In this paper, for mmWave cell-free massive MIMO systems, an energy efficiency optimization scheme that integrates a dynamic-grouped-adaptive (DGA) connected hybrid precoding at access points (APs) with deep reinforcement learning (DRL) is proposed. The proposed DGA-connected architecture divides antennas into groups and supports adaptive activation of radio frequency chains and phase shifters using switches. To further address the dynamic collaborative transmission challenges among APs with this architecture, we propose an action-embedded multi-agent soft actor-critic (AE-MASAC) algorithm incorporating a Transformer-based multi-head attention mechanism to facilitate collaborative behavior among AP agents. The simulation results show that the DGA-connected architecture can significantly reduce hardware power consumption compared to fully-connected and sub-connected structures. The AE-MASAC algorithm has advantages in convergence speed and scalability over traditional centralized DRL and distributed MADRL methods, providing a general framework for high-dimensional joint optimization problems in dynamic wireless communications systems.
The demand for increased spectral efficiency (SE) in downlink multi-user multiple-input multiple-output (MU-MIMO) systems has led to the development of various multiple access techniques. Among these, rate-splitting multiple access (RSMA) can improve SE over space division multiple access (SDMA) and non-orthogonal multiple access (NOMA) by splitting each user’s data into common and private parts. The common parts from all users are then encoded into a single common message, which is transmitted alongside the private messages using an RSMA-specific precoder. Recent results showed that the precoder can be modified to enable physical layer security (PLS) by integrating secrecy rate constraints into the design. However, this approach only secures the private messages, while the common part of each user remains unprotected. To address this limitation, we propose a hybrid-PLS RSMA scheme that symmetrically encrypts all user data and embeds key information within the PLS-secured private messages. Additionally, the RSMA precoder is adapted to ensure that the private messages are sufficiently sized to carry the key information. The precoder design is formulated as a non-convex SE maximization problem and solved using successive convex approximation (SCA). Analytical and numerical evaluations demonstrate that the proposed hybrid-PLS RSMA can significantly improve SE compared to partially secure RSMA and secure SDMA, when complete user data security is required.
The fluid antenna system (FAS) enhances the reliability of wireless communication due to its flexibility, which is attributed to its numerous ports. This flexibility can be leveraged to mitigate fading at the receiver side. In this paper, we extend the current research on FAS with maximum ratio transmission (MRT), referred to as MRT-FAS. Specifically, within a cell-free network, we evaluate the performance of the the user equipped with a single FAS, which receives MRT-precoded signals transmitted from multiple access points (APs). We first derive the distribution of the received signal power for the MRT-FAS user when multiple-input signals are precoded using MRT, considering both path-loss and small-scale fading. Subsequently, both the joint probability density function and cumulative distribution function are given. Based on these derivations, we express the outage probability of the FAS user with an integral-form expression. Numerical results reveal that the use of a single fluid antenna on the receiver end significantly enhances its performance, while simultaneously reducing the number of fixed antennas required at each AP.
Quantum cryptography has revolutionized secure communication by harnessing fundamental quantum mechanics principles to ensure unprecedented information-theoretic security. Among existing protocols, the standard three-stage quantum cryptographic protocol supports direct bidirectional exchange without needing pre-shared keys and inherent eavesdropping detection. However, scaling such protocols to multi-receiver (one-to-many) communication scenarios faces significant issues in terms of time complexity and QBER rate, as a single sender must sequentially transmit to each of N receivers, which leads to linear (O(N)) complexity and 3N transmission steps. This paper proposes a scalable multi-receiver quantum secure direct communication (QSDC) scheme that optimizes the three-stage protocol by implementing recursive, parallel transmission. In the proposed protocol, selected receivers can be turned into a temporary sender in the successive rounds, dramatically improving scalability by reducing required transmission steps from O(N) to O(log4N). This recursive approach substantially reduces communication overhead and requires moderate additional resources (approximately N/2 additional senders and N/4 receivers for propagation). Additionally, this protocol preserves the inherent security features of the original protocol, including intrinsic eavesdropping detection and resilience to interception. Our results demonstrate the protocol’s efficiency and resilience in a large-scale quantum network. The proposed protocol showcases a practical approach for implementing efficient, resilient, and scalable QSDC in next-generation quantum networks.
In this paper, we study a wideband multi-user terahertz (THz) communication system, where a hybrid intelligent reflecting surface (HIRS) comprising both active and passive reflecting elements is deployed to assist this system. With true-time-delay (TTD) devices being deployed, the HIRS can alleviate the beam squint effect while balancing system performance and hardware expenditure. We intend to maximize the sum-rate of considered system by jointly optimizing transmit beamforming and HIRS configuration, which is a challenging task. Via relaxation and alternating optimization (AO) framework, we develop an algorithm to tackle this problem. Simulation results demonstrate the effectiveness of our proposed solution.
This paper investigates the performance of a full-duplex (FD) integrated sensing and communication (ISAC) system comprising an FD base station (BS), an FD user, and a monostatic target. The system enables simultaneous uplink and downlink communication between the FD BS and FD user, while the BS detects the target by sensing the reflected signals. Signal-to-interference-plus-noise ratio (SINR) expressions are formulated for uplink and downlink communication, as well as target sensing. Analytical expressions are derived for the ergodic capacities of downlink and uplink communication, leading to the communication sum rate. Additionally, using the target sensing SINR, the sensing rate is determined. The study further establishes that optimizing the power apportioning parameter at the BS can maximize the communication sum capacity while achieving the desired sensing rate. The accuracy of analytical results is validated through Monte Carlo simulations.
Rapid digitalization in e-health and biotechnology is transforming the way biological and medical data are measured, processed, and secured. These breakthroughs are driven by cutting-edge technologies that leverage distributed computing to provide scalability, enhanced security, and precise data acquisition. However, the efficient management of micro and nanoscale biological data remains a significant challenge, particularly due to the complexities involved in extracting, transmitting, and processing this data across scalable Internet of Things (IoT) infrastructure. To tackle these challenges, this paper introduces a novel and highly efficient Digital Twin (DT) framework that integrates bio-network technologies, specifically the Internet of Bio-Nano Things (IoBNT), along with decentralized deep learning algorithms, including Federated Learning (FL) and Convolutional Neural Networks (CNN). The proposed framework employs CNNs for robust pattern recognition, FL to reduce bandwidth requirements and enhance data privacy, and IoBNT devices to ensure precise data acquisition with minimal error. Our approach has demonstrated an outstanding AI-Enabled DT framework with an accuracy of 98.21% across 33 bacterial categories, showcasing the framework’s potential for scalable, secure, and efficient biological data analysis. This work paves the way for practical, high-performance data-driven DTs in e-health and biotechnological applications, which hold the potential to redefine how data is leveraged for advancing medical innovation.
Co-site interference in multi-radio systems, caused by insufficient isolation between high-power transmitters and receivers, induces deterioration of receiver communication performance. While existing co-site interference cancellation methods rely on fixed-position auxiliary receive antennas (ARA) requiring additional radio frequency (RF) cancellation circuits, this paper proposes a hardware-efficient movable auxiliary receive antenna (MA-ARA) interference cancellation (IC) method, which can minimize residual interference power by dynamically optimizing the position of MA-ARA. First, a position optimization model is formulated, followed by the closed form expression for residual interference power related to the position of MA-ARA is derived in dual-interference scenarios. Subsequently, swarm intelligence algorithms are utilized to solve the position optimization problem. Results demonstrate a 50.93 dB interference cancellation ratio (ICR) and 35.19 dB SINR improvement of dual-interference scenarios which is 80% faster convergence than brute force algorithm (BFA), and 60.03 dB ICR of single-interference scenarios which is 7dB higher than that of conventional RF cancellation method, confirming the effectiveness of the MA-ARA position optimization method.
The terahertz (THz) band holds significant potential for communication, localization, and sensing in the future 6G ecosystem, owing to its high speed and large capacity. Power amplifiers (PAs) are crucial components in THz systems for increasing transmission power to overcome the high propagation loss of the THz frequency band. However, the inherent nonlinear characteristics of PA make out-of-band regrowth and in-band distortion of the input signal when the power of the input signal is high, which degrades the quality of the received signal. In THz systems, due to the extremely large signal bandwidths, existing hardware struggles to support the sampling and processing rates required for 5-10 times of the baseband bandwidth, causing traditional digital nonlinear correction methods to fail. To address this issue, we propose an anti-aliasing post-correction (APC) method at the receiver to compensate the nonlinearity of the front-end of the transmitter with baseband sampling rate. The proposed APC model reconstructs high-sampling-rate data from low-sampling-rate symbols and employs anti-aliasing filters to prevent spectral aliasing caused by higher-order nonlinearities. Simulations show that the APC method can improve the signal error vector magnitude (EVM) from 6.42% to 2.34% with a baseband sampling rate and processing rate.
In mobile communications, while the use of higher frequency bands is increasing with each successive generation, the importance of effective utilization of relatively low frequency bands with excellent radio propagation characteristics is also becoming increasingly recognized. In order to make effective use of the ultra high frequency (UHF) band, the application of multiple-input multiple-output (MIMO) channel transmission is examined. In MIMO transmission, it is important to ensure sufficient antenna spacing because narrow antenna spacing increases spatial correlation and adversely affects transmission performance. However, the UHF band has a long wavelength, which increases the size of antennas in an array configuration at a base station. Therefore, it is necessary to configure a highly efficient antenna configuration. In this paper, we conducted an experiment in which we actually transmitted radio waves and measured the received signal. After confirming the transmission performance of the equipment used in the 12x12 MIMO transmission experiment, 12x12 MIMO transmission was performed in the UHF band in a real outdoor environment, and the relationship between the three types of transmit antenna configurations and bit error rate was clarified.
Over-the-Air computation (OTA) has emerged as a promising access scheme for Federated Learning (FL) in wireless AI networks, owing to its integrated communication and computation capabilities. With the expansion of distributed AI application scenarios, FL has evolved from single-task to multi-task. However, the shared use of time and frequency resources in Federated Multi-Task Learning (FMTL) inevitably introduces inter-task interference, bringing new challenges to communication system design. This paper studies the optimization problem of minimizing the aggregation error of the FMTL communication in multiuser single-input multiple-output (SIMO) orthogonal frequency division multiplexing (OFDM) systems. We propose an efficient joint optimization algorithm for subcarrier allocation, transmit power control, and receive beamforming, which iteratively updates them via alternating optimization. Experimental results show that jointly optimizing communication resources in frequency, power, and beam domains can significantly improve the performance of OTA-FMTL. The proposed algorithm achieves near-optimal loss performance. Compared to the counterpart with full-band frequency reuse, it significantly reduces the training loss, and compared to the counterpart with no frequency reuse, it improves the training convergence speed by over 25%.
This paper presents a quantum deep Q-network (QDQN) approach for adaptive resource control in augmented reality (AR) systems supported by space-air-ground integrated network (SAGIN) architecture. Specifically, we jointly optimise the image sampling rates of ground cameras, task offloading decisions at unmanned aerial vehicles (UAVs), and bandwidth allocation to minimise a weighted multi-objective function of latency and bandwidth utilisation. The resulting problem is formulated as a mixed-integer non-linear programming problem, which poses significant computational challenges for conventional optimisation methods. By leveraging the expressibility of parameterised quantum circuits (PQCs) and the capability of advanced reinforcement learning, the proposed QDQN solution efficiently addresses the problem and enables adaptive decision-making. Simulation results demonstrate that the QDQN achieves significantly better training stability and faster convergence compared to the classical DQN approach. Notably, the proposed method reduces task delay by up to 100 times compared to conventional benchmarks, highlighting the substantial advantages of quantum-enhanced reinforcement learning in complex resource-constrained AR scenarios.
Vertical Federated Learning (VFL) enables edge devices to collaboratively train models using different features of the same users. However, conventional VFL relies on aligned samples for training, which restricts data utilization and leads to suboptimal model performance, particularly as the proportion of aligned samples decreases with more participating parties. To address this limitation, this paper presents the Pretrained Vertical Federated Learning with Diffusion Model (VFLDiff), which enhances data utilization and improves model accuracy by incorporating both aligned and unaligned samples. VFLDiff is designed with two steps. First, the active party pretrains a Diffusion Model with its entire dataset, including unaligned samples, thus feature representations can be learned robustly. Second, the pretrained Diffusion Model's encoder is used as the active party's bottom model in the VFL training phase, which enables collaborative learning with the passive party. Experimental results demonstrate that VFLDiff consistently outperforms traditional VFL algorithms and pretrained convolutional neural network baselines across various scenarios, especially at low data alignment ratios.
Experimental evidences are presented to illustrate the resiliency of an Orbital Angular Momentum (OAM) assisted Self-healing Link in the presence of a near and a far field partial blockage. The importance of the resiliency of a communication link is attracting more attention, especially in a 5G beyond and 6G systems. For the first time, we experimentally demonstrated that self-healing is possible for a far field partial obstructer located within wireless communication link for an OAM beam at 28 GHz. In addition, more experimental results were presented to reinforce our previous publications on the self-healing property with near field partial obstructer. We also experimentally confirmed that for both the far and near-field partial obstructers, the mmWave OAM-assisted self-healing beam exhibits a relatively stronger received signal compared to a non-self-healing non-OAM beam. Furthermore, we also have demonstrated the viability of the partially obstructed OAM communication link by measuring the integrity of their receiver (Rx) constellation signals. In addition, using an asymmetrical partial blockage within the OAM link, the Rx signal strength of the positive and negative OAM mode polarities is shown to be directly related to the sign of the displacement of the partial obstructer with respect to the z axis of the link.
We propose a scalable semantic communication (SC) framework for cell-free MIMO (CF-MIMO) networks with wireless access and fronthaul links, comprising multiple devices, distributed access points (APs), and a central processing unit (CPU). Each device sends its local sensing observations to the CPU with the aid of APs. Then, the CPU performs task-oriented inference with neural networks (NNs). Due to the heterogeneous propagation environments, the sets of active devices and APs randomly vary at each transmission. Consequently, SC in the CF-MIMO must remain scalable for arbitrary node populations. We establish an end-to-end SC procedure as a permutation-invariant set function. The overall system is decomposed into component NNs reused across all devices and APs, thereby eliminating the need for retraining when the network size varies. To further enhance the generalization ability, we propose an attention-based weighted aggregation mechanism. Numerical results demonstrate the superiority of the proposed scheme.
The increasing demand for sustainable energy use and the growing concerns about hidden energy waste in buildings, usually referred to as the phantom load, highlight the critical challenge of optimizing it. The phantom load refers to the energy consumed by plug-in devices that remain in standby or idle mode, which can account for up to 32% of a building’s total energy consumption, representing a significant opportunity for energy savings. To address this, we propose an intelligent building energy management system that integrates an Edge-Enabled Digital Twin (EEDT) with a fuzzy logic-based control strategy to effectively identify and reduce the phantom load. By combining multi-feature clustering with fuzzy control rules, the system accurately detects phantom load behavior and minimizes its impact through automated shutdown, delayed operation, or user reminder. Furthermore, a Long-Short-Term Memory (LSTM)-based prediction module forecasts the next day’s user energy consumption trend based on data from the preceding two days. This enhances user awareness and enables proactive energy management. An experimental case study in a university research lab demonstrated the system’s effectiveness, showing it can save approximately 40% of the total energy per plug-in device weekly. The fuzzy decision-making framework, utilizing 27 optimized rules, achieved a phantom load reduction of up to 82%. These results indicate the system’s strong potential for plug-in energy optimization and its potential contribution to emission reduction goals in smart building environments.
