Federated learning is a distributed machine learning paradigm that emerged as a solution to the need for privacy protection in artificial intelligence. Like traditional machine learning, federated learning is threatened by multiple attacks, such as backdoor attacks, Byzantine attacks, and adversarial attacks. The weaknesses are exacerbated by the inaccessibility of data in federated learning, which makes it more difficult to defend against these threats. This points to the need for further research into defensive approaches to make federated learning a real solution for distributed machine learning paradigm with securing data privacy. Our survey provides a taxonomy of these threats and defense methods, describing the general situation of this vulnerability in federated learning. We also sort out the relationship between these methods, their advantages and disadvantages, and discuss future research directions regarding the security issues of federated learning from multiple perspectives.

Effectively improving the optimization performance of artificial electric field algorithm (AEFA) and broadening its application domain can aid in providing robot path planning in three-dimensional (3D) complex scenes. This paper effectively proposes an improved AEFA (I-AEFA) and creatively applies it to robot path planning. The algorithm introduces three mechanisms to enhance the exploration ability and convergence accuracy of the population: parameter adaptation, reverse learning, and Cauchy mutation. Next, the benchmark terrain model accurately models the 3D environment, and the global path planning problem is solved using a combination of I-AEFA and cubic spline interpolation. Then, a large number of virtual simulation experiments are conducted to evaluate the algorithm's three improved mechanisms, various control point counts, as well as single and multi-robot configurations before migrating the algorithm to the graphical modeling and analysis software (GMAS) for hardware-in-the-loop simulation experiments. Finally, the experimental results are analyzed qualitatively and quantitatively using a variety of visualization techniques and two nonparametric test methods, demonstrating that the I-AEFA proposed in this paper has good optimization performance and is highly effective, reliable, and scalable for solving robot path planning problems.

The use of artificial intelligence (AI) has increased since the middle of the 20th century, as evidenced by its applications to a wide range of engineering and science problems. Air traffic management (ATM) is becoming increasingly automated and autonomous, making it lucrative for AI applications. This paper presents a systematic review of studies that employ AI techniques for improving ATM capability. A brief account of the history, structure, and advantages of these methods is provided, followed by the description of their applications to several representative ATM tasks, such as air traffic services (ATS), airspace management (AM), air traffic flow management (ATFM), and flight operations (FO). The major contribution of the current review is the professional survey of the AI application to ATM alongside with the description of their specific advantages: (i) these methods provide alternative approaches to conventional physical modeling techniques, (ii) these methods do not require knowing relevant internal system parameters, (iii) these methods are computationally more efficient, and (iv) these methods offer compact solutions to multivariable problems. In addition, this review offers a fresh outlook on future research. One is providing a clear rationale for the model type and structure selection for a given ATM mission. Another is to understand what makes a specific architecture or algorithm effective for a given ATM mission. These are among the most important issues that will continue to attract the attention of the AI research community and ATM work teams in the future.

Anti-ship missile coordinated attack mission planning is a complex multi-objective optimization problem with multiple combinations of platforms, strong decision-making constraints, and tightly coupled links. To avoid the coupling disorder between path planning and firepower distribution and improve the efficiency of coordinated attack mission planning, a firepower distribution model under the conditions of path planning is established from the perspective of decoupling optimization and the algorithm is implemented. First, we establish reference coordinate system of firepower distribution to clarify the reference direction of firepower distribution and divide the area of firepower distribution; then, we construct an index table of membership of firepower distribution to obtain alternative firepower distribution plans; finally, the fitness function of firepower distribution is established based on damage income, missile loss, ratio of efficiency and cost of firepower distribution, and the mean square deviation of the number of missiles used, and the alternatives are sorted to obtain the optimal firepower distribution plan. According to two simulation experiments, the method in this paper can effectively solve the many-to-many firepower distribution problem of coupled path planning. Under the premise of ensuring that no path crossing occurs, the optimal global solution can be obtained, and the operability and timeliness are good.

Unmanned Aerial Vehicle (UAV) imagery has emerged as a critical data source in remote sensing, playing an important role in vehicle detection for intelligent traffic management and urban monitoring. Deep learning–based detectors rely heavily on large-scale, high-quality annotated datasets, however, collecting and labeling real-world UAV data are both costly and time-consuming. Owing to its controllability and scalability, synthetic data has become an effective supplement to address the scarcity of real data. Nevertheless, the significant domain gap between synthetic data and real data often leads to substantial performance degradation during real-world deployment. To address this challenge, this paper proposes a high-fidelity synthetic data generation framework designed to reduce the Sim2Real gap. First, UAV oblique photogrammetry is utilized to reconstruct real-world 3D model, ensuring geometric and textural authenticity; second, diversified rendering strategies that simulate real-world illumination and weather variations are adopted to cover a wide range of environmental conditions; finally, an automated ground-truth generation algorithm based on semantic masks is developed to achieve pixel-level precision and cost-efficient annotation. Based on this framework, we construct a synthetic dataset named UAV-SynthScene. Experimental results show that multiple mainstream detectors trained on UAV-SynthScene achieve competitive performance when evaluated on real data, while significantly enhancing robustness in long-tail distributions and improving generalization on real datasets.

Aiming at the triangular fuzzy (TF) multi-attribute decision making (MADM) problem with a preference for the distribution density of attribute (DDA), a decision making method with TF number two-dimensional density (TFTD) operator is proposed based on the density operator theory for the decision maker (DM). Firstly, a simple TF vector clustering method is proposed, which considers the feature of TF number and the geometric distance of vectors. Secondly, the least deviation sum of squares method is used in the program model to obtain the density weight vector. Then, two TFTD operators are defined, and the MADM method based on the TFTD operator is proposed. Finally, a numerical example is given to illustrate the superiority of this method, which can not only solve the TF MADM problem with a preference for the DDA but also help the DM make an overall comparison.