This comprehensive survey paper examines the applications of Artificial Intelligence (AI) in Unmanned Aerial Vehicle (UAV)-enabled wireless networks. With the increasing demand for efficient and adaptive communication systems, the integration of AI with UAV networks promises to revolutionize various aspects of wireless communication. The paper first outlines the background and motivation behind AI integration, highlighting the potential for enhanced network performance, autonomy, and adaptability. It then delves into the key AI applications across different network layers, including data sensing and collection, placement and trajectory optimization, radio resource management, routing and topology control, edge computing and caching, as well as security and privacy enhancement. For each application, the paper discusses relevant AI techniques, main findings, optimization objects, and the potential benefits and challenges. The survey also identifies open issues, such as the practical implementation gap, standardization issues, and real-world application barriers, and proposes future directions to address these challenges and further advance the field. In conclusion, the integration of AI with UAV-enabled Wireless Networks (UWNs) holds tremendous potential for transforming wireless communication, enabling new applications and services with unprecedented capabilities.

Unmanned Aerial Vehicles (UAVs) are increasingly deployed across military and civilian domains due to their operational flexibility, low maintenance costs, and high mobility. With the growing complexity of UAV applications and tasks, robust support from computing power networks is essential. These networks, acting as resource integration paradigms, furnish UAVs with pooled resources to tackle extensive computational demands. In this paper, we develop a framework for trading computing power resources, modeling the transaction process through a three-stage Stackelberg game to facilitate sequential decision-making. We theoretically demonstrate the existence of a Nash equilibrium and introduce a Dynamic Game Reinforcement algorithm to identify optimal strategies. Our experimental results affirm the framework’s efficacy and the superior performance of our algorithm. Additionally, we explore how variables like UAV quantity and network congestion influence the market dynamics of the computing power network.

With the continuous development of communication and computation technologies, large bandwidth services place higher demands on computing capacity and throughput in High-Speed Rail (HSR) scenarios. On the one hand, Millimetre-Wave enables high data transmission rates, but leads to high Doppler frequency deviation as well as large path loss. On the other hand, the development of Mobile Edge Computing greatly alleviates user computing congestion. In this paper, we propose the Adaptive Joint Communication and Computation Resource Allocation scheme to solve the energy optimization problem of a dual-band UAV and Mobile Relay relay-assisted HSR offloading system. This scheme works well to optimize the performance of the system through resource allocation. In addition, since the optimization problem is modeled as a Mixed-Integer Nonlinear Program, we propose the Pre- and Post-state connected Parameterized Deep Q-Network algorithm, which is based on Deep Reinforcement Learning approach, for offloading decision and bandwidth resource allocation. Simulation results show that the proposed algorithms result in lower system energy consumption, while ensuring a high task completion rate as well.

Unmanned Aerial Vehicle (UAV)-assisted Vehicular Edge Computing Networks (VECNs) have emerged as a promising solution to enhance service quality for ground vehicle users. However, the growing demands from users and the limited computing and storage resources of UAVs present significant challenges in designing an efficient edge service caching scheme to minimize latency. Moreover, the integration of service caching and task offloading complicates the support of complex tasks by a single UAV. To address these challenges, this paper proposes a novel two-tier UAV-assisted VECNs framework. In this framework, multi-rotor UAVs function as hovering nodes for computational offloading, while a fixed-wing UAV serves as a mobile auxiliary cloud platform, forming a cohesive UAV group. User tasks are structured into a task chain based on the available UAVs. We integrate a joint service chain caching and task offloading scheme that considers UAV computing and storage capacities, duplicate caching, and dynamic transmission latency. To optimize task chain completion latency, we propose an Attention-based Multi-Agent Deep Q-Network (A-MADQN) algorithm. This algorithm incorporates an attention mechanism to narrow the UAV selection space, enabling the selected UAVs to collaboratively make caching and task offloading decisions. Numerical results demonstrate that the proposed algorithm significantly enhances system processing efficiency and reduces task completion latency compared to the benchmark approaches.

This paper exploits multi-modal Physical (PHY)-layer features in terms of artificial fingerprint, In-phase/Quadrature (IQ) imbalance and Angle of Arrival (AoA) to propose a novel PHY-layer authentication framework for a Millimeter Wave (mmWave) Multiple-Input Multiple-Output (MIMO) Unmanned Aerial Vehicle (UAV)-enabled communication system. First, we resort to the AoA-based spatial fingerprint to effectively address the challenge of channel fingerprint instability induced by high-speed UAV mobility. To further enhance the low discriminability of hardware fingerprints caused by refined manufacturing techniques, artificial Gaussian noise is injected into the transmission signals to assist the receiver in better distinguishing between legitimate and illegitimate UAVs. Then, we jointly combine with inherent IQ imbalance and AoA features to design a hybrid authentication scheme and thus construct a multi-dimensional fingerprint space for a comprehensive characterization of UAV identities. To theoretically evaluate the effectiveness of the proposed authentication framework, the analytical closed-form expressions of performance metrics like false alarm and detection probabilities are also exactly derived based on the statistical signal processing technology and composite hypothesis testing. Finally, we provide large simulation results to validate the correctness and feasibility of the proposed theoretical models, and also discuss the relation between system security and communication service quality under different artificial fingerprint level.

In emergency communication scenarios, exploiting Unmanned Aerial Vehicles (UAVs) as relays to provide wireless communication services for ground users has emerged as a promising application. A key challenge in this resource-constrained application is deploying the minimum number of UAVs to form an aerial backhaul network to ensure coverage, which composes the Number and Placement Optimization for the Backhaul-Aware Network Deployment (NPO-BAND) problem. In this paper, we first formulate the NPO-BAND problem based on the geometric disk coverage model. Then, we propose a low-complexity heuristic method to solve this NP-hard problem. The proposed method contains a Very Important Point-Choosing (VIPC) strategy and a Backhaul-Aware Local Coverage (BALC) algorithm. Specifically, the VIPC strategy weighs up the backhaul connectivity constraint and the ground user coverage to choose the VIP, while the BALC algorithm solves the extended 1-center problem to determine the deployment location of each UAV. Simulation results show that the proposed method can effectively reduce the number of deployed UAVs, saving up to 25%-50% of that compared to existing methods across varying numbers and area sizes in clustered distribution patterns of ground users.

The integration of communication networks and artificial intelligence enables the effective collection of data over smart building networks, facilitating more accurate predictions of Building Energy Consumption (BEC). However, existing schemes for BEC prediction suffer from limited dynamic adaptability, risks of privacy leakage, and the inability to accurately capture actual energy consumption patterns. To improve prediction accuracy while ensuring privacy and dynamic adaptability, we propose a novel BEC prediction design that incorporates dynamic threshold participation and privacy-preserving mechanisms. Specifically, we design a three-tier network architecture integrated with threshold participation tokens to support dynamic access and dropout of building entities during the BEC model construction process. Furthermore, we develop a Context-Aware Transformer (CAT) network integrated into Federated Learning (FL) to enhance feature sensitivity and facilitate the sharing of knowledge derived from Internet of Things (IoT) data and BEC features. Finally, we evaluate the performance of our design using real-world data, and the results demonstrate that our design achieves superior performance in distributed BEC prediction.

This review provides a comprehensive survey of the most recent developments in metasurfaces for applications in domains including wireless-optical switching and communications. In particular, we focus on discussion of multi-parameter optical field regulation and potential applications in system performance enhancement. By designing nanostructured arrays with specific geometries, metasurfaces can be used to effectively manipulate parameters including phase, amplitude, and polarization, thereby enabling the switching, transmission, testing, analysis, and processing of optical signals. Notably, the introduction of phase-change materials offers a novel approach that allows metasurfaces to achieve more flexible wireless-optical switching at higher speeds. In wireless-optical communication systems, multiplexing of the different degrees of freedom of the light beams can improve the data transmission capacity and rate significantly. Finally, we present our own metasurface design with its unique passive parallel beam splitting capacity, and we demonstrate the superiority of this design in applications including wireless-optical inter-rack connections in data centers and industrial inspection based on optical cross-connectors.

In this paper, a multi-agent deep reinforcement learning-based joint resource allocation and task offloading control technique is proposed to minimize the energy consumption of user equipment (UE) supporting 5G ultra reliability low latency communications (URLLC) and enhanced mobile broadband (eMBB) services while meeting strict quality of service (QoS) requirements in 5G multi-radio access technology (RAT) networks. An optimization problem involving transmission power, channel resource, user association, offloading rate, and central processing unit (CPU) frequency is formulated using a queueing system-based mathematical design to support services with different characteristics while minimizing the energy consumption. It is proven in this paper that this problem is nondeterministic polynomial (NP) hard, in which multi-agent deep reinforcement learning (DRL) is used to solve the problem. To increase the learning efficiency and stability of deep reinforcement learning, prioritized experience replay (PER) and delayed target network and policy updates are applied. Simulation results show that the proposed scheme provides an improved energy consumption performance compared to the benchmarked schemes.

Improving the energy efficiency of information transmission is very critical to the development of future Internet of Things (IoT). Considering the sporadic characteristics for IoT transmissions, the energy consumption of a specific transmission session significantly varies with channel condition and Quality of Service (QoS) requirements. In this study, we focus on the analysis and optimization for wireless relaying communications’ statistical energy consumption. Particularly, we investigate a wirelessly-powered DF relaying communication system. Under Time Switching (TS) and Power Splitting (PS) modes, we analyze and minimize the statistical energy consumption of transmitting a fixed amount of data using mathematical analysis. Through showing some selected numerical examples, we discuss various design tradeoffs. These results will provide some important guidelines for the design of green IoT communication systems.

Dynamic graphs are increasingly utilized for detecting anomalous behaviors in nodes within the Internet of Things (IoT). Graph generative models play a pivotal role in addressing the challenge of imbalanced node categories in dynamic graphs. However, these models encounter several limitations, including the monotonicity of adjacency relationships, the complexity in constructing multi-dimensional features for nodes, and the absence of an end-to-end method for generating multiple categories of nodes. In this study, we introduce a novel graph generation model, designated as Conditional Graph Generation Model (CGGM), aimed specifically at generating samples from minority classes. The architecture comprises two principal modules: a conditional graph generation module and a graph-based anomaly detection module. The generative module adjusts to matrix sparsity by downsampling a noise adjacency matrix and integrates a multi-dimensional feature encoder based on multi-head self-attention to capture latent feature dependencies. Furthermore, a latent space constraint coupled with distribution distance is utilized to approximate the latent distribution of real data. The graph-based anomaly detection module employs the generated balanced dataset to predict node behaviors. Extensive experiments demonstrate that CGGM surpasses contemporary state-of-the-art methods in accuracy and divergence. The results further reveal that CGGM can produce diverse data categories, thereby enhancing the performance in multi-category classification tasks.

Intelligent Connected Vehicles (ICVs) generate massive heterogeneous multi-modal data during operation, and due to the limited computing resources on board, graded data encryption protection is of great significance for balancing data security and efficient utilization. However, the current data grading processes struggle to address the evolving inference attacks and dynamic operational environments, and traditional grading approaches relying on static expert judgment or information-theoretic metrics. To bridge this gap, this paper proposes a novel inference strength-driven data grading framework, where inference strength quantifies the susceptibility of one dataset to infer another through adversarial reasoning. The framework employs a systematic methodology combining graph theory, optimization, and Large Language Models to construct an inference library and calculate inference strength. The framework also provides a PageRank-based algorithm to generate interpretable data grading lists for both static policy and vehicle-end application, prioritizing core data protection while respecting computational constraints. Validated on the Audi A2D2 dataset and real vehicle controller, our approach demonstrates improved protection utility compared to default grading baselines. The results highlight its potential to enhance data security in ICVs through prioritized protection of core data under computational constraints.