Lightweight convolutional neural networks (CNNs) have simple structures but struggle to comprehensively and accurately extract important semantic information from images. While attention mechanisms can enhance CNNs by learning distinctive representations, most existing spatial and hybrid attention methods focus on local regions with extensive parameters, making them unsuitable for lightweight CNNs. In this paper, we propose a self-attention mechanism tailored for lightweight networks, namely the brief self-attention module (BSAM). BSAM consists of the brief spatial attention (BSA) and advanced channel attention blocks. Unlike conventional self-attention methods with many parameters, our BSA block improves the performance of lightweight networks by effectively learning global semantic representations. Moreover, BSAM can be seamlessly integrated into lightweight CNNs for end-to-end training, maintaining the network’s lightweight and mobile characteristics. We validate the effectiveness of the proposed method on image classification tasks using the Food-101, Caltech-256, and Mini-ImageNet datasets.

The power inversion (PI) algorithm lacks specific constraints on desired signals. Thus, the beampattern has fluctuation in all directions other than the jamming sources. This phenomenon will damage the reception of desired signals. In high signal-to-noise ratio (SNR) application, the desired signal is inevitably suppressed by the PI algorithm, resulting in a deterioration to the out signal-to-interference-and-noise ratio (SINR). This paper proposes an improved PI algorithm based on derivative constraint. Firstly, the proposed method uses subspace projection to extract jamming-free data, the derivative constraint is imposed to the non-jamming data, and subsequently the Lagrange multiplier can be used to calculate the array weight vector. Simulation results demonstrate that, the proposed algorithm in this paper has a higher output SNR, flat gains in non-jamming directions, and applicability of high SINR than the PI algorithm, thus verifying the effectiveness of the algorithm.

Automatic modulation classification(AMC) is an essential technique in both civil and military applications. While deep learning has surpassed traditional methods in accuracy, distinguishing high-order modulations remain challenging. Current efforts prioritize complex network designs, neglecting the integration of deep features and tailored feature engineering to reslove high-order ambiguities. Therefore, a multi-feature extraction framework is proposed, which directly concatenates the deep feature extracted by a newly designed lightweight neural network and the proposed spectrum secondary features or de-noised high-order statistical features. The proposed features and lightweight network both demonstrate superior overall accuracy than other competing features or networks. Furthermore, the effectiveness of the feature extraction framework is also validated. The average classification accuracy on high-order modulation sets reaches 67.39% on the RadioML2018.01A dataset, increasing more than 2% compared with the other competitive networks under the framework. The results indicate the effectiveness of the proposed feature extraction framework for its representational ability by combing the deep features with the proposed domain features.

This paper introduces a hybrid configuration design to enhance the precision of satellite antenna position measurement. By fixing the circular array antenna on the antenna mounting surface and integrating coordinate system transformation relationships with interferometric direction finding (DF) and positioning technology, accurate estimation of the antenna position is ensured. This method optimizes the quality and stability of data fusion by integrating pulse parameter characteristics, satellite orbit and attitude information, as well as the field of view information from observation stations, using techniques such as maximum-ratio-combining (MRC) and orbit extrapolation. Specifically, the sampling-importance resampling particle-filtering and Kalman-filtering (SIR-PF-KF) hybrid filtering prediction technology is employed to precisely predict and correct the three-dimensional (3D) position errors of the L-array antenna. Through data processing of five to nine orbits, accurate estimation of the antenna’s 3D position is achieved, achieving an estimation accuracy of 3 μm, significantly improving the accuracy of on-orbit rapid calibration. Experimental results show that the interferometer positioning accuracy is improved from 7.9 km before antenna position correction to within 0.2 km after correction, verifying the effectiveness and practicability of this method, which aims to address issues with positioning accuracy.

The Global Position System (GPS) is a reliable method for positioning in most scenarios, but it falls short in harsh environments like urban vehicular scenarios, where numerous trees or flyovers obstruct the signals. This presents an unprecedented challenge for autonomous vehicles or applications requiring high accuracy. Fortunately, vehicular ad-hoc networks (VANET) offer an effective solution, where vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communications are used to enhance location awareness. In V2I communications, the roadside units (RSU) transmit beacon packets, and the vehicle receives numerous packets from different RSUs to establish communication. To further improve localization accuracy, a cross-covariance matrices-alternating least square (CCM-ALS) algorithm is proposed. The algorithm relies on ALS of the CCM for obtaining the position of vehicles in V2I communications. The algorithm is highly precise compared to traditional angle of arrival (AOA) positioning and not inferior to direct position determination (DPD) approaches while being low in complexity, which is crucial for moving vehicles. The numerical results verify the superiority of the proposed method.

During actual high-speed flights, the electromagnetic (EM) properties of aircraft radomes are influenced by dielectric temperature drift, leading to substantial drift in the boresight errors (BSEs) from their room temperature values. However, applying thermal loads to the radome during ground-based EM simulation tests is challenging. This paper presents an EM equivalent physical model (EEPM) for high-speed aircraft radomes that account for the effects of dielectric temperature drift. This is achieved by attaching dielectric slices of specific thicknesses to the outer surface of a room-temperature radome (RTR) to simulate the increase in electrical thickness resulting from high temperatures. This approach enables accurate simulations of the BSEs of high-temperature radomes (HTRs) under high-speed flight conditions. An application example, supported by full-wave numerical calculations and physical testing, demonstrates that the EEPM exhibits substantial improvement in approximating the HTR compared to the RTR, facilitating precise simulations of the BSEs of HTRs during high-speed flights. Overall, the proposed EEPM is anticipated to considerably enhance the alignment between the ground-based simulations of high-speed aircraft guidance systems and their actual flight conditions.

Exact estimation of space object attitude parameters is a great challenge. The effectiveness of conventional attitude estimation approaches based on target sizes suffers a significant reduction when occlusion exists. This paper proposes an innovative approach to estimate the attitude parameters for space objects based on inverse synthetic aperture radar (ISAR) image sequences. The formulation for nonlinear size constraints (NSC) is developed by accounting for the characteristics of object size variation in ISAR image sequences. The multi-start framework for global optimization and the Broyden-Fletcher-Goldfarb-Shanno (BFGS) based quasi-Newton iterative method are combined with and used for more accurate estimation of space object’s attitude parameters. Furthermore, the Cramer-Rao lower bound (CRLB) of attitude parameter estimates is derived. Comparative experiments demonstrate the effectiveness and robustness of the proposed method.

In the traditional radar unmanned aerial vehicle (UAV) detection process, coherent integration and micro-Doppler (m-D) parameter estimation are carried out separately. The target discrimination process needs to obtain the position information of the target, which will lose energy. In this paper, a long time integration method of radar signal based on rotating target radon Fourier transform (RTRFT) is proposed. This method modifies the distance and frequency terms in the traditional generalized radon Fourier transform (GRFT), and adds the frequency sinusoidal modulation term. Then, based on the cardinality balanced multi-target multi-Bernoulli (CBMeMBer) filter, the position of the target is detected in the high-dimensional space obtained by RTRFT. This method can combine coherent integration and micro-motion parameter estimation. Simulation experiments show that the proposed method can estimate the main translational parameters and rotational micro-motion parameters of the target while detecting the target, and the target detection performance is improved.

Electronic reconnaissance units commonly utilize an interferometer direction-finding system to measure the incoming direction of radar radiation signals. This approach enables the accurate determination of threat source locations, which is essential for devising route plans oriented toward flight path generation. When a frequency diverse array (FDA) system is adopted by ground radars, errors are introduced into the angle measurements of the passive direction finding system. To address this issue, this study starts with FDA model establishment and equiphasic surface characteristics analysis and analyzes the principles of FDA deception in identifying one-dimensional single-baseline interferometer directions. Additionally, the Cramer-Rao bounds of the signal carrier frequency estimation error and angle measurement error during the interferometer’s direction finding process are considered. The simulation results verify that the one-dimensional single-baseline interferometer direction finding system can be deceived by the FDA radar, and the FDA with a sine frequency offset exhibits the optimum deception effect.

This paper addresses a dynamic vehicle routing problem with stochastic requests in a dual-channel distribution center that utilizes shared vehicle resources to serve two types of customers: offline corporate clients (CCs) with fixed and stochastic batch demands, and online individual customers (ICs) with single-unit demands. To manage stochastic batch demands from CCs, this paper proposes three recourse policies under a differentiated resource-sharing scheme: the waiting-tour-based (WTB) policy, the advance-tour-based (ATB) policy, and the advance-customer-based (ACB) policy. These policies differ in their response priorities to random requests and the scope of route reoptimization. The problem is formulated as a two-stage stochastic recourse programming model, where the first stage establishes routes for fixed demands. In the second stage, we construct three stochastic recourse programming models corresponding to the proposed recourse policies. To solve these models, this paper develop rolling horizon algorithms integrated with mathematical programming models or metaheuristic algorithms. Extensive numerical experiments validate the effectiveness of the proposed algorithms and policies. The results indicate that both the ATB and ACB policies lead to cost savings compared to the WTB policy, especially when stochastic demands are urgent and delivery resources are quite limited. Specifically, when the number of ICs is small, the expected total cost savings can exceed 12%, and in some scenarios, savings of over 20% can be achieved. When the number of ICs is large, some scenarios can achieve cost savings exceeding 7%. Furthermore, the ACB policy yields lower costs, fewer worsened ICs, fewer trips, and less vehicle time than the ATB policy.

In an aircraft final assembly line (AFAL), the rational scheduling of assembly workers to complete tasks in an orderly manner is crucial for enhancing production efficiency. This paper addresses the multi-skilled worker scheduling problem in the AFAL, where the processing time of each task varies due to the assigned workers’ skill levels, referred to as variable duration. The objective is to minimize the makespan, i.e., the total time required for all workers to complete all tasks. A mixed integer linear programming model is formulated under complex constraints including assembly precedence relations, skill requirements, worker skill capabilities, and workspace capacities. To solve the model effectively, a multi-pass priority rule-based heuristic (MPRH) algorithm is proposed. This algorithm integrates 14 activity priority rules and nine worker priority rules with worker weights. Extensive experiments iteratively the best-performing priority rules, and the most effective rule subsets are integrated through a lightweight multi-pass mechanism to enhance its efficiency. The computational results demonstrate that the MPRH can find high-quality solutions effectively within very short central processing unit central processing unit (CPU) time compared to GUROBI. A case study based on real data obtained from an AFAL confirms the necessity and the feasibility of the approach in practical applications. Sensitivity analyses provide valuable insights to real production scenarios.

Aiming at the characteristics of autonomy, confrontation, and uncertainty in unmanned aerial vehicle (UAV) swarm operations, case-based reasoning (CBR) technology with advantages such as weak dependence on domain knowledge and efficient problem-solving is introduced, and a recommendation method for UAV swarm operation strategies based on CBR is proposed. Firstly, we design a universal framework for UAV swarm operation strategies from three dimensions: operation effectiveness, time, and cost. Secondly, based on the representation of operation cases, certain, fuzzy, interval, and classification attribute similarity calculation methods, as well as entropy-based attribute weight allocation methods, are suggested to support the calculation of global similarity of cases. This method is utilized to match the source case with the most similarity from the historical case library, to obtain the optimal recommendation strategy for the target case. Finally, in the form of red blue confrontation, a UAV swarm operation strategy recommendation case is constructed based on actual battle cases, and a system simulation analysis is conducted. The results show that the strategy given in the example performs the best in three evaluation indicators, including cost-effectiveness, and overall outperforms other operation strategies. Therefore, the proposed method has advantages such as high real-time performance and interpretability, and can address the issue of recommending UAV swarm operation strategies in complex battlefield environments across both online and offline modes. At the same time, this study could also provide new ideas for the selection of UAV swarm operation strategies.

The precision and quality of machining in computer numerical control (CNC) machines are significantly impacted by the state of the tool. Therefore, it is essential and crucial to monitor the tool’s condition in real time during operation. To improve the monitoring accuracy of tool wear values, a tool wear monitoring approach is developed in this work, which is based on an improved integrated model of densely connected convolutional network (DenseNet) and gated recurrent unit (GRU), which incorporates data preprocessing via wavelet packet transform (WPT). Firstly, wavelet packet decomposition (WPD) is used to extract time-frequency domain features from the original time-series monitoring signals of the tool. Secondly, the multidimensional deep features are extracted from DenseNet containing asymmetric convolution kernels, and feature fusion is performed. A dilation scheme is employed to acquire more historical data by utilizing dilated convolutional kernels with different dilation rates. Finally, the GRU is utilized to extract temporal features from the extracted deep-level signal features, and the feature mapping of these temporal features is then carried out by a fully connected neural network, which ultimately achieves the monitoring of tool wear values. Comprehensive experiments conducted on reference datasets show that the proposed model performs better in terms of accuracy and generalization than other cutting-edge tool wear monitoring algorithms.

In this paper, a grey Kalman filter model is proposed for lithium battery charge state estimation. Firstly, this paper establishes a recursive relation equation between the front and back terms through the grey model (GM). Secondly, the state space expression is constructed based on the recursive relationship equation. Next, the Kalman filter algorithm is integrated to form a grey Kalman filter model. Finally, the charge state is estimated based on public lithium battery data. In this paper, the state of charge is estimated from three different aspects, including different driving cycles, randomly mixed driving cycles, and the estimation of the state of charge by different temperatures under the same driving cycle conditions. On this basis, the model is applied to a life scenario using the charge state of 20 electric vehicles. The results show that the proposed model has good accuracy.

Agile earth observation satellites (AEOSs) represent a new generation of satellites with three degrees of freedom (pitch, roll, and yaw); they possess a long visible time window (VTW) for ground targets and support imaging at any moment within the VTW. However, different observation times demonstrate different cloud cover distributions, which exhibit different effects on the AEOS observation. Previous studies ignored pitch angles, discretized VTWs, or fixed cloud cover for every VTW, which led to the loss of intermediate observation states, thus these studies are not suitable for AEOS scheduling considering cloud cover distribution. In this study, a relationship formula between the cloud cover and observation time is proposed to calculate the cloud cover for every observation time, and a relationship formula between the observation time and pitch angle is designed to calculate the pitch angle for every observation time in the VTW. A refined model including the pitch angle, roll angle, and cloud cover distribution is established, which can make the scheme closer to the actual application of AEOSs. A hybrid genetic simulated annealing (HGSA) algorithm for AEOS scheduling is proposed, which integrates the advantages of genetic and simulated annealing algorithms and can effectively avoid falling into a local optimal solution. The experiments are conducted to compare the proposed algorithm with the traditional algorithms, the results verify that the proposed model and algorithm are efficient and effective for AEOS scheduling considering cloud cover distribution.

To overcome the limitations of conventional approaches that adopt monolithic architectures and overlook critical dynamic interactions in evaluating combat effectiveness and subsystem contributions within amphibious operations, this paper proposes an integrated framework combining complex system network modeling with dynamic adversarial simulation for evaluating mission-critical system-of-systems (SoS). Specifically, the contribution rate of unmanned aerial vehicles (UAVs) to the amphibious joint landing SoS (AJLSoS) is quantified. Firstly, a standardized network topology model is developed using operation loop theory, systematically characterizing node functionalities and their interdependencies. Secondly, the ideal Lanchester equation is augmented according to the model’s static operational capability, and an amphibious operational simulation model is constructed based on the modified equation, enabling dynamic simulation of force attrition and engagement duration as key performance indicators of AJLSoS. To validate the theoretical framework, a battalion-level amphibious campaign scenario is developed to compute effectiveness metrics across multiple control scenarios and the contribution rate of UAVs to AJLSoS is analyzed. This study not only provides actionable insights for operational mission planning of UAVs in the context of amphibious operations but also demonstrates high adaptability to diverse operational contexts.

A cooperative guidance law is proposed in a two-on-two engagement scenario with large-heading-errors by choosing zero-effort miss distance as a sliding surface, which consists of an attacker, a protector, a defender, and a target, based on fixed-time sliding mode control theory. Based on the nonlinear method of fixed-time sliding mode control, the performance of the cooperative guidance law remains satisfactory even with large-heading-errors scenarios where the linearization-based approaches might be invalid. By virtue of this law, the attacker pursues the target with the assistance of the protector, which can intercept the defender in the engagement scenario. Furthermore, if the attacker is intercepted by the defender, the guidance law of the protector could guarantee that the protector attacks the target. A robust adaptive term is included in the guidance law to deal with the case of the unknown disturbance upper bound of the defender-target team. Finally, the feasibility of the guidance law is verified by nonlinear numerical simulations, and the superiority of it is illustrated by comparing with the linearization guidance law.

As the Mars probe, which has limited on-board ability in computation is unable to carry out the large-scale landmark solution, it is necessary to achieve optimal selection of landmarks while ensuring autonomous navigation accuracy during landing phase. This paper proposes an optimal landmark selection method based on the observability matrix for the Mars probe. Firstly, an observability matrix for navigation system is constructed with Fisher information quantity. Secondly, the optimal configuration of the landmark distribution is given by maximizing the scalar function of the observability matrix. Based on the optimal configuration, the greedy algorithm is used to determine the number of the landmarks at each moment adaptively. In addition, considering the fact that the number of the observable landmarks gradually decreases during the landing process, the convergence threshold of the greedy algorithm is set to a dynamic value regarding landing time. Finally, mathematical simulation verification is conducted, and the results show that the proposed optimal landmark selection method has higher navigation accuracy compared with the random landmark selection method. It can effectively suppress the influence of the measurement model errors and achieve a higher landing accuracy.

Two-photon fluorescence microscopy, based on the principles of two-photon excited fluorescence and second harmonic generation, enables real-time non-invasive in vivo imaging of skin and cells, providing a means to assess human health status. In this paper, a miniaturized two-photon imaging system is designed and fabricated to withstand extreme vibration and shock environments. The mechanical stability of the optical and structural components of the miniaturized probe is evaluated under random vibration and shock vibration tests using finite element simulation methods and ray tracing techniques. During the environmental testing, the maximum stress on the probe is 11.5 MPa, which is well below the threshold for structural failure. The largest structural displacement occurs at the collimator, where random vibrations produce an offset of 10.9 μm. This offset is analyzed by using geometric optics and point spread functions. Under the maximum collimator offset, the theoretical resolution, as calculated by the point spread function, shifted from 463.28 nm to 463.48 nm. Additionally, a lateral offset of 127 nm is observed at the center position, which does not significantly impact the imaging performance. Finally, environmental and imaging performance tests are conducted. The system’s measured resolution after the environmental tests is 530 nm, consistent with its resolution prior to testing. Imaging tests are also performed on the skin’s stratum corneum, granular layer, spinous layer, and basal cell layer, revealing clear cellular structural information. These results confirm the device’s potential for applications in extreme shock and vibration environments.

This paper proposes a differential-fatness-based active disturbance rejection control (ADRC) for high-speed steering control of tracked tank systems. Firstly, a high-speed steering model is established by considering the lateral component of the centrifugal force acting on the tank on the basis of modeling and analyzing the dynamic model of the low-speed steering system. Secondly, we propose a differential-flatness ADRC approach by converting the under-actuated system to a fully driven flat one. Moreover, we prove the differential flatness of the steering system, which facilitates a two-channel ADRC development. Finally, we show that both the states of the flat system and the original under-actuated system can track the reference trajectory. On the external interference condition, the system is observed to re-track the target signal within 2 s.

This paper investigates the high-performance control issues of systems affected by time-varying disturbances and measurement noise. Conventionally, active disturbance rejection control (ADRC) is a favorable control strategy to reject unknown disturbances and uncertainties. However, its control performance is limited because standard extended state observer (ESO) struggles to effectively estimate time-varying disturbances. The emergence of high-order ESO (HESO) alleviates the limitation. Unfortunately, it deteriorates the noise suppression capability when the disturbance rejection is enhanced. To tackle this challenge, an improved ADRC with cascade HESO (CHESO) is proposed. A comprehensive theoretical analysis associated with the performance of HESO is given for the first time. The presented analyses provide an intuitive understanding of the performance of HESO. Then, a novel CHESO is developed. The convergence of CHESO is proved via input-to-state stable theory. Extensive frequency domain analyses indicate that CHESO has stronger disturbance rejection and high-frequency noise attenuation performance than ESO and HESO without increasing the observer bandwidth. Comparative simulations conducted on a servo control system validate the effectiveness and preponderance of the proposed method.

Aiming at the terminal defense problem of aircraft, this paper proposes a method to simultaneously achieve terminal defense and seize the dominant position. The method employs a λ-return based reinforcement learning algorithm, which can be applied to the flight assistance decision-making system to improve the pilot’s survivability. First, we model the environment to simulate the interaction between air-to-air missiles and aircraft. Subsequently, we propose a λ-return based approach to improve the deep Q learning network (DQN), deep advantageous actor criticism (A2C), and proximity policy optimization (PPO) algorithms used to train manoeuvre strategies. The method employs an action space containing nine manoeuvres and defines the off-target distance at the end of the scene as a sparse reward for algorithm training. Simulation results show that the convergence speed of the three improved algorithms is significantly improved when using the λ-return method. Moreover, the effect of the fetch value on the convergence speed is verified by ablation experiments. In order to solve the illegal behavior problem in the training process, we also design a backtracking-based illegal behavior masking mechanism, which improves the data generation efficiency of the environment model and promotes effective algorithm training.