Unsourced Multiple Access With Common Alarm Messages: Network Slicing for Massive and Critical IoT

Khac-Hoang Ngo, Giuseppe Durisi, Alexandre Graell i Amat, Petar Popovski, Anders E. Kalør, Beatriz Soret

TL;DR

This paper investigates how to integrate high-priority "common alarm" messages into the Unsourced Multiple Access (UMA) framework, which is designed for massive IoT. Through a rigorous information-theoretic analysis, the authors demonstrate that while orthogonal network slicing (H-OMA) allows these two traffic types to coexist with high energy efficiency, nonorthogonal slicing (H-NOMA) fails due to the difficulty of perfectly canceling interference from common alarm signals.

What Problem Does This Paper Address?

The Internet of Things (IoT) is generally split into two camps: massive IoT (millions of low-cost sensors sending small data bursts) and critical IoT (high-reliability, low-latency applications like autonomous vehicles). Traditionally, these are handled by separate protocols (mMTC and URLLC in 5G).

The authors tackle the challenge of heterogeneous traffic within the Unsourced Multiple Access (UMA) framework. In UMA, all devices share a single codebook, and the receiver just wants a list of what was sent, not who sent it. The specific gap addressed here is the "common alarm" scenario: when a physical event (like a gas leak) occurs, multiple sensors might simultaneously transmit the same high-priority alarm message, alongside their own unique standard data. The goal is to find the most energy-efficient way to slice the network so that both types of messages are received reliably when the number of active users is random and unknown.

Methods

The researchers developed a mathematical model for message generation where each user can independently generate a standard message, an alarm message, or both (see the tree representation in Figure 1).

![Fig.

  1. A tree representation of the message generation for user k.](/figures/art_f0341d54272c42dfb1c6dbcbfb3fac81)

They compared two primary network slicing strategies:

  1. Heterogeneous Orthogonal Multiple Access (H-OMA): The transmission frame is split into two distinct blocks—one for alarms and one for standard data.
  2. Heterogeneous Nonorthogonal Multiple Access (H-NOMA): Alarm and standard messages are superposed and transmitted over the entire frame simultaneously.

To evaluate these, they derived new random-coding achievability bounds for the Gaussian Multiple Access Channel. These bounds provide a theoretical "speed limit" for the minimum energy per bit ($E_b/N_0$) required to meet specific targets for misdetections and false positives.

![Fig.

  1. Illustration of a frame in the two considered network slicing strategies, namely, H-OMA and H-NOMA.](/figures/art_56d746fe3e74475fb41300f4e13b9ba1)

Datasets

Not discussed in the paper. As a theoretical information-theory study, the authors rely on synthetic Gaussian noise models and mathematical derivations rather than empirical datasets.

Key Insights and Ideas

  • The Power of Coherence: In the alarm block, because multiple users transmit the exact same codeword, their signals add up coherently. This creates a massive boost in effective SNR, allowing the alarm to be decoded with extremely high reliability using very few channel uses.
  • The H-NOMA Trap: While nonorthogonal access is often touted for its spectral efficiency, this paper finds it energy-inefficient for this specific IoT mix. The "kicker" is interference cancellation. To decode standard messages in H-NOMA, the receiver must first subtract the alarm signal. However, even if the alarm message is decoded correctly, the receiver must also perfectly estimate the number of users who sent it. Any error in this estimation leaves residual interference that drowns out the weaker standard messages.
  • Dynamic Range Matters: For H-OMA to be efficient, devices must be able to transmit alarm messages at a much lower power than standard messages. If the hardware has a limited "dynamic range" (the ratio between the highest and lowest power it can support), the energy efficiency of the alarm traffic drops significantly.

Key Results

  • H-OMA Efficiency: With a marginal "backoff" (a slight increase in energy) for standard traffic, H-OMA can support critical alarms at very high energy efficiency. For example, if many users detect an alarm, the required $E_b/N_0$ for the alarm can be as low as -30 dB.
  • H-NOMA Failure: H-NOMA requires significantly higher energy for the alarm traffic to satisfy the reliability requirements of the standard traffic. The authors conclude that "reliability diversity"—the idea of using different reliability targets to separate superposed signals—is hard to exploit when the number of active users is unknown.
  • User Sensitivity: The more sensitive the users are to the alarm (i.e., the more users transmit the common message), the more efficient H-OMA becomes due to the coherent signal gain.

Limitations and Future Work

The authors acknowledge several simplifications that provide avenues for future research:

  • Gaussian Channel Assumption: The study assumes a perfect Gaussian channel. Real-world IoT channels involve fading, phase rotations, and synchronization errors, which would likely make the coherent addition of alarm messages much harder to achieve.
  • Single Alarm Type: The model only considers one common alarm message. Future work could explore scenarios with multiple different alarm types requiring different reliability levels.
  • Power Adaptation: The paper suggests looking into dynamic power adaptation, where users might boost their standard message power only when no alarm is present.

Summary updated

Abstract

We investigate the coexistence of massive and critical Internet of Things (IoT) services in the context of the unsourced multiple access (UMA) framework introduced by Polyanskiy (2017), where all users employ a common codebook and the receiver returns an unordered list of decoded codewords. This setup is suitably modified to introduce heterogeneous traffic. Specifically, to model the massive IoT service, we assume that a standard message originates independently from each IoT device as in the standard UMA setup. To model the critical IoT service, we assume the generation of alarm messages that are common for all devices. This setup requires a significant redefinition of the error events, i.e., misdetections and false positives. We further assume that the number of active users in each transmission attempt is random and unknown. We derive a random-coding achievability bound on the misdetection and false positive probabilities of both standard and alarm messages on the Gaussian multiple access channel. Using our bound, we demonstrate that orthogonal network slicing enables massive and critical IoT to coexist under the requirement of high energy efficiency. On the contrary, we show that nonorthogonal network slicing is energy inefficient due to the residual interference from the alarm signal when decoding the standard messages.
BibTeX
@article{ngo2024unsourced,
  author = {Khac-Hoang Ngo and Giuseppe Durisi and Alexandre Graell i Amat and Petar Popovski and Anders E. Kalør and Beatriz Soret},
  title = {{Unsourced Multiple Access With Common Alarm Messages: Network Slicing for Massive and Critical IoT}},
  journal = {IEEE Trans. Communications},
  year = {2024},
  doi = {10.1109/tcomm.2023.3325470},
}
Figure