Generative Diffusion Model Driven Massive Random Access in Massive MIMO Systems

Massive random access is an important technology for achieving ultra-massive connectivity in next-generation wireless communication systems. It aims to address key challenges during the initial access phase, including active user detection (AUD), channel estimation (CE), and data detection (DD). This paper examines massive access in massive multiple-input multiple-output (MIMO) systems, where deep learning is used to tackle the challenging AUD, CE, and DD functions. First, we introduce a Transformer-AUD scheme tailored for variable pilot-length access. This approach integrates pilot length information and a spatial correlation module into a Transformer-based detector, enabling a single model to generalize across various pilot lengths and antenna numbers. Next, we propose a generative diffusion model (GDM)-driven iterative CE and DD framework. The GDM employs a score function to capture the posterior distributions of massive MIMO channels and data symbols. Part of the score function is learned from the channel dataset via neural networks, while the remaining score component is derived in a closed form by applying the symbol prior constellation distribution and known transmission model. Utilizing these posterior scores, we design an asynchronous alternating CE and DD framework that employs a predictor-corrector sampling technique to iteratively generate channel estimation and data detection results during the reverse diffusion process. Simulation results demonstrate that our proposed approaches significantly outperform baseline methods with respect to AUD, CE, and DD.

💡 Research Summary

**
This paper addresses the three fundamental tasks of massive random access (MRA) in massive multiple‑input multiple‑output (MIMO) systems—active user detection (AUD), channel estimation (CE), and data detection (DD)—by proposing a unified two‑stage framework.

Stage 1: Variable‑Pilot‑Length Transformer‑AUD (VPL‑AUDNet).
Traditional transformer‑based AUD solutions assume a fixed pilot length, requiring retraining when the pilot overhead changes. The authors introduce a pilot‑length‑adaptive module (PLAM) that explicitly feeds the pilot length into the network and dynamically scales the attention weights, enabling a single model to operate across a wide range of pilot lengths (e.g., 4, 8, 16 symbols) and antenna configurations without re‑training. In addition, a Spatial Correlation Module (SCM) extracts inter‑antenna correlation from the received signal matrix, enriching the transformer’s representation in correlated or non‑Rayleigh channels. The overall architecture consists of preprocessing → PLAM → SCM → multi‑head attention → decoder → binary thresholding, trained with binary cross‑entropy to directly output the active‑user support set. Experimental results show that VPL‑AUDNet outperforms covariance‑based and earlier transformer‑based detectors by 5–7 % in detection accuracy and 3–4 % in F1‑score, especially when pilot resources are scarce.

Stage 2: Generative Diffusion Model (GDM) for Joint CE and DD (JCEDD).
The second stage treats the joint estimation of the high‑dimensional channel matrix H and the transmitted data symbols X as a reverse diffusion problem. In the forward diffusion, Gaussian noise is gradually added to the concatenated vector (