Respiration monitoring via radio signals enables contactless health sensing but suffers from interference caused by nearby motion. We propose a robust respiration sensing framework using Cell-free Massive MIMO (CF-mMIMO), which leverages spatial macro-diversity for interference resilience. Specifically, we analyze respiration sensing in single-antenna channels using Power Spectral Density (PSD) to reveal the impact of interference on the breathing channel’s movement spectrum. Based on this, we introduce a new metric, Sensing-Signal-to-Interference Ratio (SSIR), to evaluate local channel quality without requiring ground truth. Then, we design a Weighted Antenna Combining (WAC) method to prioritize reliable sensing links and suppress distortion. Experimental validation using a 64-antenna CF-mMIMO testbed with 100 Orthogonal Frequency-Division Multiplexing (OFDM) subcarriers over an 18 MHz bandwidth confirms the framework’s robustness. In the presence of interference, the WAC method achieves a mean waveform correlation of 0.81 with ground truth, significantly outperforming single-antenna (0.52), averaging-based methods (0.53), and existing Wi-Fi approaches. Finally, we analyze the impact of time, frequency, and spatial resource allocation on both communication and sensing performance. Results show that increasing bandwidth and antenna count benefits both communication and sensing. With a sufficient number of antennas, respiration sensing remains accurate even with long coherence times (1 second) and narrow bandwidths (3 subcarriers), enabling its integration into communication systems with negligible overhead, making it practically “for free”. This makes CF-mMIMO a promising architecture for robust and scalable Integrated Sensing and Communication (ISAC) health monitoring.

Integrating dense channel fingerprints into deep learning (DL) becomes a promising way to realize precise three-dimensional (3D) indoor localization. However, most existing methods are frequency-dependent, which limits the localization precision when operating in different frequency bands. To address this challenge, this paper proposes a masked Transformer encoder (MTE) model capable of using the channel state information (CSI) data of an arbitrary number of sub-channels (frequency bands) as input. The proposed MTE model can locate a UE using frequency-scalable CSI data, to realize resilient localization. We first introduce how to transform CSI data into sequential data suitable for Transformer-based models, with length of the sequence determined by the number of sub-channels. Based on this, an MTE model is designed to achieve resilient FP localization with frequency-scalability, i.e., capable of processing the CSI data of an arbitrary number of sub-channels. Next, we construct a 3D CSI FP dataset using ray-tracing (RT) simulations based on real-world indoor scenarios and versatile electromagnetic (EM) coefficients. The reliability of the dataset is verified by measurement data. Extensive experiments demonstrate that the MTE model outperforms many state-of-the-art baselines, classical time-series models, and alternative Transformer-based methods, especially under arbitrary sub-channel CSI data. Moreover, we demonstrate that the MTE model also offers many advantages in terms of training and storage costs through comparisons with conventional models.

sponsorship: This research is supported by iSEE-6G project under the Horizon Europe Research and Innovation program with Grant Agreement No. 101139291. (iSEE-6G project under the Horizon Europe Research and Innovation program|101139291)

This paper demonstrates the feasibility of respiration pattern estimation utilizing a communication-centric cell-free massive MIMO OFDM Base Station (BS). The sensing target is typically positioned near the User Equipment (UE), which transmits uplink pilots to the BS. Our results demonstrate the potential of massive MIMO systems for accurate and reliable vital sign estimation. Initially, we adopt a single antenna sensing solution that combines multiple subcarriers and a breathing projection to align the 2D complex breathing pattern to a single displacement dimension. Then, Weighted Antenna Combining (WAC) aggregates the 1D breathing signals from multiple antennas. The results demonstrate that the combination of space-frequency resources—specifically in terms of subcarriers and antennas—yields higher accuracy than using only a single antenna or subcarrier. Our results significantly improved respiration estimation accuracy by using multiple subcarriers and antennas. With WAC, we achieved an average correlation of 0.8 with ground truth data, compared to 0.6 for single antenna or subcarrier methods—a 0.2 correlation increase. Moreover, the system produced perfect breathing rate estimates. These findings suggest that the limited bandwidth (18 MHz in the testbed) can be effectively compensated by utilizing spatial resources, such as distributed antennas.

This work focuses on channel estimation in extremely large aperture array (ELAA) systems, where near-field propagation and spatial non-stationarity introduce complexities that hinder the effectiveness of traditional estimation techniques. A physics-based hybrid channel model is developed, incorporating non-binary visibility region (VR) masks to simulate diffraction-induced power variations across the antenna array. To address the estimation challenges posed by these channel conditions, a novel algorithm is proposed: Visibility-Region-HMM-Aided Polar-Domain Simultaneous Orthogonal Matching Pursuit (VR-HMM-P-SOMP). The method extends a greedy sparse recovery framework by integrating VR estimation through a hidden Markov model (HMM), using a novel emission formulation and Viterbi decoding. This allows the algorithm to adaptively mask steering vectors and account for spatial non-stationarity at the antenna level. Simulation results demonstrate that the proposed method enhances estimation accuracy compared to existing techniques, particularly in low-SNR and sparse scenarios, while maintaining a low computational complexity. The algorithm presents robustness across a range of design parameters and channel conditions, offering a practical solution for ELAA systems.

In this paper, we present a spatially consistent A2G channel model based on probabilistic LOS/NLOS segmentation to parameterize the deterministic path loss and stochastic shadow fading model. Motivated by the limitations of existing Unmanned Aerial Vehicle (UAV) channel models that overlook spatial correlation, our approach reproduces LOS/NLOS transitions along ground user trajectories in urban environments. This model captures environment-specific obstructions by means of azimuth and elevation-dependent LOS probabilities without requiring a full detailed 3D representation of the surroundings. We validate our framework against a geometry-based simulator by evaluating it across various urban settings. The results demonstrate its accuracy and computational efficiency, enabling further realistic derivations of path loss and shadow fading models and thorough outage analysis.

Accurate Probability of Line-of-Sight$(P_{\text{LoS}})$modeling is important in evaluating the performance of Unmanned Aerial Vehicle (UAV)-based communication systems in urban environments, where real-time communication and low latency are often major requirements. Existing$P_{L o S}$models often rely on simplified Manhattan grid layouts using International Telecommunication Union (ITU)-defined builtup parameters, which may not reflect the randomness of real cities. Therefore, this paper introduces the Urban Line-ofSight Simulator (ULS) to model$P_{\text{LoS}}$for three random city layouts with varying building sizes and shapes constructed using ITU built-up parameters. Based on the ULS simulated data, we obtained the empirical$P_{L o S}$for four standard urban environments across three different city layouts. Finally, we analyze how well Manhattan grid-based models replicate$P_{L o S}$results from random and real-world layouts, providing insights into their applicability for time-critical communication systems in urban IoT networks.

Communications between unmanned aerial vehicles (UAVs) play an important role in deploying aerial networks. Although some studies reveal that drone-based air-to-air (A2A) channels are relatively clear and thus can be modeled as free-space propagation, such an assumption may not be applicable to drones flying in low altitudes of built-up environments. In practice, low-altitude A2A channel modeling becomes more challenging in urban scenarios since buildings can obstruct the line-of-sight (LOS) path, and multipaths from buildings lead to additional losses. Therefore, we herein focus on modeling low-altitude A2A channels considering a generic urban deployment, where we introduce the evidence of the small-size first Fresnel zone at the millimeter-wave (mmWave) band to approximately derive the LOS probability. Then, the path loss under different propagation conditions is investigated to obtain an integrated path loss model. In addition, we incorporate the impact of imperfect beam alignment on the path loss, where the relation between path loss fluctuation and beam misalignment level is modeled as an exponential form. Finally, comparisons with the 3GPP model show the effectiveness of the proposed analytical model. Numerical simulations in different environments and heights provide practical deployment guidance for aerial networks.

For reliable and efficient communications of aerial platforms, such as unmanned aerial vehicles (UAVs), the cellular network is envisioned to provide connectivity for the aerial and ground user equipment (GUE) simultaneously, which brings challenges to the existing pattern of the base station (BS) tailored for ground-level services. Thus, we focus on the coverage probability analysis to investigate the coexistence of aerial and terrestrial users, by employing realistic antenna and channel models reported in the 3rd Generation Partnership Project (3GPP). The homogeneous Poisson point process (PPP) is used to describe the BS distribution, and the BS antenna is adjustable in the down-tilted angle and the number of the antenna array. Meantime, omnidirectional antennas are used for cellular users. We first derive the approximation of coverage probability and then conduct numerous simulations to evaluate the impacts of antenna numbers, down-tilted angles, carrier frequencies, and user heights. One of the essential findings indicates that the coverage probabilities of high-altitude users become less sensitive to the down-tilted angle. Moreover, we found that the aerial user equipment (AUE) in a certain range of heights can achieve the same or better coverage probability than that of GUE, which provides an insight into the effective deployment of cellular-connected aerial communications.

Unmanned aerial vehicles (UAVs) have aroused great attention for future wireless communications since the drone can be used as an aerial base station and also as a cellular user to perform on-demand tasks. Because the reliable and robust connection between UAV and ground station (GS) plays an essential role in the UAV-based air-to-ground (AG) communications, the accurate radio channel model is needed. In this paper, we conducted AG wideband radio channel measurements in a hilly environment at 6.5 GHz with a bandwidth of 500 MHz, which is one of the high bands for ultra-wideband (UWB) communications in IEEE standards. The large-scale characteristics, including path loss and shadow fading, are represented by the close-in (CI) model. The small-scale characteristics that illustrate the fading and the multipath effect are analyzed with three critical parameters, including the small-scale fading, Rician K-factor and root-mean-square (RMS) delay spread. Afterward, the tapped delay line (TDL) model is built according to the detected multipath in the power delay profile (PDP), and the channel impulse response (CIR) is discussed for wideband channel modeling. The results can be used for the design and analysis of the AG systems.

Powered by the Internet of Things (IoT), the cellular, vehicular, and other emerging networks, such as space-air-ground integrated networks, are expected to comprehensively evolve into the new era of the Internet of Everything (IoE) in which the propagation links between the IoT devices or sensors become diversified. Thus, a general propagation channel model is required. The line-of-sight (LOS) probability plays an essential role in the channel modeling, especially for built-up environments where the blockages from buildings are unfavorable to the signal transmission. The 4-D LOS probability model with considering the 3-D environment and frequency is comprehensively investigated in this article. Using the geometry-based stochastic method, the LOS probability is derived for arbitrary sizes, heights, and orientations of buildings in finely defined urban scenarios. The major contribution is the universality, simplicity, and good extension of the proposed model with comprehensive considerations of the influence of frequency, the type of urban, and the height of transceiver. The simulation results show that the propagation with higher frequency has a higher LOS probability. Moreover, the size, height, and density of building in urban environments are closely related to the LOS probability. The good agreements of the comparisons with the ray-tracing (RT) model and standard models show the accuracy of our proposed model. These results will be useful in the modeling of various channels and the design of IoT wireless communications systems.