Leveraging the spatiotemporal continuous aware ness of low Earth orbit satellites, space-based spectrum monitoring systems can construct a spectrum situation map (SSM) to facilitate spectrum surveillance and management in mobile cognitive communication networks. However, due to limited on orbit processing capabilities and unfavorable ground-to-satellite transmission environments, spectrum measurements will inevitably be incomplete and corrupted by anomalies. Existing SSM construction methods assume stationary spectral environments and do not consider the dynamic changes in spectrum situation. Considering spectrum state is evolving constantly over time, this paper proposes a robust time-aware online streaming tensor (R TAST) completion algorithm by exploiting the time-frequency space correlation and temporal properties in real-world spectral measurements. Based on compressed wideband sampling, the proposed R-TAST algorithm integrates rank estimation, anomaly removal, and dynamic spectral tensor completion to achieve spectrum situation completion and evolutionary prediction. Numerical analyses conducted on simulated and realistic ray tracing based datasets demonstrate the effectiveness and efficiency of the proposed R-TAST in comparison with state-of-the-art streaming tensor completion and prediction algorithms.

Mega-constellation networks (MCNs) of low Earth orbit (LEO) satellites are poised to serve as critical enablers for next-generation 6G wireless systems. These satellite infrastructures not only provide ubiquitous Internet access to terrestrial users but also facilitate relay-assisted data transmission for space-based remote sensing and positioning services. However, the inherent challenges of ubiquitous coverage and overlapping service regions in dense LEO constellations necessitate rigorous optimization of user-satellite association strategies, especially when the serving satellites are from multiple constellations. This paper provides insights into user access selection and resource optimization for mega-constellations that cover extensive terrestrial areas. The selection of multiple satellites from various constellations is predicated on the calculation of their coverage areas and the geolocation of urban areas. By explicitly modeling the transmission traffic requests and data collection process, the access strategy is investigated to maximize network throughput while maintaining a balance in quality-of-service (QoS). Thus, an optimization algorithm is proposed for user access selection that synergistically combines graph convolutional attention networks (GCAN) and deep reinforcement learning (DRL). Simulation results based on the Starlink Phase I and Phase IV models show that the proposed algorithm achieves performance improvements in both throughput and access quality compared to other benchmark algorithms.

Automatic Modulation Recognition (AMR) is essential for efficient spectrum utilization, cognitive radio, and secure wireless communications. However, deploying accurate AMR models on resource-constrained devices remains challenging due to substantial computational overhead. Importantly, minimizing inference computational cost—distinct from conventional neural network lightweighting—is critical for meeting strict runtime constraints on resource-limited platforms. This paper proposes the Adaptive Modulation Inference (AMI) framework, a dynamic inference solution for efficient and adaptive AMR on limitedresource platforms such as satellites and unmanned aerial vehicles (UAVs). AMI integrates Adaptive Input Skipping (AIS) and Budget-Efficient Exiting (BEE) to dynamically tailor computation based on signal difficulty and real-time resource budgets. AIS employs Layer and Channel Gates for coarse-grained skipping and fine-grained pruning, while BEE adjusts early-exit thresholds based on entropy and Top-1 confidence. Implemented on a lightweight 1D MobileNetV2 backbone, AMI achieves up to 56% reduction in average computational cost with less than 1% accuracy loss on both RML22 and HisarMod2019.1 datasets, outperforming existing dynamic inference strategies applied in the AMR domain.

Mega-constellation networks of low Earth orbit (LEO) satellites are poised to become an integral part of future 6G networks. Satellite infrastructures can provide Internet access services to terrestrial users and act as relay satellites for on-board remote sensing and positioning services. Given the extensive coverage of LEO satellite constellations and the overlapping coverage areas between satellites, the selection of appropriate access satellites for terrestrial users is critical. This paper provides insights into user access selection and resource optimization for large LEO constellations covering large terrestrial areas. By explicitly modeling the transmission traffic requests and data collection process, the access strategy is investigated to maximize network throughput while maintaining a balance on quality of service. Thus, an optimization algorithm is proposed for user access selection that synergistically combines graph convolutional attention networks and deep reinforcement learning (DRL). Simulation results based on the Starlink Phase I model show that our algorithm achieves performance improvements in both throughput and access quality compared to random access and standalone DRL frameworks.

Traditional signal recognition requires the design of multiple different deep neural networks to handle different signal recognition tasks, which not only fails to take into account the correlation among different subtasks, but also leads to large model size and higher computational complexity. In this work, we propose a multitask-learning-based multiband signal recognition network$(\text {M}^{2}\text {-Net})$to simultaneously recognize the location of occupied frequency bands, modulation types, and signal types. The proposed$\text {M}^{2}\text {-Net}$consists of two main parts: 1) shared feature extraction network (SFEN) and 2) multitask classification header (MCH). In SFEN, a plug-and-play multitask feature extraction convolution and an adaptive threshold denoising module are introduced to provide better shared feature extraction and denoising performance. In MCH, the shared features obtained from SFEN are further processed for different recognition tasks. Furthermore, during the multitask model training, homoscedastic uncertainty is introduced as a task-dependent weight to adaptively balance the training loss of different tasks. To evaluate the recognition performance of the proposed method, we construct a multiband signal dataset and compare$\text {M}^{2}\text {-Net}$with several state-of-the-art models in signal recognition field. Experiment results show that the proposed$\text {M}^{2}\text {-Net}$has significant performance improvements in terms of recognition accuracy and model complexity, especially under low signal-to-noise ratio conditions.

This paper presents GBSense, an innovative compressed spectrum sensing system designed for GHz-bandwidth signals in dynamic spectrum access (DSA) applications. GBSense introduces an efficient approach to periodic non-uniform sampling, capturing wideband signals using significantly lower sampling rates compared to traditional Nyquist sampling. By integrating time-interleaved analog-to-digital conversion, GBSense overcomes the hardware complexity typically associated with traditional multicoset sampling, providing precise, adjustable sampling patterns without the need for analog delay circuits. The system’s ability to process signals with a 2 GHz radio frequency bandwidth using only a 400 MHz average sampling rate enables more efficient spectrum monitoring and access in wideband cognitive radios. Lab tests demonstrate 100% accurate spectrum detection when the spectrum occupancy is below 100 MHz and over 80% accuracy for occupancy up to 200 MHz. Additionally, an integrated system utilizing a low-power Raspberry Pi processor achieves a low processing latency of around 30 ms per frame, demonstrating the system’s potential for DSA applications in next-generation wireless networks.

Wideband signal recognition is a crucial task in the cognitive wireless communication, involving accurate classification of different signal types, modulation types, center frequencies, etc. However, most conventional approaches treat the recognition of different parameters as multiple independent tasks, and often face performance bottlenecks due to the complexity and diversity of wideband signals. To overcome these challenges, we propose a multi-task learning (MTL) network that integrates multiple tasks of signal recognition into an end-to-end model to accomplish spectrum sensing, modulation recognition, and signal classification simultaneously. By employing a shared feature extraction network and a multi-task classification header, the proposed framework effectively captures the correlations and shared information among different tasks, thereby enhancing overall recognition performance. To validate the effectiveness of the proposed scheme, we compare its performance with other state-of-the-art recognition and classification networks. Experimental results demonstrate the significant performance of the proposed MTL network in spectrum sensing, modulation recognition, and signal classification tasks.

Semantic status update (SSU) communication is envisioned to provide semantic-aware and energy efficient semantic information (SI) delivery in future intelligent Internet of Things (IoT) applications. In this paper, we integrate the knowledge base (KB)-enabled semantic network into a discrete time Markov chain, and introduce a new metric in the SSU communication system, named semantic utility loss (SUL), which captures the timeliness and estimation accuracy of SI. The transmitter samples and extracts SI from the physical process, and sends the SSU. To combat semantic noise, the receiver can update KB at the cost of energy consumption to keep semantic match with the transmitter, i.e., inferring informative SI from received SSU. To minimize the weighted sum of SUL and overall energy cost incurred by transmitting SSU and updating KB, we formulate an infinite horizon average cost Markov decision process. We prove that the optimal joint transmission and updating (JTU) policy has a double threshold structure concerning SUL. Simulation results show the superiority of the proposed JTU policy over the zero-wait and sample-at-change baseline policies. In addition, we reveal that under the optimal JTU policy, the SSU communication framework outperforms the non-SSU framework in providing informative and energy efficient SI delivery.

In cognitive communication networks, the high frequency and wide bandwidth of millimeter-wave signals exacerbate in-phase/quadrature-phase (IQ) imbalance in the transceiver hardware, leading to mirror signal interference, which increases the false alarm probability of spectrum sensing and results in significant performance degradation in communication systems. However, traditional IQ imbalance estimation algorithms typically estimate IQ amplitude and phase imbalance separately, leading to increased hardware costs. To address this issue, we propose a convolutional block attention module (CBAM)-based algorithm to jointly estimate the IQ amplitude and phase imbalance parameters. By exploiting the correlation between IQ amplitude and phase imbalances, the proposed algorithm can not only further improve the accuracy of IQ amplitude and phase imbalance estimation at low signal-to-noise ratios, but also improve the estimation speed through joint estimation scheme. Simulation results demonstrate that the proposed algorithm has lower processing delay and higher estimation accuracy compared to traditional algorithms, and lower space complexity than other deep learning-based algorithms.

Automatic modulation recognition (AMR) is crucial for efficient modulation type recognition in modern wireless systems, but its performance suffers from low signal-to-noise ratios (SNR). This paper proposes an adaptive denoising automatic modulation recognition network (AD-AMR Net) to improve its performance, which incorporates an adaptive denoising module (ADM) to alleviate noise and a feature extraction module (FEM) to extract multi-scale features from the denoised signals. Com-pared to conventional AMR schemes, AD-AMR Net demonstrates better recognition accuracy under low SNRs, achieving 84.8% average accuracy in the 0 – 10 dB range versus 80 – 82% for conventional methods. Overall, the proposed model achieves 64.6% accuracy on the RadioML dataset, substantially outperforming previous deep learning techniques. Key innovations of adaptive denoising approach and multi-scale feature extraction enable major performance gains in low SNR scenarios.

Increased matrix dimensionality and shorter channel coherence time pose critical challenge to obtain channel state information (CSI) in millimeter-wave massive multiple-input multiple-output communication systems. The accuracy of CSI derived through beam training is often hampered by codebook design, while the CSI secured by channel estimation typically underutilizes the prior information of the channel matrix. To address these issues, we introduce a novel channel estimation algorithm that incorporates a weighted nuclear norm minimization approach which adopts fast wide-beam training to determine the weight factor. The simulation results demonstrate that the proposed method achieves more accurate estimation performance with reliable convergence when compared with traditional schemes.

In this paper, we focus on the optimal problem of average age of information (AAoI) in grant-free massive access, and propose an age-critical frameless ALOHA (ACFA) random access protocol, where the AAoI is implicitly reduced by banning the transmission of activated user equipments (UEs) recovered successfully in the last frame. In particular, we analyze the dense and sparse access models according to the activation probability, and present these scenarios with time-stamped sampling either at the beginning of the frame or in the first slot transmitting the packet. In order to qualify the AAoI of proposed protocol, we define two virtual rates and establish an iterative framework to analyze the access successful probability (ASP) of the protocol in asymptotic regime, and derive the closed-form expressions of AAoI as a function of ASP and virtual rate in all cases. Further, we formulate the optimal problems of normalized AAoI in all cases, and obtain the selection of access parameters by asymptotic analysis and simulations, respectively. Finally, we compare our protocol with state-of-the-art schemes, and the simulation results show that the ACFA random access protocol outperforms these benchmark schemes, and has great potential of access-banned policy in minimizing AAoI for frame-based protocols.

The goal of the next-generation mobile communication system is higher data-rates, lower latency, and higher energy-efficient performance, which bring about the demands for fast beam tracking in time-varying mobile communication. With the development of large-scale antenna array technology, highly directional beams can be formed with limited radio frequency chains. However, traditional exhaustive searching scheme has unacceptable overhead that leads to great challenges for applying to mobile millimeter-wave environments. Fast beam tracking scheme therefore has been recognized as a key technology in millimeter wave communication. To address this issue, this paper proposes a data-driven multi-armed beam tracking scheme to select the beamforming/combining vectors that achieve the target quality of service based on the real-time measurement, rather than the prior knowledge such as channel and user mobility information in beamforming design. To further speed up the beam tracking process, multi-armed beam is created to sample multiple spatial directions simultaneously. Simulation results show that the proposed data-driven multi-armed beam tracking method could achieve fast beam tracking performance with high resolution and reduced training overhead.

Compressive sensing (CS) techniques have been proposed for wideband spectrum sensing applications to achieve sub-Nyquist-rate sampling. The complexity of CS recovery algorithm and the detection performance against noise are two of the main challenges of the implementation of compressive spectrum sensing (CSS). We hereby propose CSS scheme based on orthogonal matching pursuit (OMP) with the aid of spectrum sparsity order estimation enabled by detecting the signal subspace dimensionality directly from sub- Nyquist measurements. The computational effort of spectrum recovery can be saved superlinearly with the reduction of iterations. With the estimated spectrum sparsity order, the OMP algorithm is proposed to run only an explicit and a fraction of iterations compared to the cases where such estimation is absent. Besides, the estimation of active channel number also enables blind and hard decision of channel occupancy where threshold adaption for energy detection is avoided. Moreover, the detection performance of the proposed CSS scheme by simulation shows superior robustness against noise compared to the energy detection scheme.

The increasing number of Internet of Things (IoT) objects has been a growing challenge of the current spectrum supply. To handle this issue, the IoT devices should have cognitive capabilities to access the unoccupied portion of the wideband spectrum. However, most IoT devices are difficult to perform wideband spectrum sensing using either conventional Nyquist sampling system or sub-Nyquist sampling system since both power-hungry sampling components and intricate sub-Nyquist sampling hardware are unrealistic in the power-constrained IoT paradigm. In this paper, we propose a blind joint sub-Nyquist sensing scheme by utilizing the surround IoT devices to jointly sample the spectrum based on the multicoset sampling theory. Thus, only the off-the-shelf low-rate analog-to-digital converters on the IoT devices are required to form coset samplers and only the minimum number of coset samplers are adopted without the prior knowledge of the number of occupied channels and signal-to-noise ratios. Moreover, to further reduce the number of coset samplers and transfer part of the computational burden from the IoT devices to the core network, we adopt the data from geo-location database when applicable. The experimental results on both simulated and real-world signals verify the theoretical results and effectiveness of the proposed scheme. At the meanwhile, it is shown that the adaptive number of coset samplers could be adopted without causing the degradation of the detection performance and the number of coset samplers could be further reduced with the assists from geolocation database even when the obtained information is partially correct.

Spectrum sensing is a proactive way in cognitive radio systems to achieve dynamic spectrum access; and compressive spectrum sensing (CSS) techniques alleviate the demand for high-speed sampling in wideband spectrum sensing. Most existing literature discusses Neyman-Pearson channel energy detection and threshold adaption schemes to achieve an optimal performance of detection in a conventional non-compressive spectrum sensing scenario. However, in the CSS, it is found that the channel energy statistics and optimal threshold depend not only on noise energy but also on compression ratio, sparsity of spectrum, and nature of recovery algorithms. To investigate the channel energy statistics of recovered spectrum, we postulate a statistical model of channel energy for CSS and propose a learning algorithm based on a mixture model and expectation-maximization techniques. In addition, having verified the validity of the postulated model, we propose a practical threshold adaption scheme for CSS aiming to maintain constant false alarm rates in channel energy detection. In simulations, it is shown that the postulated channel energy statistic models with parameters learned by the proposed learning algorithm fit well with empirical distributions under circumstances of various channel models and recovery algorithms. Moreover, it is presented that the proposed threshold adaption scheme maintains the false alarm rate near the predefined constant, which in turn validates the postulated model.

The increasing number of Internet of things (IoT) objects has been a growing challenge of the current spectrum supply. To handle this issue, the IoT devices should have cognitive capabilities to detect and access the unoccupied portion of the wideband spectrum. However, most IoT devices are difficult to perform wideband spectrum sensing using either conventional Nyquist sampling system or sub-Nyquist sampling system since both the power-hungry sampling components and specialized sub-Nyquist sampling hardware are unrealistic in the power-constrained IoT paradigm. In this paper, we propose a blind sub-Nyquist sensing scheme by utilizing the surround IoT devices to jointly sample the spectrum based on the multi-coset sampling theory. Thus, only the off-the-shelf low- rate analog-to-digital converters (ADCs) on the IoT devices are required to form coset samplers and only the minimum number of coset samplers are adopted without the prior knowledge of the number of occupied channels. The experimental results on both the simulated and real-time signals verify the theoretical results and the effectiveness of the proposed scheme. At the meanwhile, it is shown that the adaptive number of coset samplers could be adopted without causing the degradation of the detection performance.

To maximize spectrum access opportunities for white space devices, incorporating real-time spectrum sensing with geolocation database is a promising approach to enhance detection resolution. Advanced spectrum sensing techniques are needed to quickly and accurately identify spectrum occupancy over a wide frequency range. However, the stringent requirements from the wideband signal acquisition and processing pose a major implementation challenge in compact devices with limited energy storage and computational capabilities. In this paper, an efficient joint scheme of sub-Nyquist wideband spectrum sensing with geolocation database is proposed to ensure accurate detection of the surrounding spectrum with reduced number of measurements. Subspace augmented greedy algorithm is modified to incorporate a priori information from geolocation database, therefore enabling local spectrum sensing to be performed only on a limited number of potentially vacant channels over TV White Space. Theoretical analysis and experimental results on the simulated and real-time signals show that the proposed joint scheme improves the sensing sensitivity with lower computation complexity, while the updated channel information from local sensing reduces the risk of interferences to the primary users.

To enable dynamic spectrum access of machine-to- machine (M2M) communications for internet of thigns (IoT) applications, incorporating real-time wideband spectrum sensing with geo- location database is a promising approach to reducing complexity and promoting low latency. However, high sampling rate and high speed signal processing of the wideband signal acquisition pose a major implementation challenge in compact M2M/IoT devices with limited power supply and low computation capabilities. In this paper, to reduce the required sampling rate without degrading the detection performance, a hybrid sub- Nyquist wideband spectrum sensing scheme with geo-location database is proposed. The iteratively reweighted least squares (IRLS) algorithm is modified to incorporate a priori information from geo-location database, therefore enabling local spectrum sensing to be performed only on a limited number of potentially vacant TV channels. Theoretical analyses and experimental results show that the proposed hybrid scheme improves detec- tion performance with reduced sub-Nyquist sampling ratio and demonstrates a better detection capability over lower signal-to- noise ratio (SNR) regions.

Nowadays, wideband spectrum sensing, as one of the vital technologies of cognitive radio (CR), has the potential to find more temporarily available frequency bands to meet the growing demands of wireless services. As the vast number of samples are required to be collected and processed, traditional wideband spectrum sensing methods become inefficient and cause large energy consumption. Therefore, many theoretical work focus on applying compressive sensing (CS) into wideband spectrum sensing to alleviate this issue. In this paper, to verify the CS-based spectrum sensing scheme in real-world scenarios, a real-time compressive spectrum sensing testbed is proposed to process the real-time data collected from the TV white space (TVWS) spectrum. The proposed testbed consists of two parts: a senor node, and a real-time signal processing platform based on National Instruments (NI) LabVIEW software to process the spectral data and control the sensor.

Nowadays, people usually connect to the Internet through a multitude of different devices. Video streaming takes the lion's share of the bandwidth, and represents the real challenge for the service providers and for the research community. At the same time, most of the connections come from indoor, where Wi-Fi already experiences congestion and coverage holes, directly translating into a poor experience for the user. A possible relief comes from the TV white space (TVWS) networks, which can enhance the communication range thanks to sub-GHz frequencies and favorable propagation characteristics, but offer slower datarates compared with other 802.11 protocols. In this paper, we show the benefits that TVWS networks can bring to the end user, and we present CABA, a connection aware balancing algorithm able to exploit multiple radio connections in the favor of a better user experience. Our experimental results indicate that the TVWS network can effectively provide a wider communication range, but a load balancing middleware between the available connections on the device must be used to achieve better performance. We conclude this paper by presenting real data coming from field trials in which we streamed an MPEG dynamic adaptive streaming over HTTP video over TVWS and Wi-Fi. Practical quantitative results on the achievable quality of experience for the end user are then reported. Our results show that balancing the load between Wi-Fi and TVWS can provide a higher playback quality (up to 15% of average quality index) in scenarios in which the Wi-Fi is received at a low strength.

Wideband spectrum sensing is regarded as one of the key features in cognitive radio systems. Compressive sensing (CS) has recently become one of the promising techniques to deal with the Nyquist sampling rate bottleneck of wideband spectrum sensing. Theoretical analyses and simulation have shown that CS could achieve high detection probability and low false alarm for wideband spectrum sensing. However, implementation of CS on the real-time signals and real-time processing poses significant challenges due to the iterative nature of the CS algorithms. In this paper, we propose a novel adaptively regularized iterative reweighted least squares (AR-IRLS) algorithm to implement the real-time signal recovery on the CS based wideband spectrum sensing. The proposed algorithm moves estimated solutions along an exponential-linear path by regularizing weights with a series of non- increasing penalty terms, which significantly speeds up the convergence of reconstruction and provides high fidelity guarantee to cope with the varying bandwidths and power levels of occupied channels. The proposed algorithm presents robustness against different sparsity levels at low compressive ratio without degradation on the reconstruction performance, and is tested on the real-time signals over TV white space spectrum after having been validated on the simulated signals. Both the simulation and real-time experiments show that the proposed algorithm outperforms the conventional iterative reweighted least squares (IRLS) algorithms in terms of convergence speed, reconstruction accuracy, and compressive ratio requirement.

Operating at ultra-high frequency (UHF), TV white space (TVWS) can achieve long-distance communication and good in-building penetration, and has attracted increasing attention of regulators, researchers and stakeholders. This paper explores the potential of TVWS for network provisioning within a cluster of buildings, through a succession of tests. Different transmission distances, from 10m to over 120m, and through multiple layers of walls as well as complex transmission environment imposed by other factors like office and construction facilities, are considered. Further, a compact ultra-wide band (UWB) printed monopole antenna is designed for the client white space terminal, and compared with a commercial directional UHF antenna on the same client. Measurement results show that the in-house compact antenna achieves fast network speed and a high signal-to-interference-plus-noise ratio (SINR), and it is orientation independent.