This document is a Guest Editorial for the second part of a Special Issue on Next Generation Multiple Access (NGMA), published in the IEEE Journal on Selected Areas in Communications. It provides a high-level overview of 20 technical papers that define the roadmap for how devices will connect in the post-5G (6G) era.
TL;DR
As 5G reaches maturity, the focus is shifting toward 6G networks that require a paradigm shift from "communication-only" to multi-functional designs. This editorial summarizes 20 pioneering papers exploring how Next Generation Multiple Access (NGMA) will leverage Reconfigurable Intelligent Surfaces (RIS), Machine Learning, and non-orthogonal schemes to support massive connectivity and ultra-low latency.
What Problem Does This Paper Address?
The transition from 5G to 6G introduces a massive scalability challenge. Current wireless networks primarily rely on Orthogonal Multiple Access (OMA), where users are separated by frequency, time, or code. However, OMA cannot keep up with the "massive and ubiquitous connectivity" required for disruptive applications like the Metaverse (AR/VR), Industry 4.0, and integrated sensing.
The core problem is designing NGMA techniques that are:
- Efficient: Handling a much larger number of users over limited radio resources.
- Multi-functional: Moving beyond just sending data to include sensing, imaging, computing, and localization.
- Intelligent: Using data-driven approaches to manage complex, dynamic network environments.
Methods
As an editorial, this text categorizes the methodologies of 20 different research papers into four primary pillars:
- Information-Theoretic Studies: Using mathematical proofs to determine the fundamental limits of new technologies, such as the interference-nulling capabilities of Reconfigurable Intelligent Surfaces (RIS) and secrecy rates in wiretap channels.
- Massive Multiple Access: Developing algorithms for "grant-free" access (where devices transmit without waiting for a handshake) and "unsourced" random access, which is critical for the Internet of Things (IoT).
- Machine Learning (ML) for NGMA: Applying Federated Meta-Learning (FML) and Multi-Agent Reinforcement Learning (MARL) to optimize device selection, power control, and medium access protocols.
- NGMA "Meets Others": Integrating multiple access with other emerging fields, such as Visible Light Communication (VLC), cache-assisted networks, and Unmanned Aerial Vehicle (UAV) base stations.
Datasets
While the editorial does not introduce a single unified benchmark, the summarized papers utilize a variety of evaluation methods:
- Simulated Environments: Most papers use Monte-Carlo simulations to validate theoretical models (e.g., Rayleigh block-fading channels for unsourced access).
- Scenario-Specific Data: Paper [A17] specifically mentions using both synthetic and real-life datasets to test task management among a large number of workers with scattered computing power.
- Standardized Models: Several papers use standard traffic models, such as Poisson and VoIP traffic, to compare new MAC protocols against existing baselines.
Key Insights and Ideas
- The RIS Advantage: Research shows that RIS can theoretically nullify interference completely in a $K$-user channel if the number of RIS elements is large enough, providing a "clean" channel even in crowded environments.
- From Bits to "Age": There is a shift in metrics. Instead of just looking at throughput, researchers are focusing on the Age-of-Information (AoI)—a measure of how "fresh" the data is—which is vital for autonomous systems.
- Unsourced Access is the Future of IoT: In massive IoT, the base station doesn't need to know who is sending (identity) before decoding what is being sent. This "unsourced" approach significantly reduces signaling overhead.
- Rainbow-Link: A novel frequency-domain scheme termed "rainbow-link" is proposed to serve narrowband users by exploiting wideband spectrum at millimeter-wave bands without the need for explicit beam-training.
Key Results
The editorial highlights several quantitative and qualitative breakthroughs:
- Interference Nulling: It was numerically shown that interference can be nulled if the number of RIS elements is slightly larger than $2K(K-1)$.
- Learning Efficiency: A new Federated Meta-Learning algorithm (NUFM) was shown to accelerate convergence in edge learning by optimizing device selection.
- Latency Reduction: Cache-assisted NOMA schemes demonstrated superior performance in terms of both latency and "cache hit ratios" compared to traditional benchmark schemes.
- Secrecy: New iterative algorithms for multiple-access wiretap channels were able to enhance secrecy rates while simultaneously increasing spectral efficiency.
Limitations and Future Work
The editors acknowledge that NGMA is still in its infancy stage. Key areas for future exploration include:
- Hardware Constraints: For example, in Visible Light Communication (VLC), the limited modulation bandwidth of LEDs remains a bottleneck that requires more advanced rate-splitting techniques.
- Imperfect Information: Many algorithms (like GAMP-based detection) perform well with perfect channel knowledge but need more robustness for "imperfect" real-world channel distribution information.
- Complexity vs. Performance: While non-orthogonal schemes (NOMA) offer higher efficiency, they often require more complex receivers. Future work must focus on low-complexity iterative algorithms that can run on power-constrained IoT devices.
Summary updated