2026· IEEE Transactions on Machine Learning in Communications and Networking· Vol 4, pp. 1261-1278· 0 citations· 36 references
TL;DR
This study proposes a novel FL framework where the CF-mMIMO participants are APs, and proposes a client selection strategy that prioritizes APs based on their average received signal power, showing competitive performance compared to baseline methods, while also addressing the scalability and privacy requirements of mMTC systems.
Abstract
The rapid growth of massive machine-type communications (mMTC), combined with advances in edge intelligence, is paving the way for low-latency, low-overhead connectivity. However, the sporadic nature of device activity in mMTC scenarios calls for efficient methods to determine which devices are active at any given time. This motivates collaborative learning within a cell-free massive multiple-input multiple-output (CF-mMIMO) architecture, where the wide geographical distribution of access points (APs) and their joint coordination make distributed learning efficient and secure. Consequently, federated learning (FL) emerges as a promising solution. Indeed, FL enables participants to train a shared model without exchanging raw local data, thereby enhancing data privacy at the AP side and lowering the fronthaul load while leveraging heterogeneous, location-dependent data. The present study proposes a novel FL framework where the CF-mMIMO participants are APs. Due to differences in device behavior, mobility patterns, and environmental factors across the network, the data collected at each AP is often non-independent and non-identically distributed (non-IID). This heterogeneity slows down the convergence of standard FL training and increase variability among client updates, particularly under heterogeneous radio feature distributions. To address this, we propose a client selection strategy that prioritizes APs based on their average received signal power. Our approach shows competitive performance compared to baseline methods, while also addressing the scalability and privacy requirements of mMTC systems. Furthermore, our study analyzes the fairness achieved by APs across devices and presents a representative, percentage-scale analysis of power-consumption gains relative to detection performance when some APs are dropped (i.e., taken out of service), examining two AP-dropping strategies. These results bring valuable insights and set guidelines towards the implementation of FL-based activity detection in CF-mMIMO networks.
Experimental findings indicate that TFL achieves quicker convergence and delivers a superior weighted ISAC utility for users, alongside improved communication and sensing performance, and a more potent blend of communication-sensing advantages compared to per-cell learning, standard federated learning, privacy-compromising federated learning, and mobility-aware federated learning approaches.
Jillella Venkateswara Rao, Pydimarri Padmaja, S. Ravikanth et al.· Journal of Intelligent Decis...· 0 citations
A cross-layer end-to-end (E2E) resource orchestration framework for green CF-mMIMO ISAC systems with distributed multi-target detection is developed and a fundamental implementation trade-off is revealed: FIS provides lower detector-processing complexity and higher detection performance, whereas PIS substantially reduces fronthaul requirements.
Z. Behdad, Özlem Tuğfe Demir, Ki Won Sung et al.· 0 citations
Cell-free massive multiple-input multiple-output (CF mMIMO) requires effective power control, but centralized processing relies on global instantaneous channel state information (CSI) and creates heavy fronthaul load. This letter focuses on low-overhead distributed power control under limited fronthaul capacity. We propose an information bottleneck (IB)-based policy that exchanges compact latent messages instead of raw local observations, and we train it using cluster-based federated learning to keep raw CSI local. The IB penalty provides an explicit information-rate proxy for online coordination, while robustness under imperfect CSI and data locality are evaluated as supporting effects rather than formal guarantees. Simulations show that the proposed method approaches a centralized benchmark with much lower effective overhead and stable behavior under channel estimation errors.
Yukun Ma, Jiayi Zhang, Zih-Yi Liu et al.· IEEE Wireless Communications...· 0 citations
Over-the-air FL with EH MDs under heterogeneous data distributions under heterogeneous data distributions is studied, and the proposed unified framework improves fairness or personalization, depending on the operating mode, while reducing communication overhead.
F. Bagci, Busra Tegin, Mohammad Kazemi et al.· 0 citations
The unprecedented growth of machine-type devices has underscored the need for fundamental solutions to support emerging massive connectivity. In particular, unsourced random access (URA) has emerged as a promising paradigm, reframing the massive connectivity problem as a coding-theoretic challenge with favorable energy and spectral efficiency. Among the widely studied URA models, the Gaussian multiple-access channel (GMAC) and multi-input multi-output (MIMO) systems are of particular significance. Sparse code design is well-suited for URA, offering scalable solutions while retaining many advantages of legacy access protocols. However, existing sparse code designs often suffer from limited sparsity control, inefficient interference cancellation, and a strong dependence on specific channel code designs, posing challenges for long-term adaptability as more powerful channel codes continue to evolve. In MIMO-URA systems, additional activity detection and channel estimation phases typically lead to increased missed detection (MD) and false alarm (FA) errors compared with the GMAC model, which does not require these phases. While prior studies have predominantly focused on minimizing MD errors, the effective mitigation of FA errors remains an open problem. To address this challenge, we propose a sparse code with slotted transmission under the GMAC model, combined with an analytical power division strategy to enhance interference cancellation. Furthermore, we introduce a novel MIMO receiver framework based on joint pattern–data–channel (JPDC) estimation, which significantly reduces FA errors by leveraging the intrinsic correlation between user activity and transmitted data. Notably, the proposed method achieves improved overall system performance without requiring additional transmission overhead or complex algorithms.
Zhentian Zhang, M. J. Ahmadi, kai-kit Wong et al.· IEEE Transactions on Wireles...· 5 citations
Reconfigurable intelligent surface (RIS) is a promising technology for enhancing coverage and connectivity in Internet of Things (IoT) networks. However, the passive nature of RIS impedes the decoupling of the BS-RIS channel and RIS-user channels from the cascaded channel, making both channel estimation (CE) and active user detection (AUD) challenging in RIS-assisted IoT networks. To address this problem, this paper formulates the task of joint CE and AUD as a multi-layer sparse recovery problem by exploiting the sparsity structures in cascaded channels, as well as their joint scaling property whereby the cascaded channel of each user equipment (UE) can be normalized relative to that of an active reference UE through a diagonal matrix. Within the framework of variational Bayesian inference, the formulated problem is transformed into a Bethe free energy (BFE) minimization problem. To solve it effectively, we propose a low-complexity BFE-based hybrid message passing (BFE-HMP) algorithm that introduces various moment matching constraints. More importantly, two decentralized implementations of the proposed BFE-HMP algorithm are developed for the emerging decentralized base station (BS) architectures. Extensive simulation results validate the superiority of the proposed BFE-based algorithms in terms of normalized mean square error and the detection error probability over the existing state-of-the-art methods.
Yufei Cao, Heng Liu, Tao Yu et al.· IEEE Transactions on Wireles...· 0 citations