With the advancement of Artificial Intelligence (AI) technology, edge networks are progressively evolving towards intelligence, and mobile intelligent agents such as Unmanned Aerial Vehicles (UAVs) and Unmanned Ground Vehicles (UGVs) are playing an increasingly vital role in this process. However, existing studies have addressed cooperation among homogeneous agents, such as multiple UAVs, without exploring collaboration between heterogeneous mobile intelligent agents. For this purpose, we propose a vehicle-mounted multi-UAV cooperative service system to collaboratively collect and process multi-source heterogeneous Internet of Things (IoT) tasks under strict latency and resource constraints. To address the complexities of heterogeneous task structures and dynamic collaboration, the paper firstly introduces a multi-source heterogeneous task scheduling mechanism, which optimizes task prioritization and resource allocation for efficient processing. In addition, a decentralized reinforcement learning approach based on Partial Reward Decoupling Heterogeneous Agent Proximal Policy Optimization (PRD-HAPPO) is employed to enhance collaboration and trajectory planning between UAVs and vehicles. Simulation results demonstrate that the proposed framework significantly improves task completion efficiency, reduces system latency, and outperforms existing Deep Reinforcement Learning (DRL) algorithms in terms of convergence and scalability.

Maritime object detection is essential for navigation safety, surveillance, and autonomous operations, yet constrained by two key challenges: the scarcity of annotated maritime data and poor generalization across various maritime attributes (e.g., object category, viewpoint, location, and imaging environment). To address these challenges, we propose Neptune-X, a data-centric generative-selection framework that enhances training effectiveness by leveraging synthetic data generation with task-aware sample selection. From the generation perspective, we develop X-to-Maritime, a multi-modality-conditioned generative model that synthesizes diverse and realistic maritime scenes. A key component is the Bidirectional Object-Water Attention module, which captures boundary interactions between objects and their aquatic surroundings to improve visual fidelity. To further improve downstream tasking performance, we propose Attribute-correlated Active Sampling, which dynamically selects synthetic samples based on their task relevance. To support robust benchmarking, we construct the Maritime Generation Dataset, the first dataset tailored for generative maritime learning, encompassing a wide range of semantic conditions. Extensive experiments demonstrate that our approach sets a new benchmark in maritime scene synthesis, significantly improving detection accuracy, particularly in challenging and previously underrepresented settings. The code is available at https://github.com/gy65896/Neptune-X.

Nonorthogonal multiple access (NOMA) has been recognized as a key enabling techniques for future communication systems since it can take advantage of differences in users’ channel conditions to achieve higher spectral efficiency compared with orthogonal multiple access. However, employing such a capacity-centric framework can result in intragroup co-channel interference for those users whose symbols are intended to be first decoded, potentially causing a decrease in reliability. Therefore, the Alamouti encoding method has been widely used in the multiuser MISO-NOMA broadcast channels to improve reliability. Although the Alamouti-based NOMA paradigm can improve reliability by providing full spatial diversity, it comes at the cost of a significant loss in spectral efficiency compared to the capacity-centric NOMA framework. To overcome these limitations, this article presents a new approach for improving the performance of multiuser MISO-NOMA broadcast channels, in which the diversity-multiplexing tradeoff is leveraged more effectively by providing multiplexing for near users and diversity for far users at the transceiver. The closed-form expressions of ergodic sum-rate and outage probability for the proposed scheme are derived. Simulation results demonstrate that the proposed scheme is capable of performing well in a wide range of scenarios and it surpasses traditional methods in terms of reliability and effectiveness.

With the development of 6G wireless networks, non-fixed monitoring methods utilizing drone networks as monitoring carriers have been receiving increasing attention. However, due to significant environmental interference affecting the hovering of drones in the air, the video signals often contain substantial random interference, which impacts the performance of current multi-object and multi-camera tracking (MOMCT) algorithms. To address this issue, this paper proposes a digital-twin-driven multi-vehicle multi-drone monitoring system, in which a simulation scenario consistent with the actual environment is created. Furthermore, a large amount of vehicle monitoring video data from multiple drone perspectives with random interference is generated, thus compensating for the shortage of multi-camera vehicle monitoring data, especially data affected by random interference. Additionally, an online tracking framework for MOMCT based on the digital twin system is set up, forming a pipeline that fully utilizes the simulated data generated by the digital-twin system for training and testing. Using this pipeline, we find that a feature detection algorithm combining AttentionNet-CBAM and ResNext101-IBN-A can effectively enhance the ability to identify target features, thereby achieving better multi-vehicle tracking results. Experimental results on the CityFlow dataset verify that the proposed method has improved the IDF1 by 1.52% and IDR by 3.58% in comparison with the state-of-the-art online MOMCT methods.

Object heading detection (OHD) offers potential for research in the control and traffic analysis sectors. Contemporary methodologies in OHD grapple with a set of distinct limitations: a constrained range of detectable object types, a discernible drop in accuracy for oriented bounding box (OBB) predictions influenced by heading estimations, and the inherent limitations associated with the exclusive perspective of bird’s-eye view imagery. This paper introduces HDDet, an advanced method devised for detecting the OBB of objects with heading direction from various viewpoints. It sequentially delineates the Circular Annotation Method (CAM), Multi-dimensional Angle Encoding (MDAE), and the CosWeight strategy. CAM initially couples oriented bounding box data with heading details for accurate target delineation. MDAE follows, optimizing angle encoding to boost the model’s training process. The culmination of this approach is CosWeight, which integrates the Rotated-IoU and heading information into the horizontal box’s loss, thereby enhancing the precision of heading predictions. Rigorous testing across diverse datasets, including the SJTU-L dataset where the heading accuracy increased from 64.41% to 94.82% over OHDet, validates HDDet’s enhanced capabilities, surpassing existing methodologies in both OBB detection and heading accuracy, and marking a significant advancement over the state-of-the-art OHDet. Additionally, the paper presents an engineering vehicle dataset, which is conducive to multi-perspective object heading detection research.

This paper investigates the beamforming design in the Cell-Free massive multi-input multi-output (MIMO) integrated sensing and communication (ISAC) system, termed as the CF-ISAC system, in presence of the channel state information (CSI) estimation error. The beamforming design is formulated into a sensing beampattern matching mean square error minimization problem under the constraints of the power budgets of the access points (APs) and the ergodic rate requirements of the users. A computationally tractable lower bound of the ergodic rate over the imperfect CSI is derived based on the Jensen's Inequality, and then a successive convex approximation based algorithm is proposed to solve the considered problem. Numerical results illustrate the beampatterns for different direction of arrival estimations of the targets. The advantage of the CF-ISAC system for radar sensing is revealed based on the relative location between the AP and the target.

Abstract Most of the shape classification methods are based on a single closed contour. However, practical shapes always have complex contours, for example, a combination of multiple open contours. How to accurately identify complex shapes is an unsolved problem. In this research, a novel method is proposed to classify complex shapes. The proposed method firstly encodes a complex shape to an angle code and a sparsity code, then input these codes to a 1‐D CNN for extracting features and classification. Experiments on two datasets show this novel method is superior in terms of classification accuracy. These two datasets are practical shape dataset collected by this paper on internet and MPEG‐7 CE‐1 Part B. The proposed method achieves higher classification accuracy than compared methods. In order to show the performance of the proposed method on each class, the accuracy on each class is analyzed. Ablation experiment is conducted to show the contribution of each module in the network. The result shows that each module is meaningful in the network, because without any module the accuracy drops.

Maneuvering-target tracking has always been an important and challenge work because the unknown and changeable motion-models can easily lead to the failure of model-driven target tracking. Recently, many neural network methods are proposed to improve the tracking accuracy by constructing direct mapping relationships from noisy observations to target states. However, limited by the coverage of training data, those data-driven methods suffer other problems, such as weak generalization abilities and unstable tracking effects. In this paper, a digital twin system for maneuvering-target tracking is built, and all kinds of simulated data are created with different motion-models. Based on those data, the features of noisy observations and their relationship to target states are found by two specially designed neural networks: one eliminates the observation noises and the other one predicts the target states according to the noise-limited observations. Combining the above two networks, the state prediction method is proposed to intelligently predict targets by understanding the information of motion-model hidden in noisy observations. Simulation results show that, in comparison with the state-of-the-art model-driven and data-driven methods, the proposed method can correctly and timely predict the motion-models, increase the tracking generalization ability and reduce the tracking root-mean-squared-error by over 50% in most of maneuvering-target tracking scenes.

Locating the propagation source is one of the most important strategies to control the harmful diffusion process on complex networks. Most existing methods only consider the infection time information of the observers, but the diffusion direction information of the observers is ignored, which is helpful to locate the source. In this paper, we consider both of the diffusion direction information and the infection time information to locate the source. We introduce a relaxed direction-induced search (DIS) to utilize the diffusion direction information of the observers to approximate the actual diffusion tree on a network. Based on the relaxed DIS, we further utilize the infection time information of the observers to define two kinds of observers-based similarity measures, including the Infection Time Similarity and the Infection Time Order Similarity. With the two kinds of similarity measures and the relaxed DIS, a novel source locating method is proposed. We validate the performance of the proposed method on a series of synthetic and real networks. The experimental results show that the proposed method is feasible and effective in accurately locating the propagation source.

Maritime ports face challenges associated with navigation safety, operational efficiency, and management. With the development of the Internet of Things, artificial intelligence simulation technologies, geographical information systems, and cloud computing technologies as well as navigation aids and decision support systems in maritime transportation, ports have the potential to better manage traffic, loading, and unloading. Recently, there has been growing attention in unmanned shipping to support the maritime industry and the military. This paper aims to extend the application of geographical theory and methodology in unmanned ship path optimization. Automatic collision avoidance concerning maneuvering capabilities of ships as well as complying with maritime traffic rules remains a challenge. This study attempts to tackle development needs associated with path optimization in maritime travel. By integrating ship movement behavior, geographical features, and the International Regulations for Avoiding Collisions at Sea, the proposed methods seek to reduce the human error associated with maritime accidents. This paper proposes economic efficiency and safety-driven unmanned ship path planning that will promote the future growth of intelligent port development.

Connected vehicles (CVs) are getting increasing attention in intelligent transportation systems (ITSs). In terms of the stringent security and anonymity issues, a novel anonymous and decentralized payment platform is demanded. Bitcoin is regarded as a potential solution. But the existing Bitcoin and its various improvements lack a beforehand anti-fraud Bitcoin deposit scheme, which is necessary to prevent potential fraud risks from Bitcoin to CVs. This paper aims to build a novel anti-fraud deposit scheme as a seamless bridge between CV networks and the Bitcoin payment platform. We propose a Bitcoin-to-Connected-Vehicle deposit scheme (Bit2CV) with an outsourcing endorsement in order to build an anti-fraud deposit transaction (dtr). Firstly, Bit2CV leverages a special Bitcoin-opcode-OP_RETURN based method to record the dtr, the CV's request, and the endorsement together on Blockchain ledger. This method can achieve security features, including non-repudiation, privacy-friendly endorsement, future audit, and anti-fraud. Meanwhile, Bit2CV is fully decentralized without any involvement of centralized authorities and fully compatible with the current Bitcoin network. We conduct the security analysis and also simulations for performance evaluation. In a typical scenario of our simulation, Bit2CV total time cost is less than 477.03 ms, and the size of the endorsement is 1,732 bytes, which is much less and smaller than transaction confirmation time and the average block size, respectively. These designs and results demonstrate that Bit2CV is feasible and practical.

In this research, an attribute-weighted one-dependence Bayes estimation algorithm based on the asymmetric correlation coefficient is proposed. The asymmetric correlation coefficients Tau_y and Lambda_y, respectively, are used to calculate the correlation between parent attributes and category labels, then the result of calculation is regarded as weight to the parent attribute. The algorithm is applied to eight types of different datasets including binary classification and multiple classification from the UCI database. By comparing the time complexity and classification accuracy, experimental results show that the algorithm can significantly improve the classification performance with less prediction error. In addition, several baseline methods such as KNN, ANN, logistic regression and SVM are used for comparison with the proposed method.

This paper investigates the wireless-powered hierarchical fog-cloud computing networks, where multiple energy-constrained users harvest energy from a hybrid access point (HAP) firstly and then use their harvested energy to offload their computation tasks to fog/cloud servers via the HAP or compute their tasks locally. To pursue multi-user fairness, an optimization problem is formulated to maximize the minimal energy balance among all users by jointly optimizing time assignments, computation central processing unit (CPU) frequencies, and the computing mode selection. Since the problem is mixed-integer combinatorial non-convex, which is intractable, a generalized Benders decomposition (GBD)-based method is proposed, which guarantees the globally optimal solution. To release the high computational complexity of the proposed GBD-based method, a penalized successive convex approximation (P-SCA)-based algorithm is designed as an alternative to obtain a suboptimal solution with low computational complexity. Numerical results show that among different optimizable factors in the system, computing mode selection is the dominant one on affecting the system performance. Moreover, for each user, local computing is a better choice, if it is with relatively poor channel gain and small local computing delay. Otherwise, fog/cloud computing may be a better choice. Additionally, for the users with relatively high channel gains, if their local computing delays are less than those selecting fog computing, cloud computing should be a better choice.

This paper investigates the optimal design of wireless- powered hierarchical fog-cloud computing networks, where energy-constrained users first harvest energy from a hybrid access point (HAP) and then offload their computation tasks to fog/cloud servers via the HAP or compute the tasks locally by us- ing the harvested energy. An optimization problem is formulated to maximize the minimal energy balance among multiple users by jointly optimizing offloading decisions, communication and computation resource allocations in the system, where computational capacity, processing delay, energy harvesting (EH) and energy consumption constraints are considered. To efficiently solve such a mixed-integer combinatorial non-convex problem, a penalized successive convex approximation (P-SCA)- based algorithm is designed, which is able to converge to a suboptimal solution with the polynomial time computational complexity. Numerical results show that compared to communication and computation resource allocation, the offloading decision is the dominant factor on affecting the system performance. It is also found that local computing is a better choice for users with relatively poor channel gains while fog/cloud computing is a better choice for users with relatively good channel gains. Specifically, cloud computing is preferred if the cloud computational capacity is strong enough and the wired-link data rate is high enough; Otherwise, fog computing is preferred. Besides, more users are served, less max-min energy balance can be obtained.

In the field of maneuvering-target tracking (MTT), the targets with changeable and uncertain maneuvering movements cannot be tracked precisely because there always exist time delays of maneuvering model estimation with traditional MT-T algorithms. To solve this problem, we propose a deep MTT (DeepMTT) algorithm based on a deep neural network, which can quickly track maneuvering targets once it has been well trained by abundant off-line trajectory data from existent ma-neuvering targets. To this end, we first build a Large-scale trajectory database to offer abundant off-line trajectory data for network training. Second, the DeepMTT algorithm is developed based on a deep neural network, which consists of three bidirectional long short-term memory layers, a filtering layer, a maxout layer and a linear output layer. The simulation results verify that our DeepMTT algorithm outperforms other state-of-the-art MTT algorithms.

By allowing mobile device to offload all or part of a latency-constrained computational task to a base station at the edge of a network, mobile edge computing (MEC) is a promising technique to help the mobile device save its energy consumption. In this paper, we consider the scenario with one mobile device and multiple edge base stations, which is usual in practice. Nonorthogonal multiple access (NOMA) technique is implemented. An optimization problem is formulated, in which the transmission power to every edge base station and the amount of data to be offloaded for computation is optimized to minimize the total energy consumption of the mobile device. However, the formulated optimization problem is non-convex. To get the global optimal solution, we decompose the formulated optimization problem into two levels. In the lower level, a convex optimization problem is required to be solved. By finding out some special property of the lower level optimization problem, the upper level optimization problem can be formulated as a monotonic optimization problem, whose global optimal solution is achievable. Numerical results verify the effectiveness of our proposed method.

This paper investigates a simultaneous wireless information and power transfer (SWIPT)-enabled non- orthogonal multiple access (NOMA) network with full- duplex (FD) relaying, where a multi-antenna source transmits information to two users. The nearby user is with multiple antennas, which receives its own information and harvests energy from the signals transmitted by the source and also help forward information to the far-end user. For such a system, an optimization problem is formulated to minimize the required transmit power by jointly optimizing beamforming vectors and power splitting (PS) ratio under the energy harvesting and users' data rate constraints of both users. As the problem is non- convex with unknown solution, a bilevel- optimization method is proposed to solve it via semidefinite relaxation (SDR) and the global optimal solution is achieved with perfect self- interference cancellation. However, since self- interference may not be cancelled perfectly in practice, a successive convex approximation (SCA) based algorithm with low complexity is proposed to obtain a near optimal solution. Numerical results show that integrating NOMA, FD relaying and SWIPT in a single communication system is able to greatly reduce the required transmit power. Besides, the effects of the parameters including the data rate threshold and the energy storage amounts, on the system performance are also discussed.

Recently, videoconferencing has been popular in multimedia systems, such as FaceTime and Skype. In videoconferencing, almost every frame contains a human face. Therefore, it is important to predict human visual attention on face videos by saliency detection, as saliency may be used as a guide to the region of interest for the content-based applications of face videos. In this paper, we propose a data-driven approach for saliency detection in face videos. From the data-driven perspective, we first establish an eye-tracking database that contains fixations of 76 face videos viewed by 40 subjects. Upon the analysis of our database, we find that visual attention is significantly attracted by faces in videos. More important, the attention distribution within face regions varies with regard to mouth movement. Since previous works have investigated that it is efficient to model face saliency in still images using a Gaussian mixture model (GMM), the variation of visual attention in videos can be modeled by dynamic GMM (DGMM). Accordingly, we propose adopting the particle filter (PF) in modeling DGMM for saliency detection of face videos, which is called PF-DGMM. Finally, the experimental results show that our PF-DGMM approach significantly outperforms other state-of-the-art approaches in saliency detection of face videos.

Single-image blind deblurring could be considered as an important preprocessing step in imaging information fusion. Its purpose is to simultaneously estimate blur kernel and latent sharp image from only one observed blurred image. Blind deblurring has been attracting increasing attention in the fields of image processing, computer vision, computational photography, etc. However, it is a typically ill-posed inverse problem, which requires regularization methods to guarantee stable image restoration results. We first proposed to robustly estimate the blur kernels by exploiting non-convex sparsity constraints on image gradients and blur kernels. The corresponding combined non-convex regularization term has the capacity of enhancing estimation accuracy. To guarantee the high-quality non-blind deblurring with estimated blur kernels, the hybrid non-convex first- and second-order TV regularizer was then introduced to stabilize the final image restoration process. The hybrid non-convex regularizer is able to achieve a good balance between sharp edges preservation and undesirable artifacts suppression. The resulting non-convex minimization problems related to blur kernel estimation and non-blind deblurring were handled using efficient numerical optimization algorithms in this paper. Numerous experiments on both synthetic and realistic images have demonstrated the good performance of the proposed blind deblurring method.

Image forensics with the presence of an adversary, such as the interplay between the sensor-based camera source identification (CSI) and the fingerprint-copy attack, has attracted increasing attention recently. In this paper, we propose a framework of CSI game with both complete information and incomplete information. A noise level-based counter anti-forensic method is presented to detect the potential fingerprint-copy attack, and unlike the state-of-the-art countermeasure of the triangle test, it does not need to collect the candidate image set. With the existence of countermeasure, a rational forger needs to balance the tradeoff between synthesizing source information and leaving new detectable evidence of raising the noise level of a forged image. The mixed-strategy other than the sequential-move assumption is adopted to solve the games. The Bayesian game is introduced to address the information asymmetry in practice. The Nash equilibrium of both the complete information game and Bayesian game are theoretically analyzed, and the expected Nash equilibrium payoff of a Bayesian game is obtained. Nash equilibrium receiver operating characteristic curves are adopted to evaluate the detection performance. Simulation results show that the information asymmetry can remarkably affect the final detection performance. To our knowledge, this paper is the first attempt in analyzing a Bayesian forensic game with practical information asymmetry.

In this paper, we investigate how features can be effectively learned by deep neural networks for audio forensic problems. By providing a preliminary feature preprocessing based on electric network frequency (ENF) analysis, we propose a convolutional neural network (CNN) for training and classification of genuine and recaptured audio recordings. Hierarchical representations which contain levels of details of the ENF components are learned from the deep neural networks and can be used for further classification. The proposed method works for small audio clips of 2 second duration, whereas the state of the art may fail with such small audio clips. Experimental results demonstrate that the proposed network yields high detection accuracy with each ENF harmonic component represented as a single-channel input. The performance can be further improved by a combined input representation which incorporates both the fundamental ENF and its harmonics. The convergence property of the network and the effect of using an analysis window with various sizes are also studied. Performance comparison against the support tensor machine demonstrates the advantage of using CNN for the task of audio recapture detection. Moreover, visualization of the intermediate feature maps provides some insight into what the deep neural networks actually learn and how they make decisions.