As electric vehicles (EVs) join the 6G Internet of Vehicles (6G-IoV), they must charge while moving and process data in real time. Static plug-in charging and on-board CPUs cannot meet these dual demands. We therefore combine two technologies. First, wireless power transfer (WPT) on electrified roads (eROAD) keeps an EV’s battery topped up throughout its trip. Second, a privacy-preserved multi-agent federated deep-RL (MA-FDRL) framework decides, every slot, where each task runs—on the vehicle, a roadside unit (RSU), or the base station (BS). Federated learning keeps raw data local; differential privacy protects model updates; deep reinforcement learning allocates radio, compute, and energy resources for many agents in real time. Experiments show the scheme cuts task-offloading latency, balances edge-server load, and scales gracefully with network size. The outcome suggests an energy-conscious architecture capable of meeting the low-latency and privacy needs of 6G-IoV.

The rapid growth of connected vehicles in sixth-generation (6G) networks demands real-time, privacy-preserving offloading of ultra-reliable, low-latency communications (URLLC) tasks. We propose Quantum-federated deep reinforcement learning (Q-FDRL). This decentralized federated learning framework tightly integrates a large artificial intelligence (AI) model (LAM), composed of a multi-layer transformer encoder with a shallow variational quantum circuit (VQC) head. This hybrid architecture enables compact, high-dimensional state representation under on-device compute constraints. Vehicular task offloading is modeled as a multi-agent Markov game-including full data partitioning, power-splitting uplink, finite-blocklength transmission, and dynamic CPU allocation- and solved via local advantage actor-critic (A2C) updates at vehicles, unmanned aerial vehicles, and roadside units. To guarantee rigorous (ε, δ)-differential privacy, we apply Gaussian-mechanism gradient clipping and secure one-time-pad encryption over quantum key distribution (QKD) links, with a moments accountant bounding the cumulative privacy loss to ε ≤ 5. In simulations over 250 episodes, Q-FDRL converges approximately ∼ 27.27% faster and reduces early-stage variance by ∼ [11.11, 42.86]% compared to centralized A2C and classical federated baselines, demonstrating its effectiveness for scalable, privacy-preserving vehicular resource management.

We study joint phase-amplitude control for near-field holographic multiple-input multiple-output systems in intelligent autonomous transport systems. Under exact Fresnel propagation and per-user quality-of-service (QoS) constraints, the design couples discrete tile phases, continuous amplitudes, a multi-user digital precoder, and non-orthogonal multiple access (NOMA) power fractions, resulting in a mixed-integer nonconvex program. We propose a hybrid quantum-classical solver, the Quantum approximate-and-gradient optimizer for the near field (QAG-NF), which alternates between a blockwise quantum approximate search over tile phases and exact-gradient updates of amplitudes, power shares, and the precoder. A penalized objective, monotone-accept rule, and Armijo backtracking procedure preserve feasibility and guarantee non-decreasing utility at accepted iterates. The continuous stage uses geometry-consistent preconditioners, including Jacobian-scaled amplitude steps and a softmax Fisher/natural gradient for NOMA powers, while the discrete stage uses few-qubit shallow circuits per block, making the method compatible with near-term noisy quantum hardware and practical simulators. Beyond solver design, we incorporate Doppler-averaged effective gains to impose conservative power floors consistent with successive interference cancellation, thereby safeguarding QoS throughout the line search. In a representative deployment, QAG-NF provides a substantial performance gain at equal base-station power, improving the sum rate by about 31.64% over a near-field projected-gradient baseline and by roughly$7.39\times $over a far-field surrogate, while also achieving higher Jain fairness.

This paper presents an optimization framework aimed at enhancing communication throughput for electric vehicles (EVs) operating on electrified roads (eROADs) while simultaneously ensuring quality of service (QoS). By integrating inductive coil-based wireless power transfer (WPT) with multiuser vehicular networks, our framework facilitates continuous inmotion charging alongside efficient communication management. To address transmission power and data rate constraints in a non-orthogonal multiple access (NOMA) uplink scenario, we tackle challenges posed by realistic mobility models, Doppler effects, and variability in network coverage. We develop a hybrid deep reinforcement learning (DRL) algorithm that integrates deep deterministic policy gradient (DDPG) and proximal policy optimization (PPO) techniques to maximize throughput. Simulation results indicate that our hybrid DRL algorithm surpasses traditional DDPG and PPO methods, achieving up to 13.74% higher throughput under low transmission power limits. Our approach manages WPT and communications efficiently, promoting EV operations in intelligent transport systems and aligning with green communication goals.

Smart wireless communications enabled by reconfigurable intelligent surfaces (RISs) have gained significant research interest in the areas of localization and sensing over the past few years. This paper investigates an unconventional approach for 3D direction-of-arrival (DoA) estimation of multiple aerial user targets using an RIS-based communication architecture. In particular, the measurements required for DOA estimation at the receivers are optimized through a deep reinforcement learning framework. The core of the proposed method lies in formulating the DoA estimation problem as a Markov decision process (MDP), which is optimized via a proximal policy optimization algorithm for its optimization. Considering a practical RIS setup with 2-bit states at each unit element, we demonstrate significant improvements in DoA estimation accuracy, in terms of reduced root mean squared error (RMSE) for various simulation scenarios of the system.

This work proposes a quantum-aided deep reinforcement learning (DRL) framework designed to enhance the accuracy of direction-of-arrival (DoA) estimation and the efficiency of computational task offloading in integrated sensing and communication systems. Traditional DRL approaches face challenges in handling high-dimensional state spaces and ensuring convergence to optimal policies within complex operational environments. The proposed quantum-aided DRL framework that operates in a military surveillance system exploits quantum computing’s parallel processing capabilities to encode operational states and actions into quantum states, significantly reducing the dimensionality of the decision space. For the very first time in literature, we propose a quantum-enhanced actor-critic method, utilizing quantum circuits for policy representation and optimization. Through comprehensive simulations, we demonstrate that our framework improves DoA estimation accuracy by 91.66% and 82.61% over existing DRL algorithms with faster convergence rate, and effectively manages the trade-off between sensing and communication and by optimizing task offloading decisions under stringent ultra-reliable low-latency communication requirements. Comparative analysis also reveals that our approach reduces the overall task offloading latency by 43.09% and 32.35% compared to the DRL-based deep deterministic policy gradient and proximal policy optimization algorithms, respectively.

The increasing complexity of real-time healthcare necessitates intelligent systems for dynamic data management and personalized assistance. This paper proposes a novel dual-LLM framework that integrates large language models (LLMs) into wireless healthcare networks. The first LLM powers an interactive artificial intelligence module (IAIM) embedded within a mobile edge computing (MEC) environment, which dynamically optimizes user-specific data routing and reconfigurable intelligent surface (RIS) configurations via a modified proximal policy optimization (PPO) algorithm. A novel Greedy Look-Ahead Algorithm (GLAA) is introduced for real-time path selection based on signal strength, emergency factors, and user-specific parameters. The second LLM, utilizing a retrieval-augmented generation (RAG) approach, serves as a personalized healthcare chat assistant that delivers context-aware patient support using real-time and historical data. Simulation results demonstrate that the proposed IAIM achieves a 9.6% reduction in network overhead compared to manual modeling and reduces latency by up to 52.5% over baseline PPO approaches, thus enabling enhanced user experience and responsiveness in healthcare systems.

In the context of ultra-reliable low-latency communication (URLLC) in Internet-of-Things (IoT) networks, conventional half-space coverage limits the flexibility of reconfigurable intelligent surface (RIS) deployment. To overcome these constraints, this paper makes use of active simultaneously transmitting and reflecting RIS (STAR-RIS), which is seamlessly integrated into digital twin (DT) and mobile edge computing (MEC) frameworks. Our primary research objective is to achieve full-space coverage by enabling simultaneous transmission and reflection of the signals while improving uplink data transmission from IoT URLLC user nodes (UNs) to the base station (BS) with the assistance of active STAR-RIS, even in the presence of imperfect channel state information (CSI). We formulate the problem of minimizing total end-to-end (e2e) latency, computed using the alternating optimization (AO) algorithm. Subsequently, we have evaluated the performance of the AO algorithm against the stochastic gradient descent (SGD) algorithm, which serves as the benchmark solution. The simulation outcomes delineate a performance evaluation under perfect and imperfect CSI scenarios. The AO algorithm outperforms SGD with latency reductions of 19.7% at$N=32$and 20.4% at$N=64$. Increasing N from 32 to 64 results in a 39.3% latency reduction for AO, surpassing SGD’s 38.8%. However, the SGD algorithm consistently exhibits lower computational complexity compared to the AO algorithm. Additionally, the energy splitting mode achieves the system’s total e2e latency reductions of 28.4% over the mode switching mode and 11.04% over time switching mode. Furthermore, active STAR-RIS optimal beamforming (ARO) achieves$\approx 10$% latency reduction over the predictive optimal beamforming (PRO), which itself surpasses active STAR-RIS with random beamforming (ARR) by$\approx 9$%. This comparison considers key factors such as the power budget, the number of RIS elements, the caching capacity of the edge computing server (ECS), the number of IoT UNs, the minimum transmission rate, and maximum transmit power at BS of active STAR-RIS.

This paper introduces a novel multi-agent federated deep reinforcement learning (MA-FDRL) framework designed to minimize vehicular task offloading latency in electrified road (eROAD) environments. The solution integrates inductive coil-based wireless power transfer (WPT) systems with full duplex multiple input and multiple output (MIMO) vehicular networks, enabling continuous charging and reducing computational delays for electric vehicles (EVs) on eROADs. The MA-FDRL framework optimizes the offloading of vehicular tasks, with support from base stations (BS), unmanned aerial vehicles (UAVs), and satellites, while ensuring data privacy through differential privacy techniques. By intelligently distributing resources across these supporting entities, the framework enhances the efficiency of task processing. Key challenges such as dynamic wireless charging, intermittent BS coverage, and privacy-preserving task offloading are addressed using a comprehensive WPT framework, a mobility model, and a differential privacy-enhanced MA-FDRL algorithm. The proposed MA-FDRL solution effectively reduces the latency of vehicle task offloading by 17.05% over proximal policy optimization (PPO) algorithm, ensures balanced task distribution between edge servers, and offers a scalable and privacy-preserving approach for future autonomous electric vehicles and connected wireless environments.

This paper investigates the deployment of a full-duplex integrated sensing and communication (ISAC) system for task offloading service to serve ultra-reliable low-latency communications (URLLC), significantly enhanced by a simultaneously transmitting and reflecting reconfigurable intelligent surface (STAR-RIS). Utilizing a non-orthogonal multiple access frame-work, this study extensively addresses the challenges associated with latency-sensitive task offloading from information receivers (IRs) to a mobile edge computing platform. We introduce a novel and sophisticated multi-agent deep reinforcement learning (MA-DRL) approach aimed at minimizing latency in task offloading under a variety of stringent ISAC-URLLC network constraints, including imperfect channel state information. The proposed MA-DRL operates on decentralized execution while maintaining centralized training, using multi-actor-critic networks to enhance learning and performance. The innovative reward decentralization framework in the present MA-DRL optimizes downlink and uplink communications through dynamic power allocation, precise beamforming, and intelligent phase shift management facilitated by the STAR-RIS. The proposed MA-DRL framework significantly outperforms existing multi-agent DRL algorithms, demonstrating substantial gain in reward maximization (i.e., linked to offloading latency minimization) by 27.46% and 52.73%.

This paper presents a novel design for a mobile edge computing (MEC) service that integrates digital twin technology with an active simultaneously transmitting and reflecting reconfigurable intelligent surface (STAR-RIS). This configuration leverages edge intelligence, aiming to strengthen ultra-reliable and low-latency communications (URLLC) within Internet-of-Things (IoT) frameworks. We explore the uplink data transmission path from singular-antenna IoT URLLC nodes (UNs) to a multi-antenna base station (BS) facilitated by an active STAR-RIS. Our focus is on framing an end-to-end (e2e) latency reduction strategy for the presented system. Due to the inherent non-convexity of this problem, we propose an efficient alternating optimization (AO) algorithm to get a solution. This algorithm decomposes the main problem into five distinct sub-problems: transmit beamforming design, optimization of caching and offloading policies, joint communication and computation optimization, and enhancement of active STAR-RIS beamforming. An extensive set of simulation outcomes indicates that our DT-enhanced optimal-phase STARRIS approach consistently surpasses benchmark methods, particularly when accounting for variables such as power constraints, the number of RIS elements, the caching capacity of the edge computing server (ECS), and the number of IoT UNs.

In this paper, we present an innovative approach to spectrum management in cognitive radio networks (CRNs) aimed at serving ultra-reliable low-latency communication (URLLC) enabled secondary users (SUs). Unmanned aerial vehicles (UAVs) are deployed for accurate and reliable spectrum sensing (SS), enhancing cooperative spectrum sensing (CSS) effectiveness. A distinctive aspect of our methodology is the integration of digital twin (DT) technology, which, to our knowledge, has not been explored previously in the context of CRNs for bandwidth assignment to URLLC-enabled SUs. This integration facilitates more sophisticated and adaptive management of spectrum resources. Moreover, we propose a deep reinforcement learning (DRL) framework incorporating a modified proximal policy optimization (MPPO) algorithm. This algorithm is designed for better stability and convergence, outperforming the standard PPO in terms of faster convergence in the present URLLC transmission latency minimization process. Simulation results indicate that our proposed DT-based spectrum management and MPPO in CRNs result in a 27.89% increase in CRN's average throughput and a 39.94% reduction in transmission latency compared to the conventional equal resource allocation scheme.

The substantial power consumption attributed to the active components within fully-connected reconfigurable intelligent surface (RIS) architecture significantly hinders the efficiency and sustainability of DT-enabled MEC networks. To tackle this challenge, we present an innovative sub-connected architecture for active RIS within the digital twin (DT) integrated mobile edge computing (MEC) framework of an Internet-of-Things (IoT) networks, capitalizing on edge intelligence to enhance ultra-reliable and low-latency communication (URLLC) services. The primary aim of our research is to improve uplink data transmission from IoT URLLC user nodes (UNs) to a base station (BS) with the aid of an active RIS, even under an imperfect channel state information (CSI). We have formulated the total end-to-end (e2e) latency minimization problem, which is solved by using an efficient alternating optimization (AO) algorithm. The algorithm breaks down the proposed non-convex problem into five subproblems, namely, beamforming design, caching and offloading policy optimization, joint communication and computation optimization, and joint active RIS phase shift and amplification factor vector optimization. We conducted a thorough analysis of the convergence properties of the proposed AO algorithm, benchmarking its performance against the established Heuristic algorithm. Our simulation results consistently demonstrate the superiority of our proposed DT-assisted optimal phase sub-connected active RIS scheme over various benchmark schemes, taking into account various factors such as the number of RIS elements, power budget constraints, imperfect CSI, edge computing server (ECS) cache capacity, number of IoT UNs, and the number of power amplifiers.

This work proposes a new model of reconfigurable intelligent surface (RIS) called cognizable RIS (CRIS) that is specifically designed to meet the unique demands of users who require extreme-ultra-reliable and low-latency Communication (xURLLC) in the sixth generation (6G) wireless networks. The programmable elements in the proposed CRIS unit can adapt to different modes of operation to provide significant performance gain. To improve reliability at the receiver, we integrate unmanned aerial vehicles with the CRIS module, which enhances network performance through beamforming and mobility. Our study focuses on maximizing the sum throughput in a multiple-input multiple-output scenario using the rate-splitting multiple access communication system. To achieve this, we introduce a novel hybridized multi-agent-based deep reinforcement learning (DRL) algorithm for optimal resource allocation that maximizes the sum throughput. We incorporate long-short-term memory (LSTM) networks into our proposed DRL to address the temporal dependencies due to stochastic channel conditions. By utilizing the proposed LSTM-based multi-agent DRL (MA-DRL) algorithm, we achieve notable gains of 11.7% and 26.9% in sum throughput over widely recognized DRL benchmark algorithms, all while adhering to xURLLC’s stringent maximum packet error probability constraint of 10−9.

This work proposes a novel design for an active reconfigurable intelligent surface (RIS)-assisted digital twin (DT) based mobile edge computing (MEC) model that leverages edge intelligence to enhance ultra-reliable and low-latency communications (URLLC) services in Internet-of- Things (loT) networks. The system model considers uplink data transmission from the single antenna IoT-URLLC nodes (UNs) to a multi-antenna base station (BS) with the aid of an active RIS under imperfect channel state information (CSI). We formulate a total end-to-end (E2E) latency minimization problem for the proposed system model. An efficient alternating optimization (AO) algorithm is proposed to tackle the non-convexity of the problem by reformulating it into five subproblems: beamforming design, caching and offloading policies optimization, joint communication and computation optimization, and active RIS phase shift optimization. Simulation results demonstrate that the proposed DT-assisted optimal-phase active RIS scheme consistently outperforms benchmark schemes, such as optimal-phase passive RIS, random-phase active RIS, and no- RIS systems, considering factors such as imperfect CSI, power budget, number of RIS elements, the caching capacity of edge computing server (ECS) and the number of loT UNs.

This paper investigates the problem of spectral efficiency maximization in an underlay cognitive radio network (CRN) utilizing rate-splitting multiple access (RSMA) transmission for MISO downlink under short packet communications and imperfect channel estimation information. In particular, we focus on an effective transmit beamforming design at the cognitive base station while satisfying the requirements of ultra-reliable and low-latency communication (URLLC), interference temperature, power budget, and minimum throughput. We model the dynamic resource allocation problem as a Markov decision process (MDP) and employ deep reinforcement learning techniques, specifically the deep deterministic policy gradient (DDPG) and proximal policy optimization (PPO) algorithms, while taking into account the time-varying channel conditions. Simulation results demonstrate that the DDPG algorithm outperforms PPO at low interference temperatures for the primary receiver, while the opposite holds at high interference temperatures. Moreover, the considered RSMA system for CRN outperforms traditional multi-user linear precoding and power-domain multiple access schemes while maintaining small packet sizes and high reliability.

Recent progress in cognitive radio networks (CRNs) promises to meet device‐to‐device communication requirements for spectrum utilisation and power control to support billions of machines/devices to be connected worldwide. The architecture of the CRN must maintain a high data rate (throughput) at low power consumption, which requires both spectrum efficient and energy efficient system design. To this aim, the proposed work adopts a CRN model, which operates in an interweave mode that allows spectrum sensing followed by opportunistic secondary user (SU) data transmission over the unused bandwidth of the primary user (PU) in a non‐overlapping frame structure. Closed form expressions of the optimal spectrum sensing duration, bandwidth, and power allocation of each secondary node are obtained to maximise the sum throughput of the overall CRN while maintaining the constraints of sensing reliability of PU, individual SU transmission outage probability, permissible interference, and residual bandwidth. Numerical results highlight that the proposed scheme maximises the sum network throughput by and over the other existing techniques.

Cooperation among the sensing nodes in cognitive radio (CR) networks improves spectrum sensing (SS) reliability. However, determining the efficient set of secondary users (SUs) to be involved in cooperative SS (CSS) becomes challenging in vehicular networks (VNs) due to the mobility of primary user (PU) and/or SU nodes. This work explores an energy detection‐based CSS scheme in CR enabled VNs (CR‐VNs) considering the uniform velocity motion of PU and SU nodes. A distance‐dependent distribution function is developed first to find the probability of a SU resides in a specific coverage of the PU transmission zone. This distribution function is then used in deriving the expression of the probability of detection and probability of false alarm using majority rule at the fusion centre. The efficacy of the mathematical analysis is studied through a large set of simulation results that highlight the enhanced SS performance in terms of the probability of detection by , and at uniform velocity over the existing works in CR‐VN.