This paper focuses on enhancing the energy efficiency (EE) of a cooperative network that features a miniature unmanned aerial vehicle (UAV) operating at terahertz (THz) frequencies and equipped with holographic surfaces to improve network performance. Unlike traditional reconfigurable intelligent surfaces (RIS), which serve as passive relays for signal reflection, this work introduces a novel concept: energy harvesting (EH) using reconfigurable holographic surfaces (RHS). These surfaces provide more powerful and focused energy delivery during wireless power transfer than RIS and are mounted on the miniature UAV. In this system, a source node enables the UAV to simultaneously receive both information and energy signals, with the harvested energy powering data transmission to a specific destination. The EE optimization problem involves adjusting non-orthogonal multiple access (NOMA) power coefficients and the UAV’s flight path while accounting for the unique characteristics of the THz channel. The problem is solved in two stages to maximize EE and meet a target transmission rate. The UAV trajectory is optimized using a successive convex approximation (SCA) method, followed by the adjustment of NOMA power coefficients through a quadratic transform technique. Simulation results demonstrate the effectiveness of the proposed algorithm, showing significant improvements over baseline methods.

Coordinated service offloading is essential to meet Quality-of-Service (QoS) targets under non-stationary edge traffic. Yet conventional schedulers lack dynamic prioritization, causing deadline violations for delay-sensitive, lower-priority flows. We present PRONTO, a multi-agent framework with centralized training and decentralized execution (CTDE) that jointly optimizes SLA-aware offloading and adaptive service orchestration. PRONTO builds on Twin Delayed Deep Deterministic Policy Gradient (TD3) and incorporates spatiotemporal, topology-aware graph attention with top-K masking and temperature scaling to encode neighborhood influence at linear coordination cost. Gated Recurrent Units (GRUs) filter temporal features, while a hybrid reward couples task urgency, SLA satisfaction, and utilization costs. A priority-aware slicing policy divides bandwidth and compute between latency-critical and throughput-oriented flows. To improve robustness, we employ stability regularizers (temporal smoothing and confidence-weighted neighbor alignment), mitigating action jitter under bursts. Extensive evaluations show superior QoS and channel utilization, with up to 27.4% lower service delay and over 18% higher SLA Satisfaction Rate (SSR) compared with strong baselines.

Reconfigurable Intelligent Surfaces (RIS) have emerged as a key solution to dynamically adjust wireless propagation by tuning the reflection coefficients of large arrays of passive elements. Reconfigurable Holographic Surfaces (RHS) build on the same foundation as RIS but extend it by employing holographic principles for finer-grained wave manipulation — that is, applying higher spatial control over the reflected signals for more precise beam steering. In this paper, we investigate shape-adaptive RHS deployments in a multi-user network. Rather than treating each RHS as a uniform reflecting surface, we propose a selective element activation strategy that dynamically adapts the spatial arrangement of deployed RHS regions to a subset of predefined shapes. In particular, we formulate a system throughput maximization problem that optimizes the shape of the selected RHS elements, active beamforming at the access point (AP), and passive beamforming at the RHS to enhance coverage and mitigate signal blockage. The resulting problem is non-convex and becomes even more challenging to solve as the number of RHS and users increases; to tackle this, we introduce an alternating optimization (AO) approach that efficiently finds near-optimal solutions irrespective of the number or spatial configuration of RHS. Numerical results demonstrate that shape adaptation enables more efficient resource distribution, enhancing the effectiveness of multi-RHS deployments as the network scales.

Rate-splitting multiple access (RSMA) has been proposed as a flexible multiple access scheme for improving interference management in sixth-generation (6G) networks. In particular, the low latency facilitated by RSMA and its robustness against user mobility and imperfect channel state information make it an ideal candidate for the ultra-reliable and low-latency (URLLC) use case in 6G networks. However, since the common message in RSMA needs to be decoded by all the users, the achievable rate of the common message is determined by the user with the poorest channel quality. To overcome this bottleneck, an active intelligent reflecting surface (IRS) can be deployed to enhance the achievable rate of the common stream. However, this comes at the expense of additional power consumption due to the active IRS. In this paper, we consider an active IRS-aided RSMA-based downlink URLLC system and study the resource allocation design for minimization of the power consumption of the base station and the active IRS under quality-of-service constraints for the URLLC users. Our simulation results reveal that active IRSs yield a lower overall power consumption and require a smaller surface size compared to passive IRSs in RSMA-based URLLC systems. Moreover, we show that active IRS-aided RSMA systems consume less power than active IRS-aided space division multiple access (SDMA) systems.

Due to the limited number of radio frequency (RF) chains in millimeter wave (mmWave) massive multiple-input multiple-output (MIMO) receivers using analog beamforming/hybrid beamforming, there is a restriction in scheduling the number of users in each transmission time interval. Therefore, fast and low-complexity user scheduling methods based on the instantaneous channel state information (CSI) are needed. In this paper, we propose novel user scheduling methods based on deep learning (DL) to reduce the size of the search space by using the learning capability of a deep neural network (DNN). We formulate the user scheduling combinatorial optimization problem as a regression problem followed by a user separation procedure through decision boundaries that are learned by a trained DNN. The decision boundaries are used to separate the users into two subsets. Then, one of the subsets is selected to be searched to find the users that maximize the sum-rate capacity. The proposed method can achieve a very low outage probability with a few number of searches. In order to achieve ergodic capacity with lower computation complexity, the proposed method is employed in combination with the genetic algorithm (GA) algorithm to take advantage of intelligent initial population selection. Our simulation results show that the proposed user scheduling methods can offer remarkably low complexity.

This paper considers the downlink of a cloud radio access network (C-RAN) consisting of a central processor (CP) and a network of connected radio units (RUs). We propose a novel resource allocation solution for the scenario with full-duplex (FD) self-backhauling RUs connected through hybrid radio-frequency (RF)/free-space optical (FSO) links to the CP for improved network throughput. This enables us to study the feasibility of the FD mode in terms of required self-interference cancellation to outperform the benchmark half-duplex hybrid RF/FSO transmission. Since the derived optimization problem for the design of the linear precoders and quantizers subject to the fronthaul capacity, zero-forcing, and power constraints, is non-convex and intractable, we develop an algorithm to solve it via an alternating optimization approach. In the simulation results, the proposed hybrid RF/FSO policy is assessed in terms of achievable rate, and we highlight the parameter range for which FD transmission is more rewarding than the time-division approach, under different weather conditions and selected RF bandwidth.

With the excessive growth in mobile users' traffic, radio resource management (RRM) techniques should undergo revolutionary changes to be competent enough to meet the ever-increasing users' demands. Virtualized wireless network (VWN) has emerged as a satisfactory solution in the fifth-generation (5G) cellular networks ensuring the required quality-of-service (QoS) of distinct slices. Yet, it seems that tackling RRM problems in VWNs using conventional optimization is not practical for real-time applications. In this paper, driven by the advancements of machine learning, we consider the throughput maximization problem in a cloud radio access network (C-RAN) assisted softly virtualized wireless network supporting different types of services and solve it with a deep Q-learning (DQL) algorithm. The performance of the proposed policy is thoroughly evaluated via simulation results with respect to the isolation rate, penalty value as well as the discount factor. It is shown that our proposed policy achieves a higher sum rate compared to the existing baseline namely a greedy search-based power allocation strategy.

In large-scale multiple-input multiple-output (LS-MIMO) systems, by exploiting hundreds of antennas at the base station, spectral efficiency, power efficiency, and link reliability can be enhanced significantly. However, by increasing the number of antennas, the computational complexity of the detectors makes the hardware implementation intractable, and therefore, LS-MIMO systems require sub-optimal low complexity detection algorithms. In this paper, two novel approaches for improving factor graphbased belief propagation with Gaussian approximation of interference (FG-BP-GAI) algorithm is proposed to reduce the computational complexity of the belief propagation (BP) based receiver without bit error rate (BER) degradation. More specifically, two novel techniques, namely odd Taylor series and odd least square, are proposed to approximate the a posteriori probability in the FG-BP-GAI policy with few polynomial terms of low degree. In the simulation results, the performance of our proposed algorithms are assessed and it is shown that our proposed improved FGBP-GAI policies can achieve lower computational complexity compared with the other approaches in the literature like MRF-BP algorithm without BER degradation.

In this paper, we consider a cooperative underlay cognitive radio network in which the primary network (PN) consists of a transmitter and receiver and the secondary network (SN) has K bidirectional half-duplex relays. In the SN, two secondary transceivers adopt multiple access broadcast protocol for the secondary data transmission and at each bidirectional relay, there exist two buffers of size L data elements. Hence, each relay can store the incoming secondary data and retransmit it in an appropriate time slot later. We propose a novel buffer-aided bidirectional relay selection policy with secondary power minimization and successive interference cancellation in which the interference between the PN and SN is eliminated. Since buffers are used at the relays, data transmission in the SN is not limited to a predefined schedule. Hence, at each time slot, based on the instantaneous buffer state information of the relays and the instantaneous or statistical channel state information of the involved links, the SN makes a decision. The SN decides optimally when to use one of the relays for the multiple access, use one of the relays for the broadcast mode or be silent provided that the data transmission in both the PN and SN are error free and the secondary power expenditure is minimized. Simulation results show that the proposed scheme minimizes the secondary power expenditure, and achieves up to 40% improvement in the secondary throughput for 6 middle relays compared to the other recently proposed policies without buffer.

In this paper, a cognitive machine-to-machine (M2M) communication network is considered, in which a cellular network shares the spectrum with the M2M communication network with M machine-type devices (MTDs), one half-duplex relay, and one MTD gateway for data gathering. One key challenge is that in the future 5G wireless networks, there will be billions of those small MTDs, and therefore, a MTD selection protocol is required for managing data transmission between MTDs. A joint buffer-aided MTD selection and power allocation protocol is proposed to maximize the MTDs' sum-rate provided that the induced interference to the cellular network is limited. In particular, in the proposed scheme, at each time slot and each subcarrier, the cognitive M2M network optimally decides on whether to be silent or to select either the relay or one of the MTDs for data transmission. To this end, for each MTD, there exists a buffer at the relay to avoid data loss. The closed-form expressions for the power coefficients of MTDs are calculated. Simulation results show that the proposed policy improves the sum-rate of the MTDs in comparison with the other proposed schemes for M2M communication without buffer.

In this paper, we consider a three-node wireless network comprising of a source and two users. Both users need to decode the transmitted data correctly. User 1 has better position to the source than user 2 most of time slots. User 1 has buffer to store the transmitted information by the source. Thus, in the case of wrong decoding at user 2, user 1 can resend data to user 2 some time slots later. In this paper, we propose a novel incremental relaying based adaptive link selection policy that exploits incremental relaying and buffer to maximize the throughput of the network. That is, based on the channel quality of the available links, each time slot is allocated either to the source or user 1 to transmit data. Both delay constrained and delay tolerant transmission schemes are studied. We model the variation of the buffer at user 1 as a Markov Chain and calculate the outage probability of the proposed policy. Our simulation results show that the proposed scheme achieves higher throughput and lower outage probability compared to the recently proposed link selection policies with or without buffer.