Deep learning has achieved excellent results in various tasks in the field of computer vision, especially in fine-grained visual categorization. It aims to distinguish the subordinate categories of the label-level categories. Due to high intra-class variances and high inter-class similarity, the fine-grained visual categorization is extremely challenging. This paper first briefly introduces and analyzes the related public datasets. After that, some of the latest methods are reviewed. Based on the feature types, the feature processing methods, and the overall structure used in the model, we divide them into three types of methods: methods based on general convolutional neural network (CNN) and strong supervision of parts, methods based on single feature processing, and methods based on multiple feature processing. Most methods of the first type have a relatively simple structure, which is the result of the initial research. The methods of the other two types include models that have special structures and training processes, which are helpful to obtain discriminative features. We conduct a specific analysis on several methods with high accuracy on public datasets. In addition, we support that the focus of the future research is to solve the demand of existing methods for the large amount of the data and the computing power. In terms of technology, the extraction of the subtle feature information with the burgeoning vision transformer (ViT) network is also an important research direction.

In challenging situations, such as low illumination, rain, and background clutter, the stability of the thermal infrared (TIR) spectrum can help red, green, blue (RGB) visible spectrum to improve tracking performance. However, the high-level image information and the modality-specific features have not been sufficiently studied. The proposed correlation filter uses the fused saliency content map to improve filter training and extracts different features of modalities. The fused content map is introduced into the spatial regularization term of correlation filter to highlight the training samples in the content region. Furthermore, the fused content map can avoid the incompleteness of the content region caused by challenging situations. Additionally, different features are extracted according to the modality characteristics and are fused by the designed response-level fusion strategy. The alternating direction method of multipliers (ADMM) algorithm is used to solve the tracker training efficiently. Experiments on the large-scale benchmark datasets show the effectiveness of the proposed tracker compared to the state-of-the-art traditional trackers and the deep learning based trackers.

The direction of ground-based interference reaching the satellite is generally very close to the spot beam of the satellite. The traditional array anti-jamming method may cause significant loss to the uplink signal while suppressing the interference. In this paper, an aperiodic multistage array is used, and a sub-array aperiodic distribution optimization scheme based on parallel differential evolution is proposed, which effectively improves the beam resolution and suppresses the grating lobe. On this basis, a two-stage signal processing method is used to suppress interference. Finally, the comprehensive performance of the proposed scheme is evaluated and verified.

This paper considers an intelligent reflecting surface (IRS)-assisted multiple-input multiple-output (MIMO) system. To maximize the average achievable rate (AAR) under outdated channel state information (CSI), we propose a twin-timescale passive beamforming (PBF) and power allocation protocol which can reduce the IRS configuration and training overhead. Specifically, the short-timescale power allocation is designed with the outdated precoder and fixed PBF. A new particle swarm optimization (PSO)-based long-timescale PBF optimization is proposed, where mini-batch channel samples are utilized to update the fitness function. Finally, simulation results demonstrate the effectiveness of the proposed method.

Multi-objective optimization (MOO) for the microwave metamaterial absorber (MMA) normally adopts evolutionary algorithms, and these optimization algorithms require many objective function evaluations. To remedy this issue, a surrogate-based MOO algorithm is proposed in this paper where Kriging models are employed to approximate objective functions. An efficient sampling strategy is presented to sequentially capture promising samples in the design region for exact evaluations. Firstly, new sample points are generated by the MOO on surrogate models. Then, new samples are captured by exploiting each objective function. Furthermore, a weighted sum of the improvement of hypervolume (IHV) and the distance to sampled points is calculated to select the new sample. Compared with two well-known MOO algorithms, the proposed algorithm is validated by benchmark problems. In addition, two broadband MMAs are applied to verify the feasibility and efficiency of the proposed algorithm.

During high-speed flight, both thermal and mechanical loads can degrade the electrical performance of the antenna-radome system, which can subsequently affect the performance of the guidance system. This paper presents a method for evaluating the electrical performance of the radome when subjected to thermo-mechanical-electrical (TME) coupling. The method involves establishing a TME coupling model (TME-CM) based on the TME sharing mesh model (TME-SMM) generated by the tetrahedral mesh partitioning of the radome structure. The effects of dielectric temperature drift and structural deformation on the radome’s electrical performance are also considered. Firstly, the temperature field of the radome is obtained by transient thermal analysis while the deformation field of the radome is obtained by static analysis. Subsequently, the dielectric variation and structural deformation of the radome are accurately incorporated into the electrical simulation model based on the TME-SMM. The three-dimensional (3D) ray tracing method with the aperture integration technique is used to calculate the radome’s electrical performance. A representative example is provided to illustrate the superiority and necessity of the proposed method. This is achieved by calculating and analyzing the changes in the radome’s electrical performance over time during high-speed flight.

The parametric scattering center model of radar target has the advantages of simplicity, sparsity and mechanism relevant, making it widely applied in fields such as radar data compression and rapid generation, radar imaging, feature extraction and recognition. This paper summarizes and analyzes the research situation, development trend, and difficult problems on scattering center (SC) parametric modeling from three aspects: parametric representation, determination method of model parameters, and application.

This paper considers the problem of sea clutter suppression. We propose the cuttable encoder-decoder-augmentation network (CEDAN) to improve clutter suppression performance by enriching the contrast information between the target and clutter. Specifically, the plug-and-play residual U-block (ResUblock) is proposed to augment the feature representation ability of the clutter suppression model. The CEDAN first extracts and fuses the multi-scale features using the encoder and the decoder composed of the ResUblocks. Then, the fused features are processed by the contrast information augmentation module (CIAM) to enhance the diversity of target and clutter, resulting in encouraging sea clutter suppression results. In addition, we propose the result-consistency loss to further improve the suppression performance. The result-consistency loss enables CEDAN to cut some blocks of decoder and CIAM to reduce the inference time without significantly degrading the suppression performance. Experimental results on measured and simulated data show that the CEDAN outperforms state-of-the-art sea clutter suppression methods in sea clutter suppression performance and computation efficiency.

Separation and recognition of radar signals is the key function of modern radar reconnaissance, which is of great significance for electronic countermeasures and anti-countermeasures. In order to improve the ability of separating mixed signals in complex electromagnetic environment, a blind source separation algorithm based on degree of cyclostationarity (DCS) criterion is constructed in this paper. Firstly, the DCS criterion is constructed by using the cyclic spectrum theory. Then the algorithm flow of blind source separation is designed based on DCS criterion. At the same time, Givens matrix is constructed to make the blind source separation algorithm suitable for multiple signals with different cyclostationary frequencies. The feasibility of this method is further proved. The theoretical and simulation results show that the algorithm can effectively separate and recognize common multi-radar signals.

Dempster-Shafer evidence theory is broadly employed in the research of multi-source information fusion. Nevertheless, when fusing highly conflicting evidence it may produce counterintuitive outcomes. To address this issue, a fusion approach based on a newly defined belief exponential divergence and Deng entropy is proposed. First, a belief exponential divergence is proposed as the conflict measurement between evidences. Then, the credibility of each evidence is calculated. Afterwards, the Deng entropy is used to calculate information volume to determine the uncertainty of evidence. Then, the weight of evidence is calculated by integrating the credibility and uncertainty of each evidence. Ultimately, initial evidences are amended and fused using Dempster’s rule of combination. The effectiveness of this approach in addressing the fusion of three typical conflict paradoxes is demonstrated by arithmetic examples. Additionally, the proposed approach is applied to aerial target recognition and iris dataset-based classification to validate its efficacy. Results indicate that the proposed approach can enhance the accuracy of target recognition and effectively address the issue of fusing conflicting evidences.

In order to solve the problem of uncertainty and fuzzy information in the process of weapon equipment system selection, a multi-attribute decision-making (MADM) method based on probabilistic hesitant fuzzy set (PHFS) is proposed. Firstly, we introduce the concept of probability and fuzzy entropy to measure the ambiguity, hesitation and uncertainty of probabilistic hesitant fuzzy elements (PHFEs). Sequentially, the expert trust network is constructed, and the importance of each expert in the network can be obtained by calculating the cumulative trust value under multiple trust propagation paths, so as to obtain the expert weight vector. Finally, we put forward an MADM method combining the probabilistic hesitant fuzzy entropy and grey relation analysis (GRA) model, and an illustrative case is employed to prove the feasibility and effectiveness of the method when solving the weapon system selection decision-making problem.

Aiming at the characteristics of multi-stage and (extremely) small samples of the identification problem of key effectiveness indexes of weapon equipment system-of-systems (WESoS), a Bayesian intelligent identification and inference model for system effectiveness assessment indexes based on dynamic grey incidence is proposed. The method uses multilayer Bayesian techniques, makes full use of historical statistics and empirical information, and determines the Bayesian estimation of the incidence degree of indexes, which effectively solves the difficulties of small sample size of effectiveness indexes and difficulty in obtaining incidence rules between indexes. Secondly, The method quantifies the incidence relationship between evaluation indexes and combat effectiveness based on Bayesian posterior grey incidence, and then identifies key system effectiveness evaluation indexes. Finally, the proposed method is applied to a case of screening key effectiveness indexes of a missile defensive system, and the analysis results show that the proposed method can fuse multi-moment information and extract multi-stage key indexes, and has good data extraction capability in the case of small samples.

Failure mode and effect analysis (FMEA) is a preventative risk evaluation method used to evaluate and eliminate failure modes within a system. However, the traditional FMEA method exhibits many deficiencies that pose challenges in practical applications. To improve the conventional FMEA, many modified FMEA models have been suggested. However, the majority of them inadequately address consensus issues and focus on achieving a complete ranking of failure modes. In this research, we propose a new FMEA approach that integrates a two-stage consensus reaching model and a density peak clustering algorithm for the assessment and clustering of failure modes. Firstly, we employ the interval 2-tuple linguistic variables (I2TLVs) to express the uncertain risk evaluations provided by FMEA experts. Then, a two-stage consensus reaching model is adopted to enable FMEA experts to reach a consensus. Next, failure modes are categorized into several risk clusters using a density peak clustering algorithm. Finally, the proposed FMEA is illustrated by a case study of load-bearing guidance devices of subway systems. The results show that the proposed FMEA model can more easily to describe the uncertain risk information of failure modes by using the I2TLVs; the introduction of an endogenous feedback mechanism and an exogenous feedback mechanism can accelerate the process of consensus reaching; and the density peak clustering of failure modes successfully improves the practical applicability of FMEA.

Cascading failures in infrastructure networks have serious impacts on network function. The limited capacity of network nodes provides a necessary condition for cascade failure. However, the network capacity cannot be infinite in the real network system. Therefore, how to reasonably allocate the limited capacity resources is of great significance. In this article, we put forward a capacity allocation strategy based on community structure against cascading failure. Experimental results indicate that the proposed method can reduce the scale of cascade failures with higher capacity utilization compared with Motter-Lai (ML) model. The advantage of our method is more obvious in scale-free network. Furthermore, the experiment shows that the cascade effect is more obvious when the vertex load is randomly varying. It is known to all that the growth of network capacity can make the network more resistant to destruction, but in this paper it is found that the contribution rate of unit capacity rises first and then decreases with the growth of network capacity cost.

Future unmanned battles desperately require intelligent combat policies, and multi-agent reinforcement learning offers a promising solution. However, due to the complexity of combat operations and large size of the combat group, this task suffers from credit assignment problem more than other reinforcement learning tasks. This study uses reward shaping to relieve the credit assignment problem and improve policy training for the new generation of large-scale unmanned combat operations. We first prove that multiple reward shaping functions would not change the Nash Equilibrium in stochastic games, providing theoretical support for their use. According to the characteristics of combat operations, we propose tactical reward shaping (TRS) that comprises maneuver shaping advice and threat assessment-based attack shaping advice. Then, we investigate the effects of different types and combinations of shaping advice on combat policies through experiments. The results show that TRS improves both the efficiency and attack accuracy of combat policies, with the combination of maneuver reward shaping advice and ally-focused attack shaping advice achieving the best performance compared with that of the baseline strategy.

Blockchain technology has attracted worldwide attention, and has strong application potential in complex product system supply chain and other fields. This paper focuses on the supply chain management issues of complex product systems, and combines the technical characteristics of blockchain, such as tamper resistance and strong resistance to destruction, to conduct research on the application of blockchain based supply chain management for complex product systems. The blockchain technology is integrated into functional modules such as business interaction, privacy protection, data storage, and system services. The application technology architecture of complex product system supply chain integrated with blockchain is constructed. The application practice in complex product system supply chain is carried out. The results show that the supply chain of complex product systems has the functions of traceability, cost reduction, and anti-counterfeiting protection. Finally, the future development direction and research focus of the complex product system supply chain based on blockchain are prospected, which provides a reference for the equipment manufacturing supply chain management in the military industry.

With the advantage of exceptional long-range traffic perception capabilities and data fusion computational prowess, the cloud control system (CCS) has exhibited formidable potential in the realm of connected assisted driving, such as the adaptive cruise control (ACC). Based on the CCS architecture, this paper proposes a cloud-based predictive ACC (PACC) strategy, which fully considers the road slope information and the preceding vehicle status. In the cloud, based on the dynamic programming (DP), the long-term economic speed planning is carried out by using the slope information. At the vehicle side, the real-time fusion planning of the economic speed and the preceding vehicle state is realized based on the model predictive control (MPC), taking into account the safety and economy of driving. In order to ensure the safety and stability of the vehicle-cloud cooperative control system, an event-triggered cruise mode switching method is proposed based on the state of each subsystem of the vehicle-cloud-network-map. Simulation results indicate that the PACC system can still ensure stable cruising under delays and some complex conditions. Moreover, under normal conditions, compared to the ACC system, the PACC system can further improve economy while ensuring safety and improve the overall energy efficiency of the vehicle, thus achieving fuel savings of 3% to 8%.

This paper mainly focuses on stability analysis of the nonlinear active disturbance rejection control (ADRC)-based control system and its applicability to real world engineering problems. Firstly, the nonlinear ADRC(NLADRC)-based control system is transformed into a multi-input multi-output (MIMO) Lurie-like system, then sufficient condition for absolute stability based on linear matrix inequality (LMI) is proposed. Since the absolute stability is a kind of global stability, Lyapunov stability is further considered. The local asymptotical stability can be determined by whether a matrix is Hurwitz or not. Using the inverted pendulum as an example, the proposed methods are verified by simulation and experiment, which show the valuable guidance for engineers to design and analyze the NL ADRC-based control system.

For the multicopter with more than four rotors, the rotor fault information is unobservable, which limits the application of active fault-tolerant on multicopters. This paper applies an existing fault-tolerant control method for quadcopter to multicopter with more than four rotors. Without relying on rotor fault information, this method is able to stabilize the multicopter with multiple rotor failures, which is validated on the hexacopter and octocopter using the hardware-in-the-loop simulations. Additionally, the hardware-in-the-loop simulations demonstrate that a more significant tilt angle in flight will inhibit the maximum tolerable number of rotor failures of a multicopter. The more significant aerodynamic drag moment will make it difficult for the multicopter to regain altitude control after rotor failure.

Small video satellites have unique advantages of short development cycle, agile attitude maneuver, real-time video imaging. They have broad application prospects in space debris, faulty spacecraft, and other space target detection and tracking. However, when a space target first enters the camera’s visual field, it has a relatively large angular velocity relative to the satellite, which makes it easy to deviate from the visual field and cause off-target problems. This paper proposes a novel visual tracking control method based on potential function preventing missed targets in space. Firstly, a circular area in the image plane is designed as a mandatory restricted projection area of the target and a visual tracking controller based on image error. Then, a potential function is designed to ensure continuous and stable tracking of the target after entering the visual field. Finally, the stability of the control is proved using Barbarat’s lemma. By setting the same conditions and comparing with the simulation results of the proportion-derivative (PD) control method, the results show that when there is a large relative attitude motion angular velocity between the target and the satellite, the tracking method based on potential function can ensure that the target does not deviate from the field-of-view during the tracking control process, and the projection of target is controlled to the desired position. The proposed control method is effective in eliminating tracking error and preventing off-target simultaneously.

In consideration of the field-of-view (FOV) angle constraint, this study focuses on the guidance problem with impact time control. A deep reinforcement learning guidance method is given for the missile to obtain the desired impact time and meet the demand of FOV angle constraint. On basis of the framework of the proportional navigation guidance, an auxiliary control term is supplemented by the distributed deep deterministic policy gradient algorithm, in which the reward functions are developed to decrease the time-to-go error and improve the terminal guidance accuracy. The numerical simulation demonstrates that the missile governed by the presented deep reinforcement learning guidance law can hit the target successfully at appointed arrival time.

To meet the requirements of modern air combat, an integrated fire/flight control (IFFC) system is designed to achieve automatic precision tracking and aiming for armed helicopters and release the pilot from heavy target burden. Considering the complex dynamic characteristics and the couplings of armed helicopters, an improved automatic attack system is constructed to integrate the fire control system with the flight control system into a unit. To obtain the optimal command signals, the algorithm is investigated to solve nonconvex optimization problems by the contracting Broyden Fletcher Goldfarb Shanno (C-BFGS) algorithm combined with the trust region method. To address the uncertainties in the automatic attack system, the memory nominal distribution and Wasserstein distance are introduced to accurately characterize the uncertainties, and the dual solvable problem is analyzed by using the duality theory, conjugate function, and dual norm. Simulation results verify the practicality and validity of the proposed method in solving the IFFC problem on the premise of satisfactory aiming accuracy.