Turbo product codes have emerged as promising candidate channel codes for hyper-reliable low-latency communication scenarios owing to the inherent parallelism of product code decoding. Recently proposed product-coded MIMO systems with polar component codes demonstrate both performance gains and latency reductions compared to traditional 1-D coded systems. However, the existing successive cancellation list (SCL)-based decoding algorithm in polar-product-coded MIMO systems suffers from high memory costs due to storing internal LLRs and still exhibits potential for further performance improvements. To enhance the performance of product-coded MIMO systems and alleviate storage overhead, this paper investigates decoding schemes based on list ordered statistics decoding (List-OSD). A hardware-friendly soft-output computation method for List-OSD is proposed, which exploits both candidate codewords and pruned paths generated during decoding. Two soft-output list ordered statistics (SOLOS) decoding schemes are presented, respectively incorporating the Pyndiah method and the proposed soft-output computation technique. Simulation results show that the two proposed decoding schemes outperform conventional decoders across various code rates and constructions in product-coded MIMO systems. For the (64, 42)2 polar-product code, the proposed decoders with a list size of 32 achieve a 0.4 dB performance gain over the SCL-based decoder with a list size of 8 at comparable latency, while reducing the byte-level memory usage by 51.2%.
The minimum mean square error (MMSE) detector is widely adopted for massive MIMO systems, but its reliance on matrix inversion results in prohibitive computational complexity. To solve this problem, linear iterative detectors, e.g., weighted Neumann series approximation (wNSA), can circumvent matrix inversion and substantially reduce complexity. However, their performance degradation in high-dimensional high-order MIMO scenarios remains unacceptable. In this paper, we propose WeFTNet, a deep neural network (DNN)-aided weighted Fourier transform (FT)-based approximate MMSE detector for massive MIMO systems. Specifically, the proposed WeFTNet integrates two unexplored features: weighted FT-based MMSE approximation and DNN-driven weighting factor optimization. Furthermore, a geometrically decreasing weighting factor model is theoretically derived and introduced to enhance the training stability and performance robustness. Numerical results demonstrate that the WeFTNet consistently achieves near-MMSE bit error rate (BER) performance in various MIMO scenarios. Compared with the other linear iterative detectors, the proposed WeFTNet succeeds in delivering up to 3.5 dB performance gains while achieving 57% complexity reduction in 256 × 128 256-QAM MIMO systems.
Automatic modulation recognition (AMR) is a cornerstone of dynamic spectrum allocation and cognitive radio. Despite the promise of deep learning-based methods, their reliance on large labeled datasets limits their practicality in scenarios with scarce labeled samples. To address this, we propose a confidence-aware semi-supervised learning (CASL) framework tailored for AMR. CASL adopts a two-stage approach: first, a time-warping-based data augmentation strategy is used to extract general features through contrastive learning. In the second stage, a self-adaptive threshold mechanism generates high-quality pseudo-labels while mitigating confirmation bias using class-level information. To enhance computational efficiency and performance, we design a pyramid learnable filter-based encoder that captures multi-scale features from large-scale, long-sequence signals, reducing inference time by about half compared with the self-attention-based encoder variant. Simulation results demonstrate substantial performance gains on four public datasets, achieving a 20.43% average accuracy improvement over state-of-the-art methods in label-scarce scenarios and up to 38.64% in low signal-to-noise ratio (e.g., 0 dB) conditions. These results highlight CASL’s potential for practical AMR applications in challenging environments.
UAV route construction methods have been studied using protocols for ad hoc networks composed of wireless multi-hop terminals installed in each house, such as smart meters. In the previous research, in order to eliminate the risk of collision due to overlapping routes when multiple UAVs navigate simultaneously, we proposed a method to lock the link used for the established route so that it cannot be used for the route of another UAV, as well as a method to hierarchize links to eliminate route construction failure due to insufficient links caused by locked routes. The previous research did not construct routes that take into account the priority of UAVs and the balance between efficiency, which appears in travel time, and safety in avoiding densely populated areas, in response to the goods to be transported and the type of delivery service. In this paper, we propose a method for prioritizing the available hierarchical altitude layers and controlling the balance between travel time and avoidance of densely populated areas, adapted to the service requirements of each UAV. The effectiveness and characteristics of the proposed method are evaluated and verified by computer simulation.
Indoor localization has emerged as a critical enabling technology for various smart applications, yet its performance is constrained by multipath propagation, signal blockage, and environmental dynamics. While recent deep learning (DL)-based approaches have demonstrated promising improvements over traditional techniques, their effectiveness is often limited by high architectural complexity, limited robustness against electromagnetic-environment variations, and weak interpretability. To address these limitations, this paper proposes a novel indoor localization framework that leverages the structural characteristics of dual-polarized channel state information (CSI) through disentangled representation learning. By decoupling the shared and exclusive latent features embedded in the two polarization channels, the proposed framework is able to isolate location-relevant properties from device-specific and electromagnetic distortions. Furthermore, a three-stage joint optimization strategy is developed to ensure effective feature disentanglement and robust localization performance. Compared with various representative localization models, the proposed framework achieves over 19% enhancement in localization accuracy, demonstrating improved adaptability under imperfect CSI, clock-synchronization impairments, different obstacle densities, dynamic scenarios, different antenna configurations, and reduced training data conditions, with better model interpretability and low inference complexity. Ablation studies and feature visualization further confirm the viability of the proposed network structure design, highlighting the effectiveness of information-theoretic disentanglement in extracting location-relevant features.
Cross-silo federated reinforcement learning (FRL) trains a shared policy across silos that must simultaneously respect a per-round uplink budget and on-device privacy. Although differential privacy (DP) and gradient compression each have mature solutions in supervised federated learning, naively stacking them in the policy-optimisation regime fails, for two structural reasons. First, whole-client DP injects noise whose ratio to the signal scales as √d/N, which for neural policies exceeds one at every practical ε, so the policy collapses to random. Second, a single shared compression mask cannot represent each client’s descent direction under non-i.i.d. environments, driving the compressor’s contraction constant toward zero and destroying learning. In this paper, we propose FedDPRL, a federated policyoptimisation algorithm built from two coupled designs. Specifically, a record-level (per-trajectory) Gaussian mechanism clips and noises each trajectory, lifting the effective sample size to R = NB and removing the √d/N barrier so that DP becomes viable on neural policies. Additionally, an error-feedback Top-k compressor with per-client masks preserves each client’s own descent direction while cutting the uplink by an order of magnitude. Experimental results across classic control, continuous control (MuJoCo), and vision with N = 20 clients show that FedDPRL matches dense DP-FedAvg-RL utility at 13–14× less uplink, while a shared mask collapses to near-random.
The design of millimeter-wave (mmWave) communication systems requires comprehensive research on mmWave channels. Although notable exceptions such as New York University and 3GPP have developed well-validated models based on extensive measurements, many other existing mmWave channel models still rely predominantly on theoretical analysis and lack sufficient support from substantial real-world measurements, particularly across multiple scenarios. Based on the measured mmWave channel data, this paper proposes an improved delay-domain clustering algorithm and a novel and more accurate cluster-based Saleh-Valenzuela (SV) channel model for both line-of-sight (LOS) and obstructed line-of-sight (OLOS) cases. This model separately extracts the first cluster amplified by the directional antenna for analysis, which is more suitable for the mmWave characteristics and has higher accuracy. The eight key cluster parameters and their optimal fitting distributions are extracted, and cluster characteristics in multiple scenarios are analyzed, which reveal the mmWave propagation mechanism. The generated fitting parameters are synthesized into simulated channels, and the root-mean-square delay spread, Kolmogorov-Smirnov (KS) distance and Quantile-Quantile analysis are used to validate this model in multiple scenarios. Finally, the overall process of mmWave channel modeling is presented. This paper provides valuable guidance and a novel channel model applicable to multiple scenarios for the simulation, design, and development of mmWave communication systems.
Orthogonal frequency division multiplexing (OFDM) has been recognized as a promising technique for underwater acoustic (UWA) communication due to its high spectral efficiency and multipath-mitigation capability, which is, however, highly vulnerable to carrier frequency offset (CFO). Under multipath UWA channels, the current CFO estimation approaches generally face a fundamental trade-off between estimation accuracy and spectral efficiency: while the pilot-based methods suffer from error propagation caused by channel uncertainty, null-based approaches lead to significant bandwidth overhead. To address this issue, a novel virtual null subcarriers (VNS)-based CFO estimation method is proposed to improve estimation performance without allocating physical null subcarriers. Specifically, by projecting the received pilot vector into the orthogonal complement of the pilot observation subspace to suppress the dominant channel-induced component, the residual term of projection, dominated by noise and CFO mismatch, is defined as VNS. Thus, estimating CFO towards minimizing the energy of VNS is capable of addressing the impairment of UWA channels while avoiding the overhead of physical null subcarriers. Since the construction of the orthogonal projection matrix is hindered by the rank deficiency arising from limited pilot overhead and the ill-conditioning caused by severe multipath channels, a dual-constraint adaptive regularization approach is designed to adaptively balance numerical stability against the preservation of sparse channel structures. Moreover, a multi-stage particle swarm optimization algorithm is adopted to facilitate a global search of the multimodal non-convex VNS-based cost function in the presence of local minima. Finally, simulations and field experiments demonstrate that the proposed methods achieve superior performance compared to the mainstream approaches.
LEO mega-constellations equipped with laser intersatellite links (LISLs) are key enablers of future 6G networks. However, solar background noise can disrupt LISLs and degrade constellation performance. Existing studies rely on time-consuming simulations that only indicate whether LISLs are disrupted, failing to reveal the relationship between constellation configurations and the spatiotemporal patterns of LISL disruptions. In this paper, we develop a constellation-level impacted model for solar background noise on LEO mega-constellations based on geometric analysis. First, we extract constellation characteristic quantities (specifically, the orbital plane normal vectors and the inter-plane LISL equatorial angle) to characterize the geometric features of LISLs. By integrating these quantities with the solar position model, we reveal that LISL disruptions are strictly confined to specific temporal intervals governed by the periodic variation of the constellation-sun geometry. We analytically derive these time intervals and define them as risk time windows. Subsequently, within these risk time windows, we determine the spatiotemporal distribution of disrupted LISLs. Guided by this analysis, we propose a Minimum Equatorial Angle Connection Strategy (MEACS) to decrease the duration of risk time windows and enhance topology integrity availability (defined as the proportion of time the constellation operates without LISL disruptions). Extensive validation against STK simulations (utilizing Starlink and OneWeb configurations) and on-orbit data demonstrates that the proposed model reduces computation time by over 99% while maintaining the relative error of LISL states below 0.1%.
The shift towards sixth-generation (6G) wireless networks calls for advanced waveforms to support emerging applications such as ultra-reliable and massive machine-type communications. Among the proposed candidates, orthogonal time frequency space (OTFS), orthogonal delay Doppler division multiplexing (ODDM), orthogonal time sequency multiplexing (OTSM), and orthogonal chirp division multiplexing (OCDM) share the common feature of exploiting the delay domain for data transmission. While these waveforms offer significant advantages in high-mobility and sensing scenarios, their integration into multi-user uplink systems reveals a critical challenge: timing misalignment (TM). Unlike OFDM, where residual timing offsets translate into phase rotations, delay-based waveforms experience actual shifts of resources, whose impact we investigate in this work.We show that the aforementioned waveforms can be unified under a time-invariant setup through a common input–output relation, which enables a unified analysis of TM in uplink resource allocation. The results reveal a fundamental trade-off between interference mitigation and guard overhead in the presence of TM. We further benchmark the observed behavior against OFDM under Doppler spread as a reference impairment and highlight the analogy between TM in delay-based allocation and Doppler spread in frequency-based allocation. The results show that delay-based waveforms can maintain a net performance advantage over OFDM in the considered setup. Beyond this unified Doppler-free analysis, we evaluate the joint impact of Doppler spread and TM on each waveform individually. We also investigate BER performance and show how coding can mitigate the resulting interference. This study therefore establishes a problem framework that motivates the development of future compensation and scheduling strategies for practical multi-user operation of delay-based waveforms.
The phenomena of distance-dependent degrees of freedom (DoFs) and spherical wavefronts in the near-field region open new avenues for the development of integrated sensing and communication (ISAC). However, most existing works that employ hybrid beamforming architectures lack the flexibility to accommodate spatially varying DoFs and thus do not fully exploit near-field characteristics. In this paper, we consider a near-field ISAC system with a hybrid analog/digital architecture, where the number of active radio frequency (RF) chains is dynamically adjusted to balance the communication rate, sensing accuracy, and power consumption. Specifically, we formulate an energy efficiency maximization problem under the transmit power and sensing beampattern constraints. We develop a model-based optimization framework that jointly optimizes the hybrid beamforming matrices and RF chain selection. To reduce the computational complexity and enhance scalability, we develop a graph neural network with constraint-preserving reparameterization (GNN-CPR), which leverages a knowledge-guided learning mechanism to learn the beamforming matrices through graph-structured feature aggregation while satisfying the constraints. Simulation results demonstrate that the proposed model-based optimization and GNN-CPR algorithms achieve higher energy efficiency than the baseline schemes. While the proposed model-based optimization algorithm provides better performance, the proposed GNN-CPR algorithm offers a significant advantage in computational efficiency.
In bandwidth- and computation-constrained communication scenarios, achieving reliable and high-fidelity transmission of 3D point cloud data under dynamic SNR conditions remains a significant challenge. This paper presents SMART-ISTC, a Spatially-Modulated Adaptive RobusT model within an Integrated Sensing-Transmission-Computing (ISTC) paradigm. The model employs a PointMamba-based dual-path encoder to extract geometric representations efficiently and introduces a Geometry- and Channel-aware Modulation (GeoCM) module that adaptively adjusts encoded features based on geometric structure and instantaneous channel state. A Geometry-Aware Decoding Module (GADM) restores dense spatial details through residual coordinate prediction and curvature-guided refinement. Evaluations on ModelNet40, ModelNet10, Shapenetv2 and Mountainous Area datasets demonstrate that SMART-ISTC maintains stable performance under fluctuating SNRs from 5 to 35 dB, achieving 3.5–4.6 dB PSNR improvements, significant Chamfer Distance reduction, and approximately 9% lower parameters and FLOPs compared with baselines. Furthermore, SMART-ISTC achieves efficient inference with an average testing time of 0.646 s per sample. Ablation studies validate the effectiveness of the key components within the model.
In the currently evolving 5G cellular networks, small cells are essential to supply reliable communication for intensive traffic demands in hotspots, and several miniaturized base stations (BSs) are deployed as clusters. Meanwhile, such cluster-deployed BSs cause frequent handovers (HOs), and the performance is degraded because of the failures of HOs and excessive signaling overheads. Consequently, HO skipping techniques have been studied recently. In this study, we consider 2-tier heterogeneous networks in which a Poisson-Poisson cluster process is introduced for the deployment of small BSs, and develop an analytical framework for evaluating a specific technique known as the periodic HO skipping scheme. Using stochastic geometry, we analyze two performance metrics, the expected downlink data rate and the HO rate for a mobile user attempting the periodic HO skipping scheme. Numerical evaluations based on the analysis validate the performance of the periodic HO skipping scheme in the 2-tier networks where clusters of small BSs are involved.
Optical wireless communication (OWC) is considered a key enabling technology to support unprecedented data rates and high traffic demands in indoor environments. In this paper, dense OWC networks with arrays of infrared (IR) lasers as optical access points (APs) are studied, where each AP illuminates a confined area and serves multiple users while respecting eye-safety and optical power constraints. To improve spectral efficiency and manage interference in dense OWC networks, we combine non-orthogonal multiple access (NOMA) with a blind interference alignment (BIA) outer precoder. The BIA precoder coordinates transmissions among multiple APs and aligns interference across groups of users, enabling NOMA to be effectively applied within each group. For efficient application of BIA-NOMA in OWC systems, a novel max-min rate optimization problem is formulated that jointly considers user grouping, two-tier power allocation across the BIA and NOMA layers, users’ traffic demands, and optical-specific constraints, thereby improving fairness and overall system performance. Given the complexity of the resulting max-min optimization problem, a dynamic solution is proposed: (i) an RF-aided algorithm establishes weak-strong user associations for effective group formation, and (ii) a dynamic two-tier power allocation algorithm optimizes power distribution among groups and users. Simulation results demonstrate that the proposed approach converges to near-optimal performance and significantly improves sum rate, fairness, and energy efficiency compared to conventional schemes, highlighting the practical feasibility of BIA-NOMA in OWC systems.
Accurate urban path loss prediction is essential for reliable wireless network planning; however, conventional empirical models often fail to capture the complex environmental variability of real urban environments. This paper proposes a geometry-conditioned multimodal framework for urban path loss prediction at 4.85 GHz that combines Tx-Rx geometry with panoramic RGB, semantic, and depth cues. To improve physical consistency and interpretability, the model learns residuals over a fitted LoS/NLoS-separated log-distance baseline rather than directly regressing absolute PL. Evaluation on real-world measurements from three Yokohama routes shows that the proposed multimodal models achieve best within-route RMSEs of 4.87, 4.62, and 4.26 dB on R3, R4, and R6, respectively. These correspond to improvements of 2.83, 1.96, and 1.89 dB over the ITU-R baseline, and 3.23, 1.70, and 1.18 dB over the geometry-only model, respectively. RGB imagery provides the most consistent gain, while semantic and depth cues offer additional but route-dependent benefits. Cross-route transfer achieves RMSEs of 5.61–7.63 dB across the evaluated route pairs. These results suggest that the learned residual correction can provide useful predictions on unseen routes, but the performance remains less consistent than in the within-route setting. This indicates that the visual features capture some transferable environmental context, while still being influenced by route-specific characteristics and spatial correlations. Broader validation with larger datasets, spatially independent evaluation, and temporally aligned imagery is needed to better assess route-invariant generalization.
Free-space optical (FSO) communication has attracted significant interest for non-terrestrial-network (NTN) applications. While C-band FSO systems benefit from high atmospheric transmittance and the maturity of optical components, certain NTN links, such as underwater and satellite communications, need to operate in the visible and short near-infrared (IR) regime (400–1000 nm). However, most visible and short near-IR FSO systems rely on intensity modulation and direct detection (IM-DD) scheme, which inherently limits transmission capacity and receiver sensitivity. To address these limitations, in this paper, we demonstrate coherent communication at 780 nm by developing a wavelength-tunable external cavity laser (ECL) and coherent receivers at this wavelength. We present detailed characterization of the tunable ECL, including its light—current characteristics and the spectrum measurement as well as sensitivity analyses of the coherent receiver. Using these key components, we successfully demodulated 12-channel wavelength-division multiplexed (WDM) 12.5-Gbaud dual-polarization (DP) 16QAM signals. We demonstrated an aggregate gross data rate of 1.2 Tb/s and a net data rate of 1 Tb/s in a multi-channel WDM-based configuration, representing the highest reported capacity in the visible and short near-IR wavelength range to date. This demonstration paves the way for high-capacity, coherent visible and short near-IR FSO communication in NTN applications.
This paper reports the design, fabrication, and indoor experimental evaluation of an inkjet-printed anomalous-reflection metasurface (MTS) for enhancing non-line-of-sight (NLoS) signal reception by redirecting reflections toward a designated NLoS region. A 4.85 GHz phase-gradient MTS is synthesized using a 10-state supercell and fabricated on paper with a foam spacer and an aluminum ground plane. To reliably quantify the intended MTS-reflected component in a multipath-rich indoor environment, a path-selective evaluation method is introduced by combining wideband 8 × 8 MIMO channel sounding with time-of-flight (ToF) gating and double-directional angular power spectrum (DDAPS) analysis. Compared with an identically sized aluminum reference, the MTS exhibits positive enhancement at the designed physical Rx angle for all tested distances, while the angle-delay and double-directional analyses support attribution of the measured improvement to the intended MTS-reflected component.
In sub-terahertz (sub-THz) communications, utilizing building reflection paths is crucial for enhancing transmission efficiency and link robustness, particularly as reliable alternatives when the line-of-sight (LoS) path is compromised. However, these paths are susceptible to environmental obstructions. Therefore, accurate modeling of building reflections is essential for the precise evaluation and design of communication systems. This paper presents the characterization and quasi-deterministic (Q-D) channel modeling of outdoor sub-THz access links, with a specific focus on the impact of vegetation blockage on these dominant reflection paths. Based on dual-band channel measurements at 154 and 300 GHz, angle-resolved channel impulse responses were obtained. In contrast to ray-tracing (RT) simulations based on simplified building models, measurement results reveal irregular attenuation and intermittent disappearance of the reflected paths. We demonstrate that this intermittency is caused by trees and hedges located in front of buildings. By comparing the measured reflection powers with 2D-RT predictions, we show that tree trunks cause complete deterministic path blockages, whereas canopies induce significant random power fluctuations. Consequently, we propose a Q-D channel model for both measured sub-THz frequencies. The model incorporates a vegetation-blockage coefficient into the deterministically computed reflection components, distinguishing the blockage effects according to the specific tree parts intersecting the path. Specifically, the blockage is modeled using double knife-edge diffraction for tree trunks and a statistical excess loss model for foliage scattering.
In display-camera communication (DCC), transmitter-receiver asynchronism and sampling rate mismatches cause temporal inter-symbol interference (ISI) and necessitate synchronization preambles, thereby fundamentally limiting effective throughput. This paper proposes a preamble-free demodulation scheme based on spatio-temporal self-synchronization and maximal-ratio combining (MRC). By superimposing zero-sum orthogonal codes onto the transmitted data, the proposed scheme exploits the rolling shutter-induced spatial mapping of the temporal phase as an inherent fingerprint to robustly estimate temporal parameters from a single captured frame. Under rigorous physical assumptions, this paper proves the mathematical uniqueness of the synchronization model and demonstrates that the zero-sum codes structurally cancel temporal ISI and direct-current (DC) components, such as ambient light and hardware offsets. MRC then optimally recovers the temporally dispersed optical energy to maximize the output signal-to-noise ratio (SNR). Furthermore, fundamental physical bounds dictated by geometric and energy limits are systematically formulated. Information-theoretic optimization reveals a highly scalable performance regime, and quantitative evaluations confirm that the proposed architecture effectively exploits the temporal domain. Specifically, it achieves a sustained Mbps-class effective throughput, representing over a 5-fold improvement over conventional uncoordinated systems under identical energy potential, thereby overcoming the receiver frame rate bottleneck.
Medical metallic implants can cause localized heating when exposed to electromagnetic fields. In this paper, the correlation between temperature increases and the spatial-averaged specific absorption rate (SAR) due to the implanted metal plates under the near-field exposure in the cellular frequency bands is presented. The numerical model consists of a realistic human model with implanted metal plates. We assumed the microwave exposure from the half-wavelength dipole antenna of both horizontal and vertical polarization. Numerical simulations based on the finite difference time-domain (FDTD) method and Pennes' bioheat transfer equation (BHTE), were performed to calculate the temperature increases and the spatial-averaged SAR, as well as the coefficient of determination. The simulation results revealed the effect of metallic implants on the correlation between temperature increase and SAR. Furthermore, the uncertainty of temperature increase caused by the different sizes of implants and frequency was estimated.
