liu, Zheng Ma, Zhiguo DingIEEE Trans. Vehicular Technology 2019 · DOI

Abstract

As next-generation wireless networks (5G/6G) demand higher spectral efficiency, Non-Orthogonal Multiple Access (NOMA) has emerged as a critical technology. However, the computational complexity of Successive Interference Cancellation (SIC) limits the practical size of NOMA user groups. This paper proposes a hybrid NOMA-TDMA framework that optimizes user clustering and power allocation. By modeling clustering as a coalitional game and utilizing Successive Convex Approximation (SCA) for power distribution, the proposed system achieves high throughput while maintaining manageable receiver complexity. Simulation results demonstrate that the proposed heuristic algorithms achieve near-optimal performance with significantly lower computational overhead compared to exhaustive search methods.

Introduction

Traditional wireless access methods, such as Time Division Multiple Access (TDMA), rely on orthogonal resources to prevent interference. While simple, these methods are spectrally inefficient. Non-Orthogonal Multiple Access (NOMA) allows multiple users to share the same resource simultaneously by utilizing the power domain. While NOMA offers superior spectral efficiency, it introduces a significant challenge: as the number of users in a NOMA group increases, the complexity of Successive Interference Cancellation (SIC) at the receiver grows exponentially.

To address this, the research explores a Hybrid NOMA system. In this architecture, users are dynamically grouped into clusters. Within each cluster, NOMA is employed; between clusters, the system utilizes TDMA. This approach seeks to balance the high capacity of NOMA with the practical implementation constraints of modern hardware.

Methodology

The authors decouple the joint optimization of user clustering and power allocation into two distinct sub-problems.

1. User Clustering via Coalitional Game Theory

The clustering problem is modeled as a coalitional game to ensure a stable and efficient distribution of users. The study evaluates two primary time-slot allocation strategies:

  • User-based Allocation: Time slots are allocated proportionally to the number of users in a cluster, prioritizing fairness.
  • Cluster-based Allocation: Each cluster receives an equal time slot regardless of size, prioritizing total system throughput.

Two algorithms were developed to solve the clustering game:

  • Preference Relation Based Algorithm (PRA): A low-complexity heuristic that allows users to switch clusters based on a preference relation. This ensures the system reaches a "Nash-stable" state where no user can improve the system payoff by moving.
  • Simulated Annealing Algorithm (SAA): A probabilistic method used to approximate the global optimum, serving as a benchmark for the PRA.

2. Power Allocation via Successive Convex Approximation (SCA)

Once clusters are formed, the system must determine the optimal power levels for each user to maximize the weighted sum rate. Because this optimization problem is non-convex, the authors employ Successive Convex Approximation (SCA). This iterative technique approximates the non-convex objective with a series of convex sub-problems, converging to a Karush-Kuhn-Tucker (KKT) point.

Key Insights and Results

The Complexity-Fairness Trade-off

A central contribution of this work is the introduction of a weight factor ($\omega$) that penalizes the complexity of SIC. This prevents the formation of excessively large clusters that would be mathematically optimal but hardware-prohibitive.

Performance Benchmarks

Simulation results, conducted in a downlink network with a 200m radius and standard path loss models, revealed several key findings:

  • Efficiency of Heuristics: The low-complexity PRA algorithm achieved approximately 95% of the performance of the SAA benchmark while converging significantly faster (typically within 1 to 3.3 cycles).
  • Throughput vs. Fairness: The cluster-based allocation scheme maximized the total weighted sum rate by forming fewer, larger clusters. Conversely, the user-based scheme achieved a higher Jain’s fairness index, ensuring that users at the cell edge maintained reliable connectivity.
  • Superiority over Baselines: The hybrid NOMA approach consistently outperformed both pure TDMA and centralized NOMA (where all users are in a single cluster), validating the hybrid architecture as a superior middle ground.

Conclusion and Future Work

This research provides a robust framework for implementing NOMA in practical wireless environments. By combining game-theoretic clustering with SCA-based power allocation, the authors demonstrate that hybrid NOMA can deliver high spectral efficiency without overwhelming receiver hardware.

Future research directions include:

  • Imperfect CSI: Evaluating the impact of imperfect Channel State Information on SIC performance.
  • MIMO Integration: Extending the framework to Multi-Input Multi-Output (MIMO) systems.
  • Dynamic Bandwidth: Exploring hybrid NOMA-FDMA to allow for dynamic frequency allocation alongside time-slot scheduling.

Summary updated

BibTeX
@article{wang2019user,
  author = {Kaidi Wang and Wei Liang and Yi Yuan and Yuanwei Liu and Zheng Ma and Zhiguo Ding},
  title = {{User Clustering and Power Allocation for Hybrid Non-Orthogonal Multiple Access Systems}},
  journal = {IEEE Trans. Vehicular Technology},
  year = {2019},
  doi = {10.1109/tvt.2019.2948105},
}
Figure