Wi-Fi-based gesture recognition faces significant performance degradation in cross-domain scenarios due to the distribution shifts between various environments, user locations, and facing orientations. To address this challenge, we propose WiSACL, a novel Wi-Fi-based gesture recognition framework that integrates innovative CSI signal processing with advanced subdomain adaptation techniques. We propose a novel Wi-Fi CSI phase difference representation that suppresses the environmental noise and highlights the motion features via phase ratio computation, wavelet denoising, and temporal differencing, which are then encoded into discriminative image representations for enhanced gesture recognition. Building upon this robust signal representation, we develop an Adaptive Distributioncalibrated Pseudo-Label (ADPL) module that progressively refines target labels through subdomain distribution alignment and confidence-aware selection. To fully exploit these pseudo-labels while mitigating the adverse effects of domain shift and label noise, we further introduce a contrastive learning (CL) model that explicitly enhances intra-class compactness and inter-class separation in the feature space. Extensive experiments on the Widar3.0 dataset demonstrate that WiSACL achieves an average accuracy of 98.62% across cross-location, cross-orientation, and cross-environment scenarios, significantly outperforming state-of-the-art methods and validating the effectiveness of our approach for robust cross-domain wireless gesture recognition.

A novel mobile edge generation (MEG) framework is proposed to support latency-sensitive generation tasks in the social-aware device-to-device (D2D) underlaying cellular network. Within this framework, user devices (UDs) and base station (BS) are organized into socially cohesive communities based on content preference and spatial proximity, enabling cooperative generation tasks via both cellular and intra-community D2D communications. A joint seed-and-content based BS-D2D (JSCB) transmission protocol is proposed to dynamically orchestrate the transmission mode between seed acquisition with local generation and direct content sharing across multiple consecutive task rounds, incorporating the spillover mechanism for handling overdue transmissions. Based on this protocol, an average task delay minimization problem is formulated to jointly optimize the UD association between cellular and D2D communication, transmission mode, D2D pairing, and BS-side beamforming. To efficiently solve the hybrid and temporally coupled problem, a joint matching and proximal policy optimization (JMPPO) algorithm is developed, where the discrete and continuous actions are decoupled with specialized modules though a hierarchical deep reinforcement learning and matching design. Numerical results validate that 1) the JSCB protocol reduces delay through adaptive transmission scheduling and cellular/D2D coordination; 2) the JMPPO algorithm outperforms both learning-based and traditional baselines in terms of average delay under the spillover and hybrid action scenarios; 3) the proposed schemes demonstrate robustness across diverse network and system conditions.

Unsupervised visual representation learning has gained significant attention in the computer vision community, driven by recent advancements in contrastive learning. Most existing contrastive learning frameworks rely on instance discrimination as a pretext task, treating each instance as a distinct category. However, this often leads to intra-class collision in a large latent space, compromising the quality of learned representations. To address this issue, we propose a novel contrastive learning method that utilizes randomly generated supervision signals. Our framework incorporates two projection heads: one handles conventional classification tasks, while the other employs a random algorithm to generate fixed-length vectors representing different classes. The second head executes a supervised contrastive learning task based on these vectors, effectively clustering instances of the same class and increasing the separation between different classes. Our method, Contrastive Learning via Randomly Generated Supervision(CLRGS), significantly improves the quality of feature representations across various datasets and achieves state-of-the-art performance in contrastive learning tasks.

WiFi-based Human Activity Recognition (HAR) is a promising approach for non-invasive, device-free monitoring of user movements, particularly in smart home and healthcare applications. However, existing studies have mainly focused on single-user scenarios, which limits practicality in real-world environments and makes them vulnerable to performance degradation from wireless channel noise and interference. We propose WiFi Multi-user Human Activity Recognition (WiMAR), a system that separates the dynamic component of Channel State Information (CSI) induced by human activities from the static component caused by the environment. Our novel hybrid CNNRABGRUM framework combines Convolutional Neural Networks with Residual Structures (CNNR) and context-aware Bidirectional Gated Recurrent Units with an Attention mechanism and Multilayer Perceptrons (ABGRUM) to extract discriminative features from the dynamic CSI component. Experimental results show that WiMAR achieves a 2.9% increase in average recognition accuracy compared to state-of-the-art models, along with accelerated training convergence and enhanced stability, demonstrating robust performance in complex multi-user environments.1

By analyzing changes in Channel State Information (CSI) during the propagation of Wi-Fi signals in indoor environ-ments, researchers can achieve contactless perception of human behaviors. Due to the high complexity of the deep models used in CSI-based human activity recognition systems, it's difficult to be employed in edge computing devices. To improve the Wi-Fi perception accuracy and reduce the system complexity, this paper proposes a lightweight human behavior perception system LiteWiHAR. The system converts the filtered CSI signal in time domain into a two-dimensional image to increase its spatial structure information. Using ConvNeXt v2 layered encoder as the backbone feature extraction network, perceptual information containing deep and shallow features in the image is extracted layer by layer. By reducing the number of depthwise separable convolution channels and compressing the number of stacked blocks in the encoder, the degree of network fragmentation is reduced and the number of model parameters is compressed. Finally, the SimAM parameterless attention module is introduced to assist the network in capturing global information. The system is lightweight while maintaining good feature extraction capability. We validated the effectiveness of LiteWiHAR on three open-source datasets and demonstrated its superiority over other state-of-the-art systems.

A simultaneously transmitting and reflecting reconfigurable intelligent surface (STAR-RIS) assisted multiuser near-field wideband communication system is investigated, in which a robust deep reinforcement learning (DRL) based algorithm is proposed to enhance the users’ achievable rate by jointly optimizing the active beamforming at the base station (BS) and passive beamforming at the STAR-RIS. To mitigate the beam split issue, the delay-phase hybrid precoding structure is introduced to facilitate wideband beamforming. Considering the coupled nature of the STAR-RIS phase-shift model, the passive beamforming design is formulated as a problem of hybrid continuous and discrete phase-shift control, and the proposed algorithm controls the high-dimensional continuous action through hybrid action mapping. Additionally, to address the issue of biased estimation encountered by existing DRL algorithms, a softmax operator is introduced into the algorithm to mitigate this bias. Simulation results illustrate that the proposed algorithm outperforms existing algorithms and overcomes the issues of overestimation and underestimation.

An adaptive next generation multiple access (NGMA) downlink scheme is provided, where non-orthogonal multiple access (NOMA) and space division multiple access (SDMA) users are served with the same orthogonal time and frequency resource to address the energy constraints and massive connectivity issues of Internet-of-Things networks. Based on this scheme, the long-term power-constrained sum rate maximization problem is investigated, where beamforming, power allocation, and user clustering are jointly optimized, subject to a long-term total power constraint. To solve the formulated problem, a spatial correlation-based user clustering approach is proposed and a resource allocation algorithm is designed based on the trust region policy optimization (TRPO) algorithm, which demonstrates stable convergence under large learning rates. Numerical results verify that the sum rate of the proposed NGMA scheme outperforms the conventional NOMA and SDMA schemes. Moreover, the spatial correlation-based clustering algorithm achieves an increasing sum rate gain compared to the channel correlation-based baseline algorithm as the spatial correlation in the channel model increases.

A reconfigurable intelligent surface (RIS)-assisted downlink nonorthogonal multiple access (NOMA) Internet of Things (IoT) network is proposed, where a Quality-of-Service (QoS)-based NOMA clustering scheme is conceived to effectively utilize the limited wireless resources among IoT devices. A throughput maximization problem is formulated by jointly optimizing the phase shifts of the RIS and the power allocation of the base station (BS) from the short-term and long-term perspectives. We aim to investigate and compare the performance of deep learning (DL) and deep reinforcement learning (DRL) algorithms for solving the formulated problems. In particular, the DL method utilizes model-agnostic-metalearning (MAML) to enhance the generalization capability of the neural network and to accelerate the convergence rate. For the DRL method, the deep deterministic policy gradient (DDPG) algorithm is employed to incorporate continuous phase-shift variables. It shows that the DL method only focuses on the maximization of the instantaneous throughput, whereas the DRL method can coordinate the power consumption over different time slots to maximize the long-term throughput. Numerical results demonstrate that: 1) the proposed QoS-based NOMA clustering scheme achieves higher IoT throughput than the conventional channel-based scheme; 2) the implementation of RISs induces approximately 5%–25% throughput gain as the number of RIS elements increases from 8 to 64; 3) DL and DRL achieve a similar throughput performance for the short-term optimization, while DRL is superior for the long-term optimization, especially when the total transmit power is limited.

Expected high demand of user applications in the WLAN is a driver for WLANs to share radio resources more efficiently. The move to 802.11be with OFDMA and MU-MIMO makes Radio Resource Management (RRM) a multi-dimensional problem in a complex wireless environment. Traditionally, the way that an RRM problem is formulated always leads to either a large state space or action space, which makes reinforcement learning impossible to be applied. In this paper, we propose a Sequential State Q-learning algorithm (SSQL) aimed at solving the Resource Unit (RU) allocation for scheduled uplink transmission to maximize system bitrate in a multi-AP 802.11be OFDMA network. The AP acts as the agent with the serving stations as ‘states’ and their RU allocations as ‘actions’. The AP observes the wireless environment, continuously refreshing the Q-values of the state-action pairs and outputs the RU allocation to optimize the objective. Through simulations, we demonstrated that the performance of SSQL is 89.67% of the global optimal with very fast convergence, which makes it more practical for use in varying wireless networks.

In this paper we evaluate the game-theory approach we have used for radio resource block allocation in sliced multi-cell networks with distributed control against a centralized control strategy. This is done in the context of an approach that allocates resources according to the QoS requirements of users. The objective is to see whether the game-theory approach is worth using throughout a network with mixed decentralized and centralized control and the conclusion is that the performance of the game-theory approach is equally suitable for both types.

In this paper, we demonstrate how isolation can be achieved in dynamic network slicing using an appropriate Connection Admission Control (CAC) mechanism. The scenario adopted is a multi-cell OFDMA network and the resource allocation is performed using a non-cooperative game. Simulation results show that such an approach leads to implicit isolation (a term we define in the paper) being achieved, at the same time reaping the capacity benefits of dynamic slicing.

In this paper, we use a Spectrum Sharing/allocation algorithm to implement a Cognitive Radio network without spectral awareness. This allows radio resources to be allocated to primary users and to secondary users in such a way that primary system and secondary system are aware of each other. By doing this, sensing of spectrum holes is avoided but the primary users will be protected and will always be allocated sufficient bandwidth to meet their QoS requirements. However, secondary users also have the concept of QoS and only those that can be given sufficient resources are actually allocated any, so avoiding wasting resources on users that cannot meet their QoS requirements. The scenario adopted is a multi-cell OFDMA network and the resource allocation is formulated as a non-cooperative game: each cell is a player whose aim is to maximise the number of secondary users that can reach their bitrates subject to the premise that all primary users in that cell can achieve their bitrate. Once spectrum is allocated we implement a power-minimisation algorithm to minimise the power consumption in the network while maintaining the number of users meeting their QoS requirements. Simulation results show the benefit of the approach.

Location based services are the hottest applications on mobile devices nowadays and the growth is continuing. Indoor wireless positioning is the key technology to enable location based services to work well indoors, where GPS normally could not work. Bluetooth has been widely used in mobile devices like phone, PAD etc. therefore Bluetooth based indoor positioning has great market potential. Radio Signal Strength (RSS) is a key parameter for wireless positioning. New Bluetooth standard (since version 2.1) enables RSS to be discovered without time consuming pre-connection. In this research, general wireless positioning technologies are firstly analysed. Then RSS based Bluetooth positioning using the new feature is studied. The mathematical model is established to analyse the relation between RSS and the distance between two Bluetooth devices. Three distance-based algorithms are used for Bluetooth positioning: Least Square Estimation, Three-border and Centroid Method. Comparison results are analysed and the ways to improve the positioning accuracy are discussed.