Intelligent transportation networks are increasingly integrating autonomous aerial vehicles (AAVs) to enable services essential for modernizing industries like transportation, logistics, and search and rescue. A significant beneficiary is the Internet of Connected Vehicles (IoCVs), where AAVs play a transformative role in supporting next-generation cellular networks. This paper addresses the challenge of minimizing the cost of delivering content to vehicles on road segments with congested traffic or insufficient infrastructure. Vehicles entering these segments request content from a AAV-hosted library that updates dynamically based on item popularity. Each vehicle submits a request, requiring the AAV to compute an optimal trajectory to maximize operational utility. Given the AAV’s limited energy, we aim to develop an energy-efficient solution. The problem is framed as a joint optimization of caching decisions, AAV trajectory planning, and radio resource allocation, formulated as a mixed integer non-linear programming (MINLP) problem. The environment’s complexity, with random vehicle arrivals and fluctuating content, makes traditional optimization techniques inadequate. To address this, we reformulate the problem as a constrained Markov decision process (CMDP) and employ a primal-dual proximal policy optimization (PPO) algorithm, enhanced by a multi-scale generative transformer (MGFormer) for improved speed and accuracy over traditional deep neural networks (DNNs). Simulation results demonstrate that the proposed framework reduces service costs by up to 30% and achieves robust performance even under high-density traffic and Doppler shifts, validating its superiority over traditional Primal-dual PPO.

Terahertz (THz) massive multiple-input multiple-output (mMIMO) systems offer ultra-high data rates but face significant challenges such as beam squint effects, high power consumption, severe path loss, and signal blockage. Incorporating reconfigurable intelligent surfaces (RIS) can mitigate these issues but complicates channel state information (CSI) acquisition. To address this, we propose a generative artificial intelligence-based hierarchical deep reinforcement learning (GAI-HDRL) framework that jointly performs channel prediction, hybrid precoding at the base station (BS), passive RIS beamforming, and digital combining at the UE to minimize transmit power. The proposed GAI-HDRL efficiently decomposes the optimization into high-level (precoding and combining) and low-level actions (CSI prediction and RIS configuration), achieving fast convergence and improved prediction accuracy. Simulation results confirm its superiority over state-of-the-art methods in terms of performance and computational efficiency, demonstrating practical viability in THz communications.

In this work, we investigate the enhancement of secure quality-of-experience (QoE) in unmanned aerial vehicle (UAV)-assisted multiuser rate-splitting multiple access (RSMA) networks under stringent secrecy constraints. The primary objective is to maximize the aggregate mean opinion scores (MOSs) of all legitimate users while guaranteeing robust physical-layer security against eavesdroppers. To this end, the original secure optimization problem is decomposed into two interconnected subproblems: joint secure beamforming and rate allocation, and UAV trajectory optimization. For the secure beamforming and rate allocation, we employ advanced convexification techniques, including epigraph reformulation, polynomial properties, and norm-bounded channel uncertainty modeling, to ensure both optimal user experience and resilience to information leakage. The UAV trajectory optimization, inherently nonconvex, is further addressed by transforming and approximating the secrecy-related constraints to support secure communication throughout the UAV’s path. Simulation results validate the effectiveness and robustness of the proposed framework in simultaneously enhancing user-perceived QoE and ensuring secure transmission, even under varying eavesdropping threats and network conditions.

We investigate the performance of a multiple re-configurable intelligence surface (RIS)-aided millimeter wave (mmWave) beamspace multiple-input multiple-output (MIMO) system with multiple users (UEs). We focus on a challenging scenario in which the direct links between the base station (BS) and all UEs are blocked, and communication is facilitated only via RISs. The maximum ratio transmission (MRT) is utilized for data precoding, while a low-complexity algorithm based on particle swarm optimization (PSO) is designed to jointly perform beam selection, power allocation, and RIS profile configuration. The proposed optimization approach demonstrates positive trade-offs between the complexity (in terms of running time) and the achievable sum rate. In addition, our results demonstrate that due to the sparsity of beamspace channels, increasing the number of unit cells (UCs) at RISs can lead to higher achievable rates than activating a larger number of beams at the MIMO BS.

Reconfigurable Intelligent Surfaces (RIS) have emerged as a promising technology for enhancing wireless communication by dynamically controlling the propagation environment. Recently, a non-diagonal RIS architecture has been proposed, enabling more advanced signal manipulation by allowing signals impinging on one element to be reflected from another element after appropriate phase-shift adjustment. This paper analyzes the energy efficiency of non-diagonal RIS-assisted wireless communication systems in high- and low-signal-to-noise-ratio (SNR) regime. We derive closed form expressions of the spectral and energy efficiency for both the non-diagonal and its diagonal counterpart, which is used as a benchmark for comparison. Simulation results reveal that non-diagonal RIS systems are the preferred choice for communication systems that prioritize spectral efficiency. Interestingly, for energy efficiency, the selection between diagonal and non-diagonal RIS architectures depends on the received SNR conditions, with diagonal RIS systems excelling at high SNR and non-diagonal RIS systems performing better at low SNR scenarios.

In this work, we consider a typical scenario in a harsh urban propagation environment which is typical for a smart city scenario where multiple reconfigurable intelligent surfaces (RISs) are deployed in different hotspot areas to overcome signal blockage between the base station and users. Our goal is to ensure uninterrupted service availability to users in different hotspot areas regardless of their location. Consistent service availability can be achieved by guaranteeing that each RIS deployed in a hotspot area can support a certain number of users. This plays a critical role in smart city applications in the context of emergency communications and ubiquitous connectivity since the design ensures service availability to as many users as possible in all relevant locations. Taking into consideration the challenges in obtaining channel state information (CSI) given the passive nature of RIS and dynamic environments, we formulate a robust fairness problem to maximize the minimum expected number of served users in proximity to each RIS while considering the available transmit power and the worst-case quality of service (QoS) constraints within the bounded CSI error model framework. The resulting problem is a mixed integer non-convex program which is highly coupled and challenging to solve in polynomial time. Thus, we resort to binary variable relaxation, convex approximation techniques, and alternating optimization to tackle the problem. Additionally, we handle the semi-infinite uncertainty constraints by employing the S-procedure and general sign-definiteness. Simulation results demonstrate the effectiveness of the proposed design in obtaining consistent and reliable service in different hotspot areas compared to the relevant benchmark schemes. In addition, the proposed design shows flexibility in serving users with their target QoS given different channel uncertainty levels.

Learning multi-scale feature representations is essential for dense prediction tasks in autonomous driving. Most existing works are based on U-shaped architectures, where high-resolution representations are progressively recovered by connecting different levels of the decoder with low-resolution representations from the encoder. We observed that rich details from low-level representation and high semantic information from high-level representations are not fully utilized in the cross-stage fusion process. Additionally, current architectures often struggle to extract efficient discriminative feature along object boundaries. To address this issue, we propose CSNet, a generic cross-stage subtraction network that extracts spatial and semantic multi-scale representations through guided contextual feature. This approach allows fine-grained features to refine deeper layers, capturing discriminative high-resolution features while filtering out redundant information. Specifically, we introduce a cross-stage subtraction module (CSM), which consists of three sub-modules: 1) a Short Path Unit, focusing on capturing complementary adjacent information; 2) Medium Path Unit for effective middle-stages features aggregation; and 3) Long Path Unit for redundant information masking and long-range context modeling. Additionally, we propose the Semantic Guided Context Reasoning (SGCR) module to reason and model contextual relations between different subtraction units. CSNet demonstrates consistent performance gains across various semantic segmentation datasets. Our model, CSNet-M, achieves 82.2% mIoU on the Camvid dataset, while CSNet-S and CSNet-M attain 79.6% and 80.5% mIoU accuracy, respectively, on the Cityscapes dataset. These results show that the proposed CSNet has the potential for enhancing real-time semantic segmentation in autonomous driving applications, offering improved accuracy and efficiency in diverse urban scenarios. The source code for this work will be published athttps://github.com/mohamedac29/CSNet.

In this study, we investigate minimization of energy consumption in multi-UAV assisted networks. We formulate an energy minimization optimization problem with UAV trajectory design, content fetching, power allocation and content placement constraints. The problem is a mixed integer nonlinear programming (MINLP); therefore, we convert the formulated problem into semi-Markov decision process (SMDP). To tackle this SMDP optimization challenge, we introduce an option-based hierarchical deep reinforcement learning (OHDRL) approach. We designate UAV trajectory planning and power allocation as the low level action space, and content placement and content fetching as the high level option space. Through simulations, we demonstrate the effectiveness of the proposed OHDRL method.

In this paper, we study the resource allocation problem in multi-RIS-aided outdoor-to-indoor communications. Specifically, we aim to provide geographical fairness to ensure that users in different hotspot areas in a smart city can be served regardless of their location. We consider a scenario where RISs are deployed to extend coverage to indoor users in different buildings where there is limited network accessibility. This design is crucial in smart cities in the context of emergency communication and ubiquitous connectivity since it ensures service availability to as many users as possible independently of the locations. Thus, to achieve geographical fairness, we formulate a max-min fairness problem to maximize the minimum number of users served by each RIS by jointly optimizing the active precoding and RIS-based beamforming subject to power and quality of service constraints. The geographical location of users is directly linked to the RIS which means that users are served by the RIS closest to them. In this case, we ensure that a certain number of users can be supported by each RIS. The formulated problem is a mixed integer nonlinear program, which is challenging to solve directly using methods of convex optimization. Accordingly, we propose an efficient successive convex approximation-based alternating optimization algorithm to tackle the complexity of the formulated problem. The presented results show the performance gain of the proposed design in providing geographical fairness compared to the relevant benchmark schemes.

Simultaneously transmitting and reflecting reconfigurable intelligent surface (STAR-RIS) is an emerging technology that enables full-space ($\mathbf{3 6 0}$ degrees) coverage on both sides of the surface. To harness the benefits of the dynamic configuration of STAR-RIS while avoiding co-channel interference, we investigate the performance of a multi-user network assisted by STARRIS. In this setup, users are divided into multiple groups, each comprising two users located on opposite sides of the STARRIS. Orthogonal time resources are allocated to each group such that the groups are served sequentially. Based on the Gamma moment matching method, we introduce a Gamma distribution to model the product of Rician, Rayleigh and mixed fading STAR-RIS channels. We then derive exact closed-form expressions for the outage probability and diversity order per user in the proposed system model. Moreover, simulation results are provided to substantiate the analytical derived expressions. Our findings highlight a reliability trade-off associated with the number of grouped users per time slot, STAR-RIS elements, and the user targeted data rates. This balance is crucial for optimizing network performance.

Coexisting satellite and terrestrial networks present a unique set of challenges and opportunities when the two networks share the same spectrum. One of these challenges is the desired signal recovery in such interference-limited scenario. In this work, we design a signal recovery scheme in coexisting satellite and terrestrial networks. We formulate an optimization problem and propose a diffusion model to perform signal recovery. The proposed diffusion model leverages the denoising mechanism to recover the signals from noisy and distorted signals. The proposed diffusion model consists of encoder to encode the input to the latent space, U-Net for denoising, attention block to integrate different relevant feature to create better context for signal recovery, and decoder to deliver the recovered signal.

The Internet of Vehicles (IoV) is undergoing swift advancements in capability and intelligence, poised to facilitate a diverse range of innovative applications. Within the IoV, edge learning empowers intelligent applications and services by leveraging data -driven tasks. Therefore, in this paper, we propose optimizing the edge learning error prediction within an edge-supported Non-Orthogonal Multiple Access (NOMA) in task-oriented communications. Specifically, we consider three autonomous vehicle perception tasks, called: object detection, traffic sign, and weather classification. For this purpose, we propose a novel approach based on the Particle Swarm Optimization (PSO) algorithm to jointly minimize the edge learning error and optimize power allocation variables. Moreover, we investigate alternative benchmark schemes, including Quantum Particle Swarm Optimization, Cuckoo Search, and Butterfly Op-timization algorithms. Satisfactorily, our simulations substantiate the superiority of the PSO algorithm over the baseline schemes, delivering superior performance with reduced computation time.

Next generation networks will involve huge number of industrial internet of things (IIoT) sensors which require reliable connectivity with low latency to manage the data transmission and processing. The design of these networks entails a lot of challenges. This article describes the 3D placement of multiple unmanned aerial vehicles (UAVs) in an IIoT network that supports non-orthogonal multiple access (NOMA). UAVs act as decode and forward (DF) relays. The 3D UAV placement problem is formulated which is highly non-convex in the coordinates. Therefore, we employ an improved adaptive whale optimization algorithm (IAWOA) to handle the problem. Even with its improved performance, IAWOA is not suitable for real-time application. Hence, we propose path aggregation network (PANet) to handle the 3D UAV placement. The simulation results show that PANet is more suitable for the online-learning.