Joint Activity and Data Detection for Massive Grant-Free Access Using Deterministic Non-Orthogonal Signatures

Nam Yul Yu, Wei Yu

Abstract

As the Internet of Things (IoT) expands toward massive Machine-Type Communications (mMTC), traditional grant-based access protocols become a significant bottleneck due to high signaling overhead and latency. This paper proposes a framework for grant-free access using deterministic, non-orthogonal signature sequences. By leveraging structured algebraic sequences—such as Cubic, Power-Residue, Sidelnikov, and Trace sequences—the authors demonstrate a method for joint activity and data detection that outperforms traditional random signatures. The proposed approach achieves coherence levels near the theoretical Welch bound, significantly reducing the hardware requirements for base stations in massive MIMO systems.

Introduction

The future of the Internet of Things (IoT) hinges on the ability of base stations to manage thousands of devices simultaneously. In many mMTC scenarios, devices transmit small packets sporadically. Traditional "grant-based" protocols, which require a multi-step handshake to request and grant channel access, are inefficient for these short bursts of data, often consuming more energy and bandwidth in signaling than in actual data transmission.

"Grant-free" access offers a solution by allowing devices to transmit immediately using unique signatures. However, this introduces a complex signal processing challenge: the base station must perform joint activity detection (identifying which devices are active) and data detection (decoding the transmitted bits) from a single, overlapping received signal. While random signatures are common in literature, they present practical hurdles in hardware implementation and lack guaranteed performance bounds. This research addresses these gaps by introducing deterministic signature designs.

The Problem: Scaling Grant-Free Access

In a typical mMTC environment, only a small fraction of the total device population is active at any given time. The base station's task is to solve an underdetermined system where the number of potential users ($N$) is much larger than the length of the signature sequence ($L$).

Most existing research relies on random Gaussian or QPSK sequences. While mathematically convenient for analysis, these sequences require significant storage at both the device and the base station. Furthermore, random sequences do not guarantee low "coherence"—a measure of how similar signatures are to one another. High coherence leads to increased detection errors, especially as the number of active devices increases.

Methodology: Deterministic Signature Design

The researchers propose a framework to design deterministic signatures by "masking" standard Discrete Fourier Transform (DFT) matrices. This approach ensures that signatures can be generated on-the-fly from a small algebraic seed, eliminating the need for large look-up tables.

1. Algebraic Masking Sequences

The study evaluates four specific types of algebraic sequences to create these masks:

  • Cubic sequences
  • Power-Residue (PR) sequences
  • Sidelnikov sequences
  • Trace sequences

2. Theoretical Bounds and Coherence

A central contribution of this work is the derivation of a sufficient condition for successful Maximum Likelihood (ML) detection. The authors prove that the reliability of detection is tied to the coherence of the signature matrix. Their proposed deterministic designs achieve a coherence of $O(1/\sqrt{L})$, which closely approaches the Welch bound—the fundamental physical limit for how distinct a set of non-orthogonal sequences can be.

3. Detection Algorithms

To validate the signatures, the authors utilized two primary algorithms:

  • Coordinate Descent ML (CD-ML): A high-performance algorithm that iteratively optimizes the likelihood function.
  • Approximate Message Passing (AMP): A more computationally efficient algorithm that is particularly effective in large-scale systems but highly sensitive to the structural properties of the signatures.

Key Insights and Results

The Power of Structure

Deterministic sequences proved mathematically superior to random ones. Because they are generated via algebraic rules, they provide consistent performance guarantees that random sequences cannot. This structure also allows the base station to detect more active users ($K$) than the length of the signature ($L$), provided the system utilizes a large number of antennas (Massive MIMO).

Antenna and Energy Efficiency

In simulations involving 200 total devices with 40 active users, the deterministic signatures reached a target error rate of $10^{-4}$ using 40 fewer base station antennas than random Gaussian signatures. This reduction in hardware requirements translates directly to lower deployment costs and improved energy efficiency for 5G and 6G infrastructure.

Non-Coherent Advantage

The design supports non-coherent access, where data bits are embedded directly into the choice of the signature sequence. This removes the need for a separate channel estimation phase, further reducing the latency and overhead for IoT devices.

Discussion and Future Work

While the proposed deterministic signatures offer clear advantages in reliability and implementation, some challenges remain. The theoretical proofs for ML success rely on asymptotic assumptions regarding the number of antennas; while simulations show strong performance with 32 to 256 antennas, the benefits may diminish in very small-scale systems.

Additionally, while the signatures themselves are easy to generate, the CD-ML detection algorithm is computationally intensive. Future research may focus on optimizing AMP-based detectors to better leverage the specific algebraic properties of these deterministic masks. Finally, the current model assumes relatively static devices. Investigating the impact of high mobility and the resulting Doppler shifts on the algebraic integrity of the signatures is a vital next step for real-world deployment.

Summary updated

BibTeX
@article{yu2024joint,
  author = {Nam Yul Yu and Wei Yu},
  title = {{Joint Activity and Data Detection for Massive Grant-Free Access Using Deterministic Non-Orthogonal Signatures}},
  journal = {IEEE Trans. Wireless Communications},
  year = {2024},
  doi = {10.1109/twc.2024.3362933},
}
Figure