In emergency communications constrained by bandwidth limitations, unmanned aerial vehicle (UAV)-based coded caching presents a promising approach for the efficient dissemination of high-bandwidth-demanding services. This paper focuses on content download and content repair in aerial caching networks, where UAVs deliver contents to both ground users and invalid UAVs. To address potential data loss due to limited power and high mobility, fault-tolerant codes are utilized to maintain data availability and reliability. Initially, we derive the expressions of communication cost and success rate for content download and content repair. The size of coded fragments, determined by the coding design, affects both the success rate and transmission cost, while the resource allocation, which influences the cooperative relationships, also impacts these two aspects. The interplay between coding design and resource allocation is thus established to jointly optimize the overall performance. Then, we design a joint optimization problem of erasure coding schemes, coding parameters, matching relations, and UAV trajectories to maximize the overall success rate. Moreover, we propose a hierarchical multi-agent parameterized deep Q-network (H-MA-PDQN) algorithm integrating a dual-component structure for long-term coding and immediate resource allocation to solve the mixed integer nonlinear programming (MINLP), and each agent employs a PDQN with hybrid discrete-continuous action space. Simulation results demonstrate that our proposed H-MA-PDQN algorithm increases the success probability by 26.7% and 66.7% and reduces the transmission cost by 27.3% and 42.9% compared with the DQN and greedy-based strategies, respectively.

In emergency scenarios, unmanned aerial vehicles (UAVs) show significant potential as aerial base stations (BSs) to establish reliable communication links and provide localization services through integrated air-ground collaboration. This paper proposes a novel energy-efficient collaborative framework based on the solo-UAV-rescuer cooperative (SURC) paradigm, which synergistically enhances both communication capacity and localization accuracy. From a system optimization perspective, we formulate an optimization problem using a normalized combination of three critical metrics: achievable data rate, localization accuracy, and energy consumption. Specifically, to maximize the system’s utility, we design a signal perception-based localization method that incorporates angle-of-arrival (AOA) localization information for guidance, and develop a beamforming scheme to facilitate high data rate communication. Building on these methods, we propose a deep reinforcement learning (DRL)-based synergistic communication and localization reinforcement (SYNCORE) approach that dynamically optimizes three key operational parameters: UAV trajectory planning, flight time, and transmission power control, achieving reliable services with energy-efficient operation. Based on the simulation results, we validate that the proposed scheme enhances communication and localization performance, while also improving energy efficiency, surpassing the baseline schemes.

In mountainous environments, terrain occlusion causes non-line-of-sight (NLoS) transmission, significantly reducing the signal propagation range. To improve emergency rescue efficiency, a mobile unmanned aerial vehicle (UAV)-based integrated localization and communication (ILAC) network should be deployed to achieve optimal performance through adaptive trajectory planning and resource allocation. However, irregular and unpredictable terrain occlusions, coupled with dynamic users, make traditional optimization ineffective and reinforcement learning (RL) inefficient. To address these challenges, this paper proposes a hybrid action space soft Actor-Critic with geographic information-based state aggregation (GeoAgg-HSAC) decision-making scheme. First, an RL state aggregation method based on graph contrastive learning is designed. Through a pre-trained graph neural network (GNN), the UAV network states experiencing the same occlusion are mapped to similar low-dimensional representations. This method reduces the state dimension and allows similar states to share policy experience, thereby improving sample efficiency and accelerating convergence. A hybrid action space SAC network is then designed, which simultaneously makes decisions for continuous UAV trajectories and discrete resource allocation. Finally, a simulation environment based on real mountain terrain and wireless data is built for the experiment. The experimental results show that the proposed scheme has significant advantages for optimizing communication and localization performance.

Next-generation mobile networks are expected to facilitate fast AI model deployment on end devices (EDs). By enabling collaborative model caching across EDs, mobile networks can efficiently support distributed AI inference services through device-to-device (D2D) cooperation. In this paper, we investigate a D2D-assisted model caching and collaborative computing framework that aims to balance the trade-off among inference delay, accuracy, and energy consumption by managing model caching, data offloading, and computation resources efficiently during the provisioning of diverse AI services. Specifically, considering the AI performance is constrained by multi-dimensional resources, a new metric named Service Hit Rate (SHR) is proposed to decouple the joint impacts of computation, communication, and caching resources on service success. Aiming to maximize the SHR, we propose a matching-aided multi-agent reinforcement learning (MARL) framework. First, a hierarchical bipartite matching algorithm is utilized for model deployment and helper assignment. Then, an attention-based MARL algorithm is employed to allocate computation resource for models cached on the same EDs. Simulation results demonstrate that the proposed algorithm significantly improves the system SHR for AI services.

Integrated sensing and communication (ISAC) has been recognized as a key technology in the low-altitude economy. Leveraging the flexibility and high maneuverability of unmanned aerial vehicles (UAVs), we propose a multi-UAV enabled ISAC system to provide communication and tracking services for multiple ground mobile targets (GMTs). By jointly optimizing the communication scheduling and UAV trajectory, we aim to maximize the system rate while guaranteeing tracking demands, subject to anti-collision and energy consumption constraints. Specifically, we decompose the original non-convex optimization problem into two subproblems and develop an efficient approach based on successive convex approximation (SCA) to solve them iteratively. Numerical results demonstrate that the proposed multi-UAV enabled system achieves superior communication performance through the joint optimization of scheduling and trajectories, while fulfilling real-time tracking requirements.

In mountainous emergency rescue operations, due to terrain occlusion, signals are often in a non-line-of-sight (NLoS) transmission state, resulting in a significant reduction in signal propagation distance. It is necessary to deploy an integrated positioning and communication (ILAC) network based on unmanned aerial vehicles (UAVs) to achieve optimal performance through trajectory planning and resource allocation. However, the complex and unpredictable terrain occlusion, coupled with dynamic user behavior, makes it difficult for traditional optimization methods to work, while the direct application of reinforcement learning (RL) is inefficient. To address these challenges, this paper proposes a hybrid action space soft actor-critic (GeoAgg-HSAC) decision framework based on geographic information state aggregation. First, a state aggregation method based on graph contrastive learning is designed. In this method, a pre-trained graph neural network (GNN) is used to map the UAV network state under similar occlusion conditions to a corresponding low-dimensional representation, thereby reducing the state dimension and allowing similar states to share policy experience, which improves sample efficiency. Then, a hybrid action space SAC network is constructed that can simultaneously make decisions on continuous UAV trajectories and discrete resource allocation. Experiments based on real mountain terrain and wireless data show that the proposed approach has significant advantages in optimizing communication and positioning performance.

Edge-cloud collaborative inference is a widely adopted approach in edge intelligence, especially for latency-sensitive tasks such as drone inspection, augmented reality, and disaster relief. To adapt to varying network bandwidth and limited computational capacity, model splitting methods divide Deep Neural Networks (DNNs) at specific split points into edge and cloud segments. However, current methods suffer from high switch latency and memory overhead, and are incompatible with Feature Pyramid Network (FPN)-based multi-branch models. To address the above issues, we propose a dynamic model splitting framework (DMSF) for edge-cloud collaborative inference. We design switchable compress–recover modules after each layer, enabling a single model to support all split points and significantly reduce memory overhead. The optimal split point is selected based on bandwidth and edge computational capacity, and the corresponding compress–recover path is activated for fast switching. For FPN-based multi-branch models, we remove the multi-scale branches from the backbone to the FPN and present hierarchical compensation modules in DMSF, making the model more suitable for splitting while maintaining perception accuracy. Experimental results show that DMSF reduces memory overhead by 75%, achieves millisecond-level switching, and outperforms the existing method with 32.0% lower latency and 12.2% higher perception accuracy.

In complex environments such as forested and mountainous regions, traditional Kalman filtering (KF) and its derivative methods for localization suffers from severe performance degradation due to inaccurate state transition model and time-varying non-zero-mean observation noise. To address these challenges, this paper makes two key contributions. First, we construct real-world UAV-based dynamic localization datasets collected in representative forest environments of China using integrated sensing and communication (ISAC) mesh devices. Second, we propose a novel framework, termed learning-enhanced adaptive Kalman filtering with non-line-of-sight (NLoS) ranging compensation (LAKF-NRC). This framework preserves the theoretical optimality of the KF in fusing predictions with measurements, while integrating two neural networks into the filtering process to jointly improve measurement and prior prediction accuracy. For measurement correction, a state-feedback TimesNet is designed to extract frequency-domain features for correcting NLoS-induced ranging errors and predicting the observation noise covariance. For prior prediction optimization, a state-feedback long short-term memory (LSTM) is developed to learn nonlinear state transition model and process noise covariance. Extensive experiments demonstrate that LAKF-NRC significantly outperforms extended Kalman filter (EKF), unscented Kalman filter (UKF) and advanced KalmanNet method in dynamic localization tasks, providing a physically consistent and generalizable solution for real-world state estimation.

Model-based shallow machine-learning methods and data-driven deep-learning (DL) methods have been advanced to address hyperspectral and multispectral image fusion (HS–MS fusion). Nonetheless, model-based approaches, which meticulously craft regularization terms within optimization models using hand-engineered priors, often struggle to pinpoint the optimal solution efficiently. DL-based methods, which train on extensive datasets to learn a non-linear mapping for generating a high spatial and spectral resolution image (HS2I), exhibit limited generalization capabilities when applied to novel and diverse test datasets. To improve the generalization ability and optimization efficiency of the existing HS–MS fusion methods, a novel deep low-rank prior (DLRP)-based self-supervised HS–MS fusion approach is devised. It incorporates a low-rank learning paradigm to produce the fused HS2I, subject to the constraints imposed by loss functions. Instead of deriving the solution via solving a low-rank approximation optimization problem, deep image prior (DIP) learned by a two-dimensional CNN and a one-dimensional CNN are integrated as the low-rank prior.

Radio maps enrich radio propagation and spectrum occupancy information, which provides fundamental support for the operation and optimization of wireless communication systems. Traditional radio maps are mainly achieved by extensive manual channel measurements, which is time-consuming and inefficient. To reduce the complexity of channel measurements, radio map estimation (RME) through novel artificial intelligence techniques has emerged to attain higher resolution radio maps from sparse measurements or few observations. However, black box problems and strong dependency on training data make learning-based methods less explainable, while model-based methods offer strong theoretical grounding but perform inferior to the learning-based methods. In this paper, we develop a deep unrolled low-rank tensor completion network (DULRTC-RME) for radio map estimation, which integrates theoretical interpretability and learning ability by unrolling the tedious low-rank tensor completion optimization into a deep network. It is the first time that algorithm unrolling technology has been used in the RME field. Experimental results demonstrate that DULRTC-RME outperforms existing RME methods.

The increasing deployment of unmanned aerial vehicles (UAVs) as mobile communication relays in urban environments necessitates accurate 3-D modeling of complex urban areas for optimal communication. Current practices involve LiDAR-based 3-D scanning to generate point cloud data; however, sensor limitations and adverse weather conditions may compromise data quality. This study proposes a new multimodal point cloud and image fusion completion network (PIFC-Net) based on a generative adversarial network (GAN), specifically tailored for large-scale urban environments. The experimental study tested five different building shapes and various objects, and the results commend the network for its effectiveness in enhancing the quality and efficiency of point cloud completion.

Electricity data plays a pivotal role in power management systems. Smart meters, as key tools for recording this data, often encounter anomalies due to meter malfunctions, operational errors, or unauthorized electricity usage, all of which jeopardize the stability of power grids. To this end, we propose the future-guidance anomaly detection network, called FG-Net, designed for real-time analysis of electricity time series data. FG-Net is designed to memorize historical data and assimilate future data, ensuring comprehensive learning of complete data information. Specifically, we leverage the comprehensive data insights gained from a complete information network to guide the predictions of the historical information network. Subsequently, we developed a self-matching feature guidance (SFG) strategy that harnesses the strengths of the complete information network to offset the limitations of the historical information network, thus providing effective guidance. The experimental results on two power grid time series datasets with different anomaly volatility, the Low Carbon London dataset and the Ausgrid Solar Home dataset, demonstrate the proposed anomaly detection method’s accuracy and efficiency.

Drone-based remote sensing has become pivotal for high-resolution dynamic monitoring. However, the differences between day and night modes will trigger a mismatch in multi-scale object features under extreme lighting conditions. In this paper, we propose a dual-stream hierarchical interaction network for multimodal drone object detection, called DHANet, which enhances the distinguishability between multi-scale objects and background for each modality. Specifically, DHANet is designed with a Modality-Adaptive Asymmetric Attention Module (M-AAM) that enhances object-level semantic representations through global and local attention mechanisms. The M-AAM employs global context attention and local positional attention to replace conventional multi-scale context extraction, thereby effectively integrating spatial-channel information of objects. Furthermore, the network is equipped with a Multimodal Scale-Attentive Convolution (M-SC) module that dynamically generates modality-specific feature aggregation weights. This design enables global cross-modality information fusion while reducing computational complexity. Experimental results on two multimodal remote sensing benchmark datasets (DroneVehicle and VEDAI) and two natural datasets (LLVIP and FLIR) demonstrate the robustness and generalizability of DHANet. The codes will be openly and freely available at https://github.com/Victoria-xin1009/-IEEE TGRS DHANet for the sake of reproducibility.

Distributed DNN inference is becoming increasingly important as the demand for intelligent services at the network edge grows. By leveraging the power of distributed computing, edge devices can perform complicated and resource-hungry inference tasks previously only possible on powerful servers, enabling new applications in areas such as autonomous vehicles, industrial automation, and smart homes. However, it is challenging to achieve accurate and efficient distributed edge inference due to the fluctuating nature of the actual resources of the devices and the processing difficulty of the input data. In this work, we propose DistrEE, a distributed DNN inference framework that can exit model inference early to meet specific quality of service requirements. In particular, the framework firstly integrates model early exit and distributed inference for multi-node collaborative inferencing scenarios. Furthermore, it designs an early exit policy to control when the model inference terminates. Extensive simulation results demonstrate that DistrEE can efficiently realize efficient collaborative inference, achieving an effective trade-off between inference latency and accuracy.

In this paper, we present a communication and localization co-design (CLCD) scheme tailored for unmanned aerial vehicle (UAV) assisted emergency networks, with the goal of enhancing rescue operation efficiency and network resource utilization. Specifically, we delve into the mechanism of mutual benefit between communication and localization in UAV wireless networks, establishing a utility function that combines communication rate and localization error. Building on this, we develop a beamforming scheme to facilitate high data rate communication, incorporating angle of arrival (AOA) localization information for guidance. A deep reinforcement learning (DRL)-based communication and localization coordinated optimization (CLCO) algorithm is further proposed to optimize the UAV trajectory and the transmit power in real-time, guaranteeing reliable communication and localization services. Extensive simulation results validate our approach, showcasing up to a 40% improvement in joint utility compared to baseline schemes.

With bandwidth constraints in emergency communications, device-to-device (D2D)-based coded caching emerges as a solution for efficiently transmitting high-bandwidth-demanding services. In this article, we investigate content sharing between emergency vehicles and mobile users via D2D communications in the emergency networks by exploiting coded caching schemes. The joint optimization of coding schemes, coding parameters, and matching relations is proposed to maximize the overall success probability of content sharing while minimizing the overall transmission cost. The interplay between coding parameters optimization and resource allocation is investigated by both download and repair process. Moreover, we propose a multi-agent parameterized deep Q-network (MA-PDQN) algorithm to solve the mixed integer nonlinear programming (MINLP), with each agent employing a PDQN with hybrid discrete-continuous action space. Simulation results show the effectiveness of the proposed algorithm in improving success probability and reducing transmission cost.

In achieving effective emergency response, the timely acquisition of environmental information, seamless command data transmission, and prompt decision-making are crucial. This necessitates the establishment of a resilient emergency communication dedicated network, capable of providing communication and sensing services even in the absence of basic infrastructure. In this paper, we propose an Emergency Network with Sensing, Communication, Computation, Caching, and Intelligence (E-SC3I). The framework incorporates mechanisms for emergency computing, caching, integrated communication and sensing, and intelligence empowerment. E-SC3I ensures rapid access to a large user base, reliable data transmission over unstable links, and dynamic network deployment in a changing environment. However, these advantages come at the cost of significant computation overhead. Therefore, we specifically concentrate on emergency computing and propose an adaptive collaborative inference method (ACIM) based on hierarchical reinforcement learning. Experimental results demonstrate our method's ability to achieve rapid inference of AI models with constrained computational and communication resources.

Many rescue missions demand effective perception and real-time decision making, which highly rely on effective data collection and processing. In this study, we propose a three-layer architecture of emergency caching networks focusing on data collection and reliable transmission, by leveraging efficient perception and edge caching technologies. Based on this architecture, we propose a disaster map collection framework that integrates coded caching technologies. Our framework strategically caches coded fragments of maps across unmanned aerial vehicles (UAVs), fostering collaborative uploading for augmented transmission reliability. Additionally, we establish a comprehensive probability model to assess the effective recovery area of disaster maps. Towards the goal of utility maximization, we propose a deep reinforcement learning (DRL) based algorithm that jointly makes decisions about cooperative UAVs selection, bandwidth allocation and coded caching parameter adjustment, accommodating the real-time map updates in a dynamic disaster situation. Our proposed scheme is more effective than the non-coding caching scheme, as validated by simulation.

Integrated sensing and communication (ISAC) has emerged as a new paradigm for the sixth generation (6G) mobile communication systems. However, embedding ISAC into a conventional communication system may degrade the mutual benefit of radar sensing and communication due to the low-spectral efficiency and weak interference management. Therefore, in this article, we propose a full-duplex (FD) non-orthogonal multiple access (NOMA)-enabled ISAC framework in which a dual functional base station (BS) operates simultaneous target detection and uplink/downlink communication with the same temporal and spectral resources. To exploit some insights into the benefit of such a framework, we investigate the sensing signal processing procedure and communication model. Towards this end, a joint transmit and receive beamforming design is studied in the cases where single-target detection with perfect channel state information (CSI) and multitarget detection with imperfect CSI are considered, respectively. The corresponding optimization problems aim to maximize the sensing signal-to-interference-plus-noise ratio (SINR), subject to uplink communication SINR requirement for each uplink user equipment (UUE) and downlink communication SINR requirement for each downlink user equipment (DUE). We propose an alternating-optimization algorithm to solve the formulated non-convex optimization problems efficiently. Specifically, at each iteration, the closed forms of the optimal sensing and uplink communication receive beamforming vectors are obtained, respectively. Then, the sub-optimal transmit beamforming vector is solved by equivalent transformation and semi-definite relaxation (SDR) method. Numerical results demonstrate that the proposed FD-NOMA ISAC system outperforms the orthogonal multiple access (OMA)-based ISAC in terms of both sensing and communication performances. Notably, the efficacy of the proposed scheme is heavily dependent on factors, such as self cancelation (self interference) and CSI uncertainty.

The advanced Internet of Things (IoT) technique has provided a promising vision in emergency response issues with its convenience in environment cognition and communication. However, the resource-poor emergency rescue scenarios hinder the applications of IoT because of the volume constraints of portable devices. Integrated sensing and communication (ISAC) would be an ideal solution, but still faces technical challenges, such as signal integration and environment adaption. To this end, this work resorts to pilot optimization to design a dual-function ISAC signal for sensing and communication. Specifically, we derive the closed-form expressions of channel estimation and ranging performances in the OFDM system and formulate a weighted optimization to minimize the bit error ratio (BER) and average ranging error (ARE) with varying requirements of communication and ranging guaranteed. On this basis, a particle swarm optimization algorithm is developed to solve the problem, and a channel feedback framework is proposed to facilitate implementation. For emergency environment adaption, we conduct simulations with the SUI channel model to validate the efficiency of our algorithm. Simulation results demonstrate that our approach has a lower BER and ARE than typical pilot schemes under the same signal-to-noise ratio (SNR).

In emergency scenarios, strong mobility and serious interference cause unstable transmission of on-site information such as close-up photos and high resolution videos, which requires a robust temporary communication network. In this paper, we focus on a UAV-assisted wireless cooperative communication and coded caching network, where emergency command vehicles and a UAV serve as content providers (CPs) to cache and transmit coded fragments or complete files for rescuers regarded as content requesters (CRs). The delivery success probability and content hit ratio are theoretically derived by incorporating the physical connectivity and social relationship between CPs and CRs. Aiming at maximizing the overall content hit ratio, we propose a multiagent two-timescale deep reinforcement learning (MA2T-DRL) algorithm to jointly optimize the transmission power and caching strategies for CPs. Specifically, we develop a two tier deep-Q networks (DQNs) framework integrating a slow-timescale DQN (ST-DQN) and a fast-timescale DQN (FT-DQN) for caching decision-making and power decision-making respectively, and then the QMIX framework is leveraged to aggregate all the outputs from local ST-DQNs. Considering the cooperative characteristics of coded caching, we further propose a novel clustering method for CPs such that CPs in the same cluster have the same willingness to serve CRs, and each cluster is regarded as the agent for training which further reduces the aggregation scale of the mixing network. Simulation results show that the proposed MA2T-DRL algorithm is efficient in model training, and presents the advantages in performance and complexity compared with the single-agent centralized training and the multiagent independent distributed training.

The distributed inference paradigm enables the computation workload to be distributed across multiple devices, facilitating the implementation of deep learning based intelligent services on extremely resource-constrained Internet of Things (IoT) scenarios. Yet it raises great challenges to perform complicated inference tasks relying on a cluster of IoT devices that are heterogeneous in their computing/communication capacity and prone to crash or timeout failures. In this paper, we present RoCoIn, a robust cooperative inference mechanism for locally distributed execution of deep neural network-based inference tasks over heterogeneous edge devices. It creates a set of independent and compact student models that are learned from a large model using knowledge distillation for distributed deployment. In particular, the devices are strategically grouped to redundantly deploy and execute the same student model such that the inference process is resilient to any local failures, while a joint knowledge partition and student model assignment scheme are designed to minimize the response latency of the distributed inference system in the presence of devices with diverse capacities. Extensive simulations are conducted to corroborate the superior performance of our RoCoIn for distributed inference compared to several baselines, and the results demonstrate its efficacy in timely inference and failure resilience.

Channel State Information (CSI) feedback, powered by Deep Learning (DL) methodologies, exhibits significant promise in enhancing spectrum efficiency within massive MIMO systems. However, DL-based approaches typically necessitate substantial CSI datasets for each specific scenario, and managing multiple learned models demands considerable storage and updating bandwidth. To overcome this costly barrier, we develop a solution for efficient training and deployment enhancement of DL-based CSI feedback, which involves a lightweight translation model to cope with new CSI environments and introduces a novel dataset augmentation based on domain knowledge. Specifically, we first develop a deep unfolding CSI feedback network, SPTM2-ISTANet+, which incorporates spherical normalization to mitigate the challenge of path loss variation. Additionally, SPTM2-ISTANet+ integrates a trainable measurement matrix and residual CSI recovery blocks to enhance efficiency and accuracy. Employing SPTM2-ISTANet+ as a foundational feedback model, we introduce an adaptive CSI feedback architecture termed CSI-TransNet. CSI-TransNet features a scenario-adaptive plug-in module for CSI translation, composed of a sparsity aligning function and a compact DL module, facilitating the reuse of pretrained models in unencountered environments. To accommodate the small datasets, we propose a lightweight and general augmentation strategy based on domain knowledge. Test results demonstrate the efficacy and efficiency of the proposed solution for accurate CSI feedback given limited measurements for unseen CSI environments.

An effective communication is crucial yet challenging due to the damage sustained by ground communication infrastructure. A promising solution for enhancing emergency communication involves leveraging long-endurance fixed-wing UAVs to act as communication relays for ground rescuers. However, the disastrous environment and dynamic characteristics of UAVs may cause intermittent connectivity states, hindering efficient data transmissions. This study addresses data transmission issues in the fixed-wing UAV networks by proposing a routing scheme that leverages pre-defined UAV trajectories and tackles the dynamic topology using a graph-based time expansion method. Specifically, we propose a graph-based modeling approach to unfold network topology over time incorporating the packet transmission constraints. Then, the time-averaged delay minimization routing issue is transformed into a static linear programming problem, which is solved at each time step to minimize the packet delay and improve the network reliability. The experiment results reveal that the proposed approach outperforms the existing methods in terms of average packet delay and packet delivery ratio.

Given the energy storage capacity and rapid response capabilities, electric vehicles (EVs) hold the potential to offer auxiliary services, including peak shaving and frequency regulation, to the power grid during emergencies. Nevertheless, effectively coordinating EVs presents a complex challenge for multiple charging stations (CSs), which must integrate dynamic traffic fluctuations into a schedule while ensuring sufficient power resources for these auxiliary services. To tackle these challenges, this paper proposes an across-realm decision-making framework to dispatch EVs to CSs for peak shaving considering dynamic traffic conditions and power constraints. We develop a cooperative multi-agent reinforcement learning (MARL) strategy, which capitalizes on the collaboration among CSs by formulating the EV dispatching problem as a Markov Game. The CSs are regarded as learning agents and a QMIX network with bidding mechanism is applied to train the joint action towards maximizing long-term team rewards. The proposed method has been tested and compared to the existing reinforcement learning and non-reinforcement learning methods, and the simulation results have demonstrated the efficiency of the proposed approach considering real-world scenarios.

The rapid growth of cloud computing has led to high energy consumption in data centers (DCs), significantly burdening the safe operation of the power grid. As DC loads are seen as emergency demand response (EDR) resources, they can be aggregated into virtual power plants (VPPs) for the energy market trading, enhancing the renewable energy consumption capacity and providing grid benefits. However, VPPs might opt to maintain the profits they gain from EDR as confidential, driven by self-interest. Additionally, the incompatibility between the EDR transactions and task scheduling can result in revenue loss for DCs. To address these challenges, we propose a joint energy trading and computation scheduling (JETCS) framework for geo-distributed DCs participating in the EDR of VPPs. We first formulate DCs' energy and revenue computation as a maxi-mization problem to balance task delay and energy consumption. Then, we employ contract theory to facilitate the energy trading between DCs and VPPs against the information asymmetry. Due to nonlinearity and the infeasibility of obtaining an analytical result, we propose the proximal Jacobian alternating direction method of multipliers (PJADMM) algorithm to find an optimal solution with low complexity. The simulation results demonstrate that our proposed framework successfully encourages VPPs to reveal their actual profits and enhances the benefits for both parties.

The advancement of mobile edge computing (MEC) is driving the utilization of mobile devices (MDs) for real-time video detection. Nonetheless, during emergency scenarios, video detection tasks encounter two primary challenges: 1) time-varying channels over time between the MDs and the unmanned aerial vehicle (UAV); 2) the need for precision while managing energy consumption on the MDs. In this paper, we collaboratively address the optimization of computation offloading, data compression, and resource allocation challenges in the context of UAV-mounted MEC networks, where the UAV offers computational support for MDs. To improve accuracy and reduce energy consumption in time-varying environments, we propose an energy-efficient Deep Deterministic Policy Gradient based computation offloading algorithm (DCOA). By pursuing the long-term goal, DCOA learns to adapt to continuously changing conditions and thus improve accuracy and energy efficiency. Additionally, we introduce a convex optimization algorithm using the Lagrange multiplier method to solve resource allocation issues for offloading tasks, further reducing energy usage. Experimental results show that DCOA achieves high accuracy with low energy consumption compared to existing algorithms.

It is essential to locate the victims timely for efficient rescue after the disaster occurs in the global positioning system (GPS) denied forest area due to the influence of the tree shading. Existing works have studied optimizing the trajectory of the unmanned aerial vehicle (UAV) to exploit the wireless signal from the ground users for localization. However, current works mainly focus on optimizing the localization accuracy while paying limited attention to the localization task completion time, which makes it challenging to meet the timeliness requirements of emergency rescue missions. To provide accurate localization services for the ground victims quickly, we propose a multi-agent deep reinforcement learning (MA-DRL)-based UAV trajectory planning algorithm, which can provide high-efficiency cooperation between the multiple UAVs by exploiting the prior information and measurements from the partners. Specifically, a low-resolution trajectory planning algorithm is proposed to reduce redundant flight distances in the pre-fly stage to localize the victim’s quantity and dispersion. Furthermore, to provide high localization accuracy and energy-efficiency victims location services quickly, we exploit the coarse user information from the pre-fly stage, integrate an adaptive forest channel model and UAV energy consumption model, and propose an MA-DRL-based UAV trajectory planning algorithm which can perform decentralized execution for the high-efficiency cooperation localization. Simulation results show that our method can finish localization missions faster with less energy consumption while guaranteeing the localization accuracy compared to other benchmark algorithms.

In this paper, we propose a UAV network adaptation scheme driven by joint communication and positioning service provisioning in an emergency scenario, where massive rescuers’ concurrent and time-varying service demands are guaranteed with a scarce spectrum. Particularly, we establish a utility function that integrates communication rate and positioning error by jointly considering the single coverage constraint for data communication and the triple coverage constraint for positioning. Based on it, we propose a genetic algorithm based multi-agent deep deterministic policy gradient (GA-MADDPG) approach that adapts the UAV deployment, role switching, and user association strategies in a hierarchical manner to accommodate the rescuers’ demands for communication and positioning services in real-time. Specifically, the MADDPG module is applied for communication and positioning UAV network deployment based on the rescuers’ spatial distribution and their service demands, while the reward-based fitness is calculated and fed in the GA module periodically to optimize the UAV roles. Extensive simulation results show that our approach improves communication-positioning utility by up to 38% among comparison schemes.

The advances of communication technology and edge intelligence are deriving new computation offloading modes in the energy Internet by integrating computing capabilities of cloud servers, edge gateways (EGs), and terminal nodes into forecasting the renewable energy generation. However, the largely dispersed data generated by abundant photovoltaic (PV) stations and limited transmission capacity will degrade the collaboration of clouds, edges, and end nodes and, as a result, fail to satisfy the delay requirements of tasks. Owing to the similarity of power data generated by PV stations with akin geographical positions and weather conditions, we can reuse and offload the selected and representative power data so as to reduce transmission costs and overloads. In this article, we propose a similarity-aware stable matching approach (SASMA) to efficiently offload prediction tasks to EGs or cloud platforms with reusing the computing results. Specifically, we analyze task similarity and build the reuse strategy for the power data, and propose a similarity graph algorithm (SGA) to select representative PV stations and derive reuse relations. We also propose a similarity-based Gale–Shapley algorithm to match reused PV stations, computing nodes with prediction models. The objective is to maximize the prediction accuracy with a stable match. Simulation results show the effectiveness of the proposed approach while examining the tradeoff between the prediction accuracy and the system delay.

Electric power grid intelligence and automation are inseparable from the support of computing resources. Electric vehicles (EVs) can provide low-cost and flexible computing offloading services for nearby grid nodes. In this paper, we propose a collaborative model between EVs and intelligent charging stations (CSs) for energy and computation trading, where EVs can contribute their computation resources to CSs during charging at CSs. However, due to selfishness, CSs may refuse to reveal their computing requirements to the EVs, leading to information asymmetry. To cope with this limitation, the contract theory is employed to incentivize interaction between potential CS-EV pair for resource trading. Furthermore, we propose a stable-matching-based algorithm to match CSs and EVs into cooperative groups to achieve mutual-beneficial utilities. Simulation results verify the effectiveness of our algorithm.

In this paper, we propose a massive access control scheme in machine-type communications (MTC) aided smart grid, which prioritizes and clusters the devices based on the latency requirement and the distance between devices. The considered use case features a single cell in the massive connectivity smart grid and a large number of devices with different priority types. For the delay-sensitive devices, a dynamic random access channel (RACH) resource allocation scheme is proposed, where a back-off mechanism is used to defer the access requests of delay-tolerance devices. In addition, we propose a cluster-based congestion control algorithm, which clusters the devices to establish local collaboration. Simulation results show the proposed scheme reduces the average blocking probability, while significantly reducing access delay for delay-sensitive devices compared with state-of-the-art access control methods.

When signals propagate through forest areas, they will be affected by environmental factors such as vegetation. Different types of environments have different influences on signal attenuation. This paper analyzes the existing classical propagation path loss models and the model with excess loss caused by forest areas and then proposes a new short-range wireless channel propagation model, which can be applied to different types of forest environments. We conducted continuous-wave measurements at a center frequency of 605 MHz on predetermined routes in distinct types of forest areas and recorded the reference signal received power. Then, we use various path loss models to fit the measured data based on different vegetation types and distributions. Simulation results show that the proposed model has substantially smaller fitting errors with reasonable computational complexity, as compared with representative traditional counterparts.

The flexible and controllable mobility makes Un-manned Aerial Vehicle (UAV) useful in collecting data from distributed wireless sensor nodes, especially in emergency situations where infrastructures are destroyed or absent. However, due to the limited battery capacity, the mission duration of UAV is severely restricted, which makes the trajectory design become challenging. In this paper, a data-collection oriented multiple UAVs path planing algorithm is proposed to minimize the data collection time. Specifically, an enhanced particle swarm optimization (E-PSO) algorithm is first proposed to optimize the visiting sequence of sensors, which can shorten the flight time. Moreover, a novel hover-on-edge algorithm is designed by jointly considering wireless communication ranges and locations of sensors in traversal sequence. The proposed algorithms are compatible with UAV flying directly above the sensor (E-PSO) and UAV hovering on the communication range of sensor (HE-PSO) which enhance the robustness of the system. Simulation results demonstrate path distance and flight time reduction of our proposed algorithm.

Virtual Power Plant (VPP) has become an important means of low-carbon development and the forecasting of user Adjusted Power Consumption (APC) is the key part of VPP. The main challenge of APC forecasting is how to extract more APC-related features under the limitation of feature dimension for lower communication latency. In this paper, we propose a Feature-based APC forecasting (F-APC) framework for reducing forecasting error and communication latency. In the F-APC framework, firstly, we propose the method to extract APC-related features using the Classification-based Auto-Encoder (CAE) neural network. Secondly, we formulate the trade-off problem between forecasting error and communication latency and the optimal dimension of feature vector is solved by the closed-form solution. Experimental results on real data show that, compared with the benchmark, the forecasting error and communication latency of our scheme with the optimal setting is reduced by 45.26% and 33.33%, respectively.

To support multifarious vehicular applications, content sharing among vehicles or between vehicles and infrastructures can enable efficient service provisioning. In this work, we investigate joint communication, caching, and computing (3C) resource allocation to support efficient content sharing between content providers (CPs) and content requesters (CRs) in vehicular networks. To tackle the high complexity of joint 3C resource allocation, we decouple the problem into a long-term content caching strategy to allocate caching resources plus a method for short-term CP-CR pairing and corresponding communication-computing resource allocation. Specifically, a popularity and social similarity (P-SS) based caching strategy is proposed by incorporating both physical and social information. As selfish CRs may refuse to reveal their quality of service (QoS) requirements to the CPs and lead to the information asymmetry, we adopt contract theory to allocate communication and computing resources for each potential CR-CP pair. We then propose a stable-matching based algorithm to match CPs and CRs for efficient content sharing. Simulation results verify that the proposed scheme can effectively solve the problem with low complexity.

The UAV deployment as well as power allocation is critical in emergency rescue with the practical diverse re-quirements of data applications. In this paper, a UAV 3-D deployment scheme driven by multi-level quality of service (QoS) is first presented. Specifically, to harvest more performance gain, a fuzzy clustering-based initialization method is proposed, which can achieve more reliable accuracy of clustering compared with traditional clustering methods. Further, considering more comprehensive effect in terms of particle diversity and iteration, a novel inertia weight update method is developed to accelerate convergence. Simulation results demonstrate that our proposed scheme outperforms other schemes.

As an emerging distributed learning paradigm, Federated Learning (FL) allows smart meters to collaboratively train a load forecasting model while keeping their private data on local devices. However, two critical issues hinder the deployment of ordinary FL algorithm in load forecasting: (i) one global model cannot fit all users well due to their heterogeneous load patterns; (ii) the training speed of FL severely depends on a few stragglers with scarce communication and computing resources. In this work, we propose a novel multi-center FL framework for load forecasting to learn multiple models simultaneously by grouping the users according to their model dissimilarity and training time. Specifically, a problem is formulated to jointly optimize the grouping strategy and forecasting model parameters, which is resolved by integrating the matching algorithm into the update process of model parameters in FL. Simulation results on real load data show that, compared with the existing load forecasting methods based on FL, the prediction error of our scheme is reduced by 8.11%, and the training time is reduced by 90.37%.

Coded caching can improve the robustness of wireless caching networks. This paper investigates the joint caching and communication optimization in terminal based wireless coded caching networks. Social characteristics of private terminals are considered to further harvest more performance gain in terms of hit ratio. To tackle the problem of unknown popularity distribution, we adopt deep reinforcement learning. Since the joint caching and communication optimization of all Content Providers (CP) results in larger-scale action and state space, we propose a novel cluster based deep reinforcement learning (CB-DRL) scheme. We present simulation results to demonstrate complexity reduction and effectiveness of the proposed algorithm.

Methods to cluster fingerprints based on Smallest-Enclosing-Circle (SEC) and to modify Weighted-K-Nearest-Neighbor (WKNN) matching algorithm for indoor fingerprint positioning system are proposed. Based on the approach to computing the smallest k-enclosing circle, the method proposed clusters fingerprints in database by introducing reference points' coordinates, instead of their received signal strength (RSS). This approach performs higher accuracy of positioning areas compared to conventional clustering algorithms, which are based on RSS. Meanwhile, this paper analyses the transmission characteristics of wireless signals in dense cluttered environments, and derives a novel path-loss-model-based weight computational method for WKNN matching algorithm. A modified-WKNN (M-WKNN) matching algorithm for indoor fingerprint positioning system is proposed and experiments are implemented in China National Grand Theatre. Results show that the location area accuracy using the proposed clustering algorithm is improved by 30% compared to that using K-means algorithm, and the positioning accuracy of M-WKNN is 11.9% and 29.1% higher than that of WKNN and KNN, respectively.