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.

This work presents an energy-efficient reinforcement learning (RL) solution based on Neuromorphic Computing (NC) to enable opportunistic spectrum access in partially observable wireless environments. To improve the energy efficiency of the underlying spectrum access strategy, we explore Neuromorphic Computing and adopt spiking neural networks. Additionally, the time-dependent aspect of the problem and the necessity for sample efficiency drive us to liquid state machines, a variant of reservoir computing. Nevertheless, a priori hyperparameter optimization of the spiking reservoir is essential for handling state- and time-varying inputs in RL agents; yet, this can undermine model robustness and impede deployment. In response, we examine homeostatic regulation for self-modulating the small-world reservoir’s dynamics, thereby maintaining desired near-chaotic behavior throughout operation. The RL model for opportunistic spectrum access is evaluated under both dynamic spectrum access (DSA), where agents identify temporal spectrum holes for transmission, and spatiotemporal spectrum access (SSA), where agents also aim to minimize coverage overspill without coordination or sharing location data. Numerical analysis demonstrates that the proposed model outperforms existing learning models in the literature for both DSA and SSA, while significantly reducing power consumption.

To fully exploit the benefits of spatial multiplexing gains within Frequency Division Duplex (FDD) Multiple-Input Multiple-Output (MIMO) systems, the development of a robust Channel State Information (CSI) feedback compression methodology with low over- the-air CSI overhead while achieving high reconstruction accuracy is critical. Such methodology must be able to effectively reduce communication overhead without compromising system level performance. Traditional codebook-based approach, as outlined in the 3rd Generation Partnership Project (3GPP) standards encounters significant challenges to balance air-interface overhead and computational complexity. In this paper, we introduce a novel and adaptable Deep Leaning (DL) based CSI codebook technique leveraging vector quantization known as the CSI-Look-Up Table (CSI-LUT). Our numerical results show that the CSI - L UT has potential in reducing > 90 % of CSI overhead for max rank 1 while still achieving better average reconstruction accuracy and system level performance than traditional codebook-based approach. We anticipate that such advancements will play a pivotal role in enhancing the efficiency and performance of CSI feedback in MIMO systems, contributing significantly to the evolving landscape of 5G and beyond.

Beam management (BM) plays a crucial role in maintaining reliable communication links in highly dynamic scenarios. To enhance BM performance, the 3rd Generation Partnership Project (3GPP) is actively exploring the use of artificial intelligence (AI) and machine learning (ML) for beam prediction in the evolution toward sixth-generation (6G) communications. The main goals of these 3GPP-based standard studies are to minimize the overhead from reference signals (RSs) and to reduce the number of beam sweepings at the user equipment (UE), which arise due to frequent beam measurements caused by UE movement and rotation. This paper delves into an AI/ML algorithm design that supports spatial domain beam prediction tailored for BM in 6G. This includes forecasting the optimal beam (pairs) and anticipating beam changes. Simulations are based on a data-driven strategy that uses RS receive power (RSRP) measurements as input for fast beam pair prediction with an advanced convolutions neural network (CNN) architecture. Results indicate that the proposed AI/ML model outperforms conventional BM techniques, reducing beam sweeping overhead and thereby validating the AI/ML's potential in BM. Additionally, our proposed algorithm achieves up to 40.58% higher beam prediction accuracy and improves the mean RSRP difference of the predicted best beam pair by up to 2.89 d$B$.

Dynamic spectrum access (DSA) has been introduced as a promising technology that allows a secondary system to access the licensed spectrum of the primary system to improve spectrum utilization. In this paper, we introduce Fed-MADRL by incorporating federated learning (FL) and multi-agent deep reinforcement learning (MADRL) to design a collaborative DSA strategy. Our Fed-MADRL scheme employs FL to enable multiple users to collaboratively optimize the system goal without sharing their training data. By keeping all the training data at the user end, FL improves the communication efficiency and strengthens user data privacy. To further reduce the communication overheads, each user only shares quantized information. We provide the convergence analysis to characterize the trade-off between the communication efficiency and the system performance. In particular, we show that the introduced method converges at a rate$\mathcal {O}(1/K^{1/4})$, where$K$is the number of FL iterations. To the best of our knowledge, Fed-MADRL is the first work that utilizes FL in DSA networks under quantized communication. Performance evaluation results show that the introduced Fed-MADRL method outperforms the independent learning method and achieves comparable performance with the centralized MADRL method, which requires much higher communication overheads.

Dynamic spectrum access (DSA) has emerged as a promising solution for spectrum usage enhancement by allowing opportunistic access of secondary users to the licensed spectrum. In this paper, we introduce Fed-MADRL, a collaborative DSA technique that exploits both federated learning (FL) and multiagent deep reinforcement learning (MADRL). FL allows numerous users to collaborate on the system goal optimization without sharing their training data. By keeping all training data at the user's end, FL simultaneously enhances communication efficiency and protects data privacy. To further reduce communication costs, each user in Fed-MADRL only shares quantized data. To the best of our knowledge, Fed-MADRL is the first effort that employs FL in DSA networks with quantized communication. Simulation results show that the introduced Fed-MADRL approach beats the independent learning method and provides comparable results to the synchronous FL method, which involves significantly greater communication overheads.

The performance of federated learning systems is bottlenecked by communication costs and training variance. The communication overhead problem is usually addressed by three communication-reduction techniques, namely, model compression, partial device participation, and periodic aggregation, at the cost of increased training variance. Different from traditional distributed learning systems, federated learning suffers from data heterogeneity (since the devices sample their data from possibly different distributions), which induces additional variance among devices during training. Various variance-reduced training algorithms have been introduced to combat the effects of data heterogeneity, while they usually cost additional communication resources to deliver necessary control information. Additionally, data privacy remains a critical issue in FL and, thus, there have been attempts at bringing Differential Privacy to this framework as a mediator between utility and privacy requirements. This article investigates the tradeoffs between communication costs and training variance under a resource-constrained federated system theoretically and experimentally, and studies how communication reduction techniques interplay in a differentially private setting. The results provide important insights into designing practical privacy-aware federated learning systems.