Multi-Cell Sparse Activity Detection for Massive Random Access: Massive MIMO Versus Cooperative MIMO

Zhilin Chen, Foad Sohrabi, Wei Yu

TL;DR

This paper tackles the challenge of detecting active devices in massive machine-type communications (mMTC), where millions of IoT devices sporadically transmit data. It compares two network architectures—non-cooperative Massive MIMO and Cooperative MIMO (C-RAN)—using the Approximate Message Passing (AMP) algorithm. The authors analytically prove and simulate that while Massive MIMO can brute-force detection errors to zero with enough antennas, Cooperative MIMO is highly efficient for cell-edge users, achieving similar performance with a fraction of the antennas.

What Problem Does This Paper Address?

In future 5G and IoT networks, massive machine-type communications (mMTC) will support tens of millions of devices. Because these devices have sporadic, event-driven traffic, only a tiny fraction are active at any given time. Before a base station (BS) can decode their data, it must first figure out which devices are actually transmitting.

Because the pool of potential devices is so large, we cannot assign unique, orthogonal signature sequences to everyone. Instead, devices must use non-orthogonal sequences, which leads to severe inter-cell interference when multiple devices transmit simultaneously. The paper asks a fundamental architectural question: to mitigate this interference and accurately detect active users, is it better to deploy a massive number of antennas at each independent base station (Massive MIMO), or should base stations share their data to jointly detect users (Cooperative MIMO)?

Methods

The authors frame the user activity detection problem as a sparse recovery problem (since user activity is sparse) and solve it using the Approximate Message Passing (AMP) algorithm with a Bayesian denoiser.

They evaluate two distinct approaches:

  1. Massive MIMO: Each BS operates independently, running AMP to detect only its own users while treating signals from neighboring cells as background Gaussian noise.
  2. Cooperative MIMO: BSs operate in a Cloud-Radio Access Network (C-RAN). Each BS runs AMP to detect users in its own cell and neighboring cells (actively recovering the interference rather than treating it as noise). The BSs compute soft decisions known as Log-Likelihood Ratios (LLRs) and forward them to a Central Unit (CU) for final aggregation and decision-making.

To analyze performance, the authors use state evolution, a mathematical tool that tracks the performance of AMP in the asymptotic regime (where the number of users and sequence lengths approach infinity). Finally, to make the Cooperative MIMO approach practical, they design a user-specific uniform quantization scheme to compress the LLRs before sending them over capacity-limited fronthaul links.

Datasets

Not discussed in the paper. The authors validate their analytical models using Monte Carlo simulations with synthetically generated Rayleigh fading channels and i.i.d. Gaussian signature sequences.

Key Insights and Ideas

  • Interference as Noise vs. Recovering Interference: Treating inter-cell interference as noise (Massive MIMO) works well if you have a massive number of antennas to overpower the interference. However, actively recovering the interference (Cooperative MIMO) significantly lowers the error floor, particularly for vulnerable cell-edge users.
  • The Antenna vs. Cooperation Trade-off: There is a direct trade-off between hardware complexity (antennas) and network complexity (cooperation/fronthaul). The authors found that cooperating across just 3 base stations yields the same cell-edge reliability as quadrupling the number of antennas in a non-cooperative setup.
  • Fronthaul Efficiency is Achievable: You don't need infinite bandwidth to make Cooperative MIMO work. Because the dynamic range of LLRs depends heavily on a specific user's large-scale fading coefficient, a generic quantization scheme fails. However, by using a user-specific uniform quantization scheme, just 3-4 bits per LLR is enough to approach the performance of a perfect, unquantized network.

Key Results

  • Massive MIMO Asymptotics: The state evolution analysis proves that as the number of antennas $M \to \infty$, the probabilities of false alarm and missed detection can be effectively driven to zero, even in the presence of inter-cell interference.
  • Cooperative MIMO Asymptotics: Unlike Massive MIMO, perfect detection cannot be achieved simply by increasing the number of cooperating BSs (due to path loss attenuation). However, it drastically improves cell-edge performance.
  • Quantitative Comparison: In a simulated 19-cell network, a Cooperative MIMO system with 3 cooperating BSs and 16 antennas per BS actually outperformed a non-cooperative Massive MIMO system equipped with 48 antennas per BS.
  • Quantization: With a quantization coverage parameter of $\zeta = 0.97$, using just 4 bits per LLR in the Cooperative MIMO setup matched the ideal infinite-capacity fronthaul performance.

Limitations and Future Work

The authors acknowledge that their analytical state evolution relies on the assumption of uncorrelated Rayleigh fading channels and i.i.d. Gaussian signature sequences. In real-world deployments, channels often exhibit spatial correlation. While the AMP framework can theoretically handle this, it would require redesigning the Bayesian denoiser to incorporate the new channel statistics—a promising direction for future work. Additionally, the Cooperative MIMO scheme assumes that base stations know the signature sequences and large-scale fading coefficients of out-of-cell users, which implies a non-trivial signaling and coordination overhead in practical networks.

Summary updated

Abstract

This paper considers sparse device activity detection for cellular machine-type communications with non-orthogonal signatures using the approximate message passing algorithm. This paper compares two network architectures, massive multiple-input-multiple-output (MIMO) and cooperative MIMO, in terms of their effectiveness in overcoming inter-cell interference. In the massive MIMO architecture, each base station (BS) detects only the users from its own cell while treating inter-cell interference as noise. In the cooperative MIMO architecture, each BS detects the users from neighboring cells as well; the detection results are then forwarded in the form of a log-likelihood ratio (LLR) to a central unit where final decisions are made. This paper analytically characterizes the probabilities of false alarm and missed detection for both architectures. The numerical results validate the analytic characterization and show that as the number of antennas increases, a massive MIMO system effectively drives the detection error to zero, while as the cooperation size increases, the cooperative MIMO architecture mainly improves the cell-edge user performance. Moreover, this paper studies the effect of LLR quantization to account for the finite-capacity fronthaul. The numerical simulations of a practical scenario suggest that in specific case cooperating three BSs in a cooperative MIMO system achieves about the same cell-edge detection reliability as a non-cooperative massive MIMO system with four times the number of antennas per BS.
BibTeX
@article{chen2019multi,
  author = {Zhilin Chen and Foad Sohrabi and Wei Yu},
  title = {{Multi-Cell Sparse Activity Detection for Massive Random Access: Massive MIMO Versus Cooperative MIMO}},
  journal = {IEEE Trans. Wireless Communications},
  year = {2019},
  url = {https://openreview.net/forum?id=qr3vj1PlsI},
}

Loading related content...

Figure