Abstract:Cell-free massive MIMO (CF-mMIMO) networks have recently emerged as a promising solution to tackle the challenges arising from next-generation massive machine-type communications. In this paper, a fully grant-free deep learning (DL)-based method for user activity detection in CF-mMIMO networks is proposed. Initially, the known non-orthogonal pilot sequences are used to estimate the channel coefficients between each user and the access points. Then, a deep convolutional neural network is used to estimate the activity status of the users. The proposed method is "blind", i.e., it is fully data-driven and does not require prior large-scale fading coefficients estimation. Numerical results show how the proposed DL-based algorithm is able to merge the information gathered by the distributed antennas to estimate the user activity status, yet outperforming a state-of-the-art covariance-based method.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to detect the activity status of users efficiently and accurately in the cell - free massive MIMO (CF - mMIMO) network with grant - free random access. Specifically, the paper proposes a completely deep - learning - based method to achieve user activity detection without prior estimation of large - scale fading coefficients. ### Problem Background With the wide application of Internet of Things (IoT) technology in various fields, the interconnection of a large number of intelligent devices has given rise to massive machine - type communications (mMTC) systems. In the upcoming 6G era, mMTC systems will face higher scalability and device battery life requirements, and more reliable and low - latency communication protocols are needed. Traditional grant - based protocols are inefficient in handling small data packets with intermittent transmissions due to their large signaling overhead, resulting in degraded network performance and unnecessary energy consumption. ### Existing Challenges 1. **Estimation of large - scale fading coefficients**: In the grant - free network, it is very difficult to estimate the large - scale fading coefficients, and imperfect estimations may be generated, thus affecting the accuracy of user activity detection. 2. **Non - orthogonal pilot sequences**: Due to the limited length of the channel coherence time interval, it is not possible to allocate orthogonal pilot sequences to each device. Therefore, non - orthogonal pilot sequences are usually used, which will lead to severe co - channel interference and complicate channel estimation and user activity detection. ### Solutions Proposed in the Paper The paper proposes a completely grant - free deep - learning (DL) method for user activity detection, which has the following characteristics: 1. **Blind detection**: This method is completely data - driven and does not require prior estimation of large - scale fading coefficients. 2. **Deep convolutional neural network (CNN)**: Use the known non - orthogonal pilot sequences to estimate the channel coefficients, and then use the deep convolutional neural network to estimate the activity status of users. 3. **Distributed antenna information fusion**: By combining the information collected by distributed antennas, effectively detect which users are in an active state. ### Numerical Results The numerical results show that the proposed DL algorithm can effectively combine the information collected by distributed antennas to estimate the user activity state, and is superior in performance to the existing covariance - based methods. ### Summary This paper aims to solve the problem of user activity detection in the CF - mMIMO network in the grant - free random access scenario. It proposes a blind detection method, which realizes efficient user activity identification through deep - learning technology without relying on the estimation of large - scale fading coefficients.

Blind User Activity Detection for Grant-Free Random Access in Cell-Free mMIMO Networks

Deep Learning-Based Approach for User Activity Detection with Grant-Free Random Access in Cell-Free Massive MIMO

Matrix Factorization Based Blind Bayesian Receiver for Grant-Free Random Access in mmWave MIMO mMTC

Asynchronous Grant-Free Random Access: Partially Uni-Directional Message Passing for Joint User Activity Detection and Channel Estimation.

Deep Learning-Based Blind Multiple User Detection for Grant-free SCMA and MUSA Systems

ML and MAP Device Activity Detections for Grant-Free Massive Access in Multi-Cell Networks

Joint User Activity and Data Detection in Grant-Free NOMA using Generative Neural Networks

Non-Bayesian Activity Detection, Large-Scale Fading Coefficient Estimation, and Unsourced Random Access with a Massive MIMO Receiver

Device Activity Detection for Massive Grant-Free Access Under Frequency-Selective Rayleigh Fading

Noncoherent Massive Random Access for Inhomogeneous Networks: From Message Passing to Deep Learning

Deep Learning-Based User Activity Detection and Channel Estimation in Grant-Free NOMA

Deep Neural Network-Based Blind Multiple User Detection for Grant-free Multi-User Shared Access

A Flexible Framework for Grant-Free Random Access in Cell-Free Massive MIMO Systems

Joint Activity Detection and Channel Estimation in Cell-Free Massive MIMO Networks with Massive Connectivity

Joint Activity and Blind Information Detection for UAV-Assisted Massive IoT Access

A Deep Learning Approach for User-Centric Clustering in Cell-Free Massive MIMO Systems

Active Detection and Channel Estimation Schemes for Massive Random Access in User-Centric Cell-Free Massive MIMO System

Block Sparse Bayesian Learning Based Joint User Activity Detection and Channel Estimation in Grant-Free MIMO-NOMA

Sensing and Deep CNN-Assisted Semi-Blind Detection for Multi-User Massive MIMO Communications

Distributed Joint User Activity Detection, Channel Estimation, and Data Detection via Expectation Propagation in Cell-Free Massive MIMO

Prior Information Aided Deep Learning Method for Grant-Free NOMA in mMTC