This work proposes two approaches, i.e., Fed-KD and Fed-TKD, to enhance communication efficiency with federated learning (FL) in Industrial Internet of Things (IIoT) radar sensing systems, which face the challenge in transmitting large volumes of sensitive data while still ensuring privacy. Fed-KD leverages knowledge distillation to transfer knowledge from complex teacher networks to simpler student models in order to reduce communication overhead in bandwidth-constrained environments. Fed-TKD improves communication efficiency further by applying tensor decomposition to reduce parameter redundancy. Experimental results using an industrial IoT radar imagery dataset show that both methods can significantly reduce communication costs while maintaining a high model accuracy, making them especially suitable for privacy-preserving industrial IoT sensing applications. Furthermore, experimental results with both IID and non-IID data distributions confirm the robustness of the proposed methods in heterogeneous environments.

Semantic communication (SemCom) is a crucial direction for future wireless communication, aiming to achieve efficient and low-redundancy communication by extracting and transmitting semantic information. It has significant application value in emerging scenarios with limited bandwidth and intensive data, such as the Low-Altitude Internet of Things (LA-IoT). However, existing SemCom frameworks face limitations in semantic compression and generalization. To address these challenges, this paper proposes a knowledge graph-enhanced SemCom system. The proposed system leverages a shared knowledge graph (KG) and deep learning models to extract semantic information from the input text, represented in the form of entity–relation–entity triples. By aligning these triples with existing knowledge in the KG, the system distinguishes between common semantics and unique semantics. Common semantics are reconstructed at the receiver using the KG, whereas unique semantics are compressed and transmitted through the channel and then fused with the shared knowledge to reconstruct the original text. Furthermore, considering the structured nature of triple-based information, we propose a structured attention mechanism to enhance transmission accuracy. Experimental results demonstrate that the proposed system achieves a BLEU score exceeding 0.9 and a METEOR score exceeding 0.93 under high SNR conditions, outperforming baseline methods in semantic fidelity while significantly reducing redundancy.

Integrated Sensing and Communication (ISAC) technology provides robust support for the development of the Internet of Things (IoT) by leveraging its powerful sensing and communication capabilities. Sensing-assisted beam prediction techniques effectively enhance the communication quality and efficiency of ISAC systems, which is crucial for achieving high-speed and stable communication among IoT devices. However, ensuring high accuracy in beam prediction typically relies on traditional centralized architectures, which intrinsically pose risks to privacy and security when the central server is compromised. Therefore, this inherent trade-off between high prediction accuracy and data security poses significant challenges in privacy-sensitive IoT deployments. To resolve this fundamental contradiction, we propose a federated CNN-Transformer method for distributed sensing-assisted beam prediction in ISAC systems. Specifically, We propose an improved federated learning (FL) framework where clients transmit not only gradients but also nonlinear features to the server for subsequent computations through task offloading while enforcing client data privacy security. Addressing this problem is particularly challenging due to scattering and noise issues in the propagation of radar sensing signals. To address this, we design a hybrid model that integrates CNN and Transformer for the client side, which effectively captures both local and global features of the sensing signals, thereby improving prediction accuracy. Experimental results demonstrate that, compared to traditional centralized methods, our proposed method not only achieves significant improvements in prediction accuracy but also offers unique advantages in terms of data privacy and security, mitigating the risk of data leakage.

A novel movable-element enabled simultaneously transmitting and reflecting surface (ME-STARS) communication system is proposed, where ME-STARS elements positions can be adjusted to enhance the degress-of-freedom for transmission and reflection. For each ME-STARS operating protocols, namely energy-splitting (ES), mode switching (MS), and time switching (TS), a weighted sum rate (WSR) maximization problem is formulated to jointly optimize the active beamforming at the base station (BS) as well as the elements positions and passive beamforming at the ME-STARS. An alternative optimization (AO)-based iterative algorithm is developed to decompose the original non-convex problem into three subproblems. Specifically, the gradient descent algorithm is employed for solving the ME-STARS element position optimization subproblem, and the weighted minimum mean square error and the successive convex approximation methods are invoked for solving the active and passive beamforming subproblems, respectively. It is further demonstrated that the proposed AO algorithm for ES can be extended to solve the problems for MS and TS. Numerical results unveil that: 1) the ME-STARS can significantly improve the WSR compared to the STARS with fixed position elements and the conventional reconfigurable intelligent surface with movable elements, thanks to the extra spatial-domain diversity and the higher flexibility in beamforming; and 2) the performance gain of ME-STARS is significant in the scenarios with larger number of users or more scatterers.

The large number of antennas in extremely large aperture array (ELAA) systems shifts the propagation regime of signals in wireless communication systems towards near-field spherical wave propagation, from beamforming to beamfocusing. The design of the two-dimensional beam codebook that contains both the angular and distance domains is challenging. To address this issue, we propose a reconfigurable intelligent surface (RIS)-assisted near-field beam training based on a novel multi-attention algorithm, which provides a fine-grained codebook with enhanced spatial resolution. Specifically, we transform the beam selection task into a location detection process, enabling more effective beam search. Experimental results unveil that the proposed method achieves beam selection accuracy up to 97% at signal-to-noise ratio (SNR) of 20 dB, and improves 10% over the baseline method at different SNRs.

The next generation mobile communication systems will experience huge transformation for multiple applications, which can provide pervasive intelligence and release more network resources for multiple terrestrial users. Moreover, digital twin (DT) technique helps each terrestrial user enable the mapping from physical world to digital space for the sake of reducing transmission latency. Thus, the integration between the terrestrial network and DT can expedite the computation offloading. Nevertheless, time-varying channel gains and dynamic UAV locations severely hinder better quality of service. In this paper, we envision a UAV-DT integrated task scheduling model to maximize the processed number of bits while minimizing the privacy protection overhead, which can further reduce the channel interference. Based on long-term task queues, we present a Lyapunov stability theory based multi-agent federated reinforcement learning (MAFRL) algorithm to optimize the CPU cycle frequency, transmission power and block size, which facilitate the integration between the communication, computation and block resources. Furthermore, we propose a blockchain-based verification mechanism to strengthen the privacy protection, and then demonstrate the performance upper bounds in terms of convergent task queues. Finally, massive simulation results show that the proposed MAFRL framework has approximately 1.7% performance gains in terms of the processed number of bits compared with state-of-the-art baseline methods.

The massive number of antennas in extremely large aperture array (ELAA) systems shifts the propagation regime of signals in internet of things (IoT) communication systems towards near-field spherical wave propagation. We propose a reconfigurable intelligent surfaces (RIS)-assisted beamfocusing mechanism, where the design of the two-dimensional beam codebook that contains both the angular and distance domains is challenging. To address this issue, we introduce a novel Transformer-based two-stage beam training algorithm, which includes the coarse and fine search phases. The proposed mechanism provides a fine-grained codebook with enhanced spatial resolution, enabling precise beamfocusing. Specifically, in the first stage, the beam training is performed to estimate the approximate location of the device by using a simple codebook, determining whether it is within the beamfocusing range (BFR) or the none-beamfocusing range (NBFR). In the second stage, by using a more precise codebook, a fine-grained beam search strategy is conducted. Experimental results unveil that the precision of the RIS-assisted beamfocusing is greatly improved. The proposed method achieves beam selection accuracy up to 97% at signal-to-noise ratio (SNR) of 20 dB, and improves 10% to 50% over the baseline method at different SNRs.

With the rapid development of cloud-edge–terminal (CET) technology, ubiquitous sensing devices are able to collaborate with edge terminals, enabling real-time, intelligent environmental awareness. For device-free sensing systems, the number of each human gesture category may vary (class imbalance), which makes previously distributed device-free sensing algorithms ineffective. In this article, we propose a novel monitoring scheme for device-free human action sensing for CET collaboration under class-imbalance conditions. Specifically, the body-coordinated velocity profile (BVP) features of wireless fidelity (WiFi) signals are used to detect human actions. To recognize human gestures, we develop a convolutional neural network (CNN) using a monitor to detect gradient changes under class imbalance. To mitigate the effects of class imbalance, a corresponding correction is applied to the loss function. To validate the effectiveness of the proposed method, we conduct numerical experiments under class-imbalance conditions. Different parameter settings and proportions of participating nodes are explored for their effects on experimental results. Additionally, numerical experiment results demonstrate that the proposed method improves recognition accuracy by 3.85%–34.1% compared to baseline algorithms. Overall, the proposed method addresses the challenge of distributed device-free sensing under class-imbalance conditions and achieves superior recognition accuracy performance.

To solve the problems of limited computing resources and data privacy in the IoE scenario of 6G networks, this paper propose an efficient transmission and secure sharing architecture of sensing data based on federated learning. The architecture considers an integrated sensing and communication (ISAC) approach, employs knowledge distillation techniques to compress and accelerate data processing models, and implements data communication technology based on airborne computing aggregation to reduce data transmission delays and improve the efficiency of data communication and computation among nodes. To address the challenge of data sharing for largescale heterogeneous network nodes in the integrated scenario, this paper adopts a sample expansion technology of distributed remote sensing data based on WGAN-GP to address the issue of insufficient data, and considers blockchain encryption technology to protect data privacy, thus promoting progress in data privacy sharing under distributed ISAC conditions and facilitating the construction of the 6G communication network.

Efficient and green transmission of communication and sensing (C&S) signals is a vital problem in Internet of Things (IoT) networks. In this article, we propose a variational autoencoder (VAE)-empowered deep learning (DL) network to fuse and reconstruct the integrated C&S signals. Specifically, we present a convolutional neural network to fuse the input communication data and SAR images into a combined representation, which can then be transmitted to other nodes in space–air–ground–ocean-integrated IoT networks. Instead of directly transmitting C&S data, the transmission of a fused feature vector can greatly save network resources and reduce network burden. Then, a mirrored deconvolutional network is constructed to recover C&S data from the transmitted feature representation. An end-to-end unsupervised training strategy is considered to train the proposed DL network without any label information and human labor. Qualitative and quantitative experiments demonstrate the feasibility of our proposed approach for transmitting and reconstructing integrated C&S signals. Further analysis on hyperparameter sensitivity and loss functions verifies the necessity and efficiency of the components in the proposed DL model. Note that the proposed efficient fusion and reconstruction schemes for C&S signals may provide the convenience to information sharing under the distributed scenarios.

Device-free gesture recognition (DFGR) is a critical technology for human-computer interaction and can be used for applications such as smart homes, and virtual reality in Wi-Fi sensing. Existing deep learning-based DFGR techniques typically require a large amount of labeled sensing data and are sensitive to the scene, which limits the development of ubiquitous sensing. In this study, in order to achieve device-free cross-scene gesture recognition from limited sensing samples, we consider the use of a small amount of Wi-Fi channel status information (CSI) data that are continuously obtained from low-cost commercial Wi-Fi devices. To this end, a few-shot learning-based cross-scene DFGR model is proposed for capturing highly discriminative information from dynamic CSI sequences. This information is then used to distinguish different gestures in the limited samples. Our experimental results using Wi-Fi signal collected at real world show that our model is able to realize 99.52% accuracy and can work well even with only one-piece data of new scene.

Gesture recognition is the central enabler of human-computer interaction (HCI). In addition to the semantic information contained in gestures, gesture-based user identification can effortlessly enhance HCI system security. Recently, the Wi-Fi-integrated sensing and communication (ISAC) technology has shown great potential in a field hitherto occupied by computer vision and radar sensing. In this work, leveraging Wi-Fi sensing, we propose a system called WiGesID that achieves joint gesture recognition and human identification (JGRHI). The basic idea behind WiGesID is to identify personalized spatiotemporal dynamic patterns from the gestures of different users. Moreover, we develop an effective approach to recognize new categories of gestures and users by computing relation scores between the features of the new category samples and the support samples. To evaluate the performance, we implemented WiGesID and conducted extensive experiments. The results demonstrate that our system outperforms the state-of-the-art method for cross-domain sensing and accurately recognizes new categories, which promotes the use of this application of Wi-Fi sensing in HCI.