Deep Learning-Based Joint Activity Detection and Channel Estimation for Cell-Free Massive MIMO

Massive machine-type communications (mMTC) is a key scenario for 6G networks featured by ultra-high connection density and sporadic data transmission, which may lead to high signaling overhead and transmission latency with traditional multiple access protocols and network structures. Grant-free non-orthogonal multiple access (GF-NOMA) and cell-free massive MIMO (CF-MIMO) offer effective solutions. This paper proposes a model-driven deep learning algorithm, namely Iterative Shrinkage-Thresholding Algorithm Network (ISTANet), for joint active user detection and channel estimation (JADCE) in high-activity CF-MIMO systems, which employs a convolutional neural network (CNN) for nonlinear transformations and incorporates skip connections to enhance training stability and reconstruction performance. The simulation results demonstrate its adaptability to varying pilot lengths and its outstanding detection performance, especially in high-activity scenarios.