This paper investigates the distributed continuous-time aggregative optimization problem for second-order multi-agent systems, where the local cost function is not only related to its own decision variables, but also to the aggregation of the decision variables of all the agents. By using the gradient descent method, the distributed average tracking (DAT) technique and the time-base generator (TBG) technique, a distributed continuous-time aggregative optimization algorithm is proposed. Subsequently, the optimality of the system’s equilibrium point is analyzed, and the convergence of the closed-loop system is proved using the Lyapunov stability theory. Finally, the effectiveness of the proposed algorithm is validated through case studies on multirobot systems and power generation systems.

Re-entry gliding vehicles exhibit high maneuverability, making trajectory prediction a key factor in the effectiveness of defense systems. To overcome the limited fitting accuracy of existing methods and their poor adaptability to maneuver mode mutations, a trajectory prediction method is proposed that integrates online maneuver mode identification with dynamic modeling. Characteristic parameters are extracted from tracking data for parameterized modeling, enabling real-time identification of maneuver modes. In addition, a maneuver detection mechanism based on higher-order cumulants is introduced to detect lateral maneuver mutations and optimize the use of historical data. Simulation results show that the proposed method achieves accurate trajectory prediction during the glide phase and maintains high accuracy under maneuver mutations, significantly enhancing the prediction performance of both three-dimensional trajectories and ground tracks.

Formation control of multiple spacecraft has attracted extensive research attention. However, achieving reliable performance under sensor failures remains a significant challenge. This paper develops an integrated framework that jointly designs distributed observers and local controllers to ensure robust formation control in the presence of external disturbances and sensor malfunctions. Treating the spacecraft formation as a single interconnected system, each spacecraft constructs a distributed observer that estimates the overall system state by incorporating both its own measurements and the predicted control information shared among the spacecraft. Based on the observer estimates, a local control law is synthesized to maintain the desired formation. Rigorous theoretical analysis and numerical simulations demonstrate that the proposed integrated approach effectively guarantees formation stability and resilience against sensor failures and disturbances.

In response to the challenges faced by unmanned swarms in mountain obstacle-breaching missions within complex terrains, such as poor task-resource coupling, lengthy solution generation times, and poor inter-platform collaboration, an unmanned swarm scheduling strategy tailored is proposed for mountain obstacle-breaching missions. Initially, by formalizing the descriptions of obstacle breaching operations, the swarm, and obstacle targets, an optimization model is constructed with the objectives of expected global benefit, timeliness, and task completion degree. A meta-task decomposition and reassembly strategy is then introduced to more precisely match the capabilities of unmanned platforms with task requirements. Additionally, a meta-task decomposition optimization model and a meta-task allocation operator are incorporated to achieve efficient allocation of swarm resources and collaborative scheduling. Simulation results demonstrate that the model can accurately generate reasonable and feasible obstacle breaching execution plans for unmanned swarms based on specific task requirements and environmental conditions. Moreover, compared to conventional strategies, the proposed strategy enhances task completion degree and expected returns while reducing the execution time of the plans.

Bird’s-eye-view (BEV) perception is a core technology for autonomous driving systems. However, existing solutions face the dilemma of high costs associated with multi-modal methods and limited performance of vision-only approaches. To address this issue, this paper proposes a framework named “a lightweight pure visual BEV perception method based on dual distillation of spatial-temporal knowledge”. This framework innovatively designs a lightweight vision-only student model based on ResNet, which leverages a dual distillation mechanism to learn from a powerful teacher model that integrates temporal information from both image and light detection and ranging (LiDAR) modalities. Specifically, we distill efficient multi-modal feature extraction and spatial fusion capabilities from the BEVFusion model, and distill advanced temporal information fusion and spatiotemporal attention mechanisms from the BEVFormer model. This dual distillation strategy enables the student model to achieve perception performance close to that of multi-modal models without relying on LiDAR. Experimental results on the nuScenes dataset demonstrate that the proposed model significantly outperforms classical vision-only algorithms, achieves comparable performance to current state-of-the-art vision-only methods on the nuScenes detection leaderboard in terms of both mean average precision (mAP) and the nuScenes detection score (NDS) metrics, and exhibits notable advantages in inference computational efficiency. Although the proposed dual-teacher paradigm incurs higher offline training costs compared to single-model approaches, it yields a streamlined and highly efficient student model suitable for resource-constrained real-time deployment. This provides an effective pathway toward low-cost, high-performance autonomous driving perception systems.

A safe and reliable path planning algorithm is fundamental for unmanned surface vehicles (USVs) to perform autonomous navigation tasks. However, a single global or local planning strategy cannot fully meet the requirements of complex maritime environments. Global planning alone cannot effectively handle dynamic obstacles, while local planning alone may fall into local optima. To address these issues, this paper proposes a multi-dynamic-obstacle avoidance path planning method that integrates an improved A* algorithm with the dynamic window approach (DWA). The traditional A* algorithm often generates paths that are too close to obstacle boundaries and contain excessive turning points, whereas the traditional DWA tends to skirt densely clustered obstacles, resulting in longer routes and insufficient dynamic obstacle avoidance. To overcome these limitations, improved versions of both algorithms are developed. Key points extracted from the optimized A* path are used as intermediate start-destination pairs for the improved DWA, and the weights of the DWA evaluation function are adjusted to achieve effective fusion. Furthermore, a multi-dynamic-obstacle avoidance strategy is designed for complex navigation scenarios. Simulation results demonstrate that the USV can adaptively switch between dynamic obstacle avoidance and path tracking based on obstacle distribution, validating the effectiveness of the proposed method.

Mega low Earth orbit (LEO) satellite networks serve as effective complements to terrestrial networks. However, the dual mobility of users and LEO satellites makes inter-satellite handovers more frequent for users. Moreover, there are both ascending and descending segments in widely deployed walker-delta constellations. Even if the locations of users do not change, when the access satellites of the communicating parties are not in the same ascending or descending segment, the end-to-end latency between them will increase. To address this challenge, the self-decision handover (SDH) strategy and the joint decision handover (JDH) strategy are proposed, and they both incorporate the routing hops as a crucial handover criterion to minimize the end-to-end latency. In addition, the shortest route hop-count algorithm is designed to assist in the handover decision-making process. Simulations demonstrate that the proposed handover strategies outperform the traditional handover strategies in terms of the number of handovers and end-to-end latency.

A methodology for the reduction of radar cross section (RCS) of cambered platforms within the target airspace is presented, which utilizes a dual-polarized ultra-wide-angle artificial electromagnetic absorbing surface. By applying the theory of generalized Brewster complex wave impedance matching, five distinct unit cell designs are developed to attain more than 95% absorption rate for dual-polarized incident waves within five angular ranges: 0°?30°, 30°?50°, 50°?60°, 60°?70°, and 70°?80°. To optimally reduce the RCS of a cambered platform, the five types of units can be evenly distributed on the surface based on the local incident angles of plane waves originating from the target airspace. As an illustrative example, the leading edge of an airfoil is taken into account, and experimental measurements validate the efficiency of the proposed structure. Specifically, the absorbing surface achieves more than 10 dB of RCS reduction in the frequency ranges from 5-10 GHz (about 66.7% relative bandwidth) for dual polarizations.

Weakly supervised semantic segmentation (WSSS) is a tricky task, which only provides category information for segmentation prediction. Thus, the key stage of WSSS is to generate the pseudo labels. For convolutional neural network (CNN) based methods, in which class activation mapping (CAM) is proposed to obtain the pseudo labels, and only concentrates on the most discriminative parts. Recently, transformer-based methods utilize attention map from the multi-headed self-attention (MHSA) module to predict pseudo labels, which usually contain obvious background noise and incoherent object area. To solve the above problems, we use the Conformer as our backbone, which is a parallel network based on convolutional neural network (CNN) and Transformer. The two branches generate pseudo labels and refine them independently, and can effectively combine the advantages of CNN and Transformer. However, the parallel structure is not close enough in the information communication. Thus, parallel structure can result in poor details about pseudo labels, and the background noise still exists. To alleviate this problem, we propose enhancing convolution CAM (ECCAM) model, which have three improved modules based on enhancing convolution, including deeper stem (DStem), convolutional feed-forward network (CFFN) and feature coupling unit with convolution (FCUConv). The ECCAM could make Conformer have tighter interaction between CNN and Transformer branches. After experimental verification, the improved modules we propose can help the network perceive more local information from images, making the final segmentation results more refined. Compared with similar architecture, our modules greatly improve the semantic segmentation performance and achieve 70.2% mean intersection over union（mIoU） on the PASCAL VOC 2012 dataset.

In this paper, the joint design of transmit and receive beamformers for transmit subaperturing multiple-input-multiple-output (TS-MIMO) radar is investigated, aiming to enhance its low probability of intercept (LPI) capability. The main objective is to simultaneously minimize the transmission power, suppress the transmit sidelobe levels, and minimize the probability of intercept, thus bolstering the LPI performance of the radar system while maintaining the desired target detection performance. An alternative optimization method is proposed to jointly optimize the transmit and receive beamformers, yielding an unified LPI optimization framework. Particularly, the proposed iterative algorithm based on the Lagrange duality theory for transmit beamforming is more efficient than the conventional convex optimization method. Numerical experiments highlight the effectiveness of the proposed approach in sidelobe suppression and computational efficiency.

The theoretical implementation aspects of scattered field prediction and angular glint calculation in near-field region are proposed in this work. First of all, a more refined expression of the Green function is developed. In this representation, an expansion center is adopted within the neighborhood of the sources. Then a high-frequency electromagnetic scattering evaluation algorithm is formulated, combining the refined physical optics (PO) and equivalent edge current (EEC) algorithm. The modified method not only retains the conciseness and efficiency of the standard code but also can be directly used in the near field (NF) scattering estimation. Afterwards, two basic concepts of the angular glint are briefly introduced and formulated. The proposed procedure makes preparation for the computation of NF linear deviation. Numerical examples demonstrate the accuracy and efficiency of the NF scattering prediction algorithm. The angular glint characteristics in near-field scenarios are also presented and analyzed in the final section.

In wireless sensor networks, ensuring communication security via specific emitter identification (SEI) is crucial. However, existing SEI methods are limited to closed-set scenarios and lack the ability to detect unknown devices and perform class-incremental training. This study proposes a class-incremental open-set SEI approach. The open-set SEI model calculates radio-frequency fingerprints (RFFs) prototypes for known signals and employs a self-attention mechanism to enhance their discriminability. Detection thresholds are set through Gaussian fitting for each class. For class-incremental learning, the algorithm freezes the parameters of the previously trained model to initialize the new model. It designs specific losses: the RFFs extraction distribution difference loss and the prototype transformation distribution difference loss, which force the new model to retain old knowledge while learning new knowledge. The training loss enables learning of new class RFFs. Experimental results demonstrate that the open-set SEI model achieves state-of-the-art performance and strong noise robustness. Moreover, the class-incremental learning algorithm effectively enables the model to retain old device RFFs knowledge, acquire new device RFFs knowledge, and detect unknown devices simultaneously.

This paper addresses weak target detection problem for bistatic radar via a pseudo-spectrum (PS) based track-before- detect (TBD). Generally, PS-TBD estimates target position and velocity by means of pseudo-spectrum construction in the discrete measurement space and accurate energy accumulation in mixed coordinates. However, the grids within the polar sensing region of the receivers in the bistatic radar are not aligned. Traditional PS-TBD can not directly process these measurements. In this paper, a PS-TBD method for bistatic radar is proposed to overcome this problem. Each cell in the measurement space of the receivers is mapped to the aligned Cartesian coordinates and predicted to the integration frame according to the assumed filter velocity. A PS is formulated centered on the predicted Cartesian position. Then the samples of the pseudo-spectra are accumulated to the nearest cell around the predicted Cartesian position. The procedure of the energy integration is derived in detail. Simulation results validate the efficacy of the proposed method in terms of detection accuracy and parameter estimation.

Despite the superior advantages of specific emitter identification in extracting emitter features from in-phase and quadrature (I/Q) signals, challenges persist due to signal-type confusion and background noise interference. To address those limitations, this paper proposes a multi-channel contrast prediction coding and complex-valued residuals network (MCPC-MCVResNet) framework. This model employs contrast prediction techniques to directly extract discriminative features from electromagnetic signal sequences, effectively capturing both amplitude and phase information within I/Q data. A core innovation of this approach is the sphere space softmax (SS-softmax) loss, which optimizes intra-class clustering density of while establishing well-defined boundaries between distinct emitters. The SS-softmax mechanism significantly enhances the model’s capacity to discern subtle variations among radiation emitters. Experimental results demonstrate superior identification accuracy, rapid convergence, and exceptional robustness in low signal-to-noise ratio environments.

To address the issue of incorrect fusion results caused by conflicting evidence due to inaccurate evidence and incomplete recognition frameworks in radar airborne target tactical intention recognition, a spatiotemporal evidence fusion algorithm is proposed. To resolve the conflict evidence fusion problem caused by inaccurate evidence, the algorithm performs discounting of evidence from both spatial and temporal dimensions. Spatial discounting is influenced by both inter-evidence inconsistency and intra-evidence inconsistency, while temporal discounting is determined by time intervals and information entropy. For the problem of conflicting evidence fusion due to an incomplete recognition framework, an open recognition architecture based on dynamic composite focal elements is proposed. This approach allocates some conflicting information to temporary composite focal elements, avoiding excessive basic probability assignment (BPA) of the empty set after fusion, which can lead to deviations from the actual fusion results. Simulation experiments comparing various methods indicate that the proposed method can effectively improve target intention recognition accuracy and demonstrates good stability.

Frequency diverse array multiple-input multiple-output (FDA-MIMO) radar has gained considerable research attention due to its ability to effectively counter active repeater deception jamming in complex electromagnetic environments. The effectiveness of interference suppression by FDA-MIMO is limited by the inherent range-angle coupling issue in the FDA beampattern. Existing literature primarily focuses on control methods for FDA-MIMO radar beam direction under the assumption of static beampatterns, with insufficient exploration of techniques for managing nonstationary beam directions. To address this gap, this paper initially introduces the FDA-MIMO signal model and the calculation formula for the FDA-MIMO array output using the minimum variance distortionless response (MVDR) beamformer. Building on this, the problem of determining the optimal frequency offset for the FDA is rephrased as a convex optimization problem, which is then resolved using the cuckoo search (CS) algorithm. Simulations confirm the effectiveness of the proposed approach, showing that the frequency offsets obtained through the CS algorithm can create a dot-shaped beam direction at the target location while effectively suppressing interference signals within the mainlobe.

Multichannel signals have the characteristics of information diversity and information consistency. To better explore and utilize the affinity relationship within multichannel signals, a new graph learning technique based on low rank tensor approximation is proposed for multichannel monitoring signal processing and utilization. Firstly, the affinity relationship of multichannel signals can be acquired based on the clustering results of each channel signal. Wherein an affinity tensor is constructed to integrate the diverse and consistent information of the clustering information among multichannel signals. Secondly, a low-rank tensor optimization model is built and the joint affinity matrix is optimized with the assistance of the strong confidence affinity matrix. Through solving the optimization model, the fused affinity relationship graph of multichannel signals can be obtained. Finally, the multichannel fused clustering results can be acquired though the updated joint affinity relationship graph. The multichannel signal utilization examples in health state assessment with public datasets and microwave detection with actual echoes verify the advantages and effectiveness of the proposed method.

Evaluating the adversarial robustness of classification algorithms in machine learning is a crucial domain. However, current methods lack measurable and interpretable metrics. To address this issue, this paper introduces a visual evaluation index named confidence centroid skewing quadrilateral, which is based on a classification confidence-based confusion matrix, offering a quantitative and visual comparison of the adversarial robustness among different classification algorithms, and enhances intuitiveness and interpretability of attack impacts. We first conduct a validity test and sensitive analysis of the method. Then, prove its effectiveness through the experiments of five classification algorithms including artificial neural network (ANN), logistic regression (LR), support vector machine (SVM), convolutional neural network (CNN) and transformer against three adversarial attacks such as fast gradient sign method (FGSM), DeepFool, and projected gradient descent (PGD) attack.

Prelaunch rolling of maritime rockets threatens the reliability of launch in rough sea conditions. In order to suppress the prelaunch rolling, this study introduces advanced smart prediction designed especially for maritime rockets. The suggested approach introduces a hybrid model that combines random forest (RF) and Adaptive boosting （AdaBoost） methods to describe the coupling mechanism of factors affecting rocket rolling and to suppress the rolling. This combination improves forecast accuracy. Thereafter, the dimensionality reduced response surfaces are used to visually present the coupling between rocket rolling and influencing factors, which reveals the prelaunch rolling mechanism. When angle between the launch device and the ship’s bow is within 80°?100°, the dynamic friction coefficient between adapters and guideways is 0.4, and the dynamic friction coefficient between the rocket and launchpad is within 0?0.15 or 0.5?0.7, the prelaunch rolling of rocket during one motion cycle of the ship is less than 0.065°, originally 0.27°, reduced by 75.93%, effectively suppressing the prelaunch rolling. This study improves the prelaunch stability of maritime rockets in rough sea conditions and establishes a mapping relationship between the factors affecting rocket rolling and the structure of the sea launch system, guiding the optimization of future sea launch systems.

Cooperative pursuit poses challenges across natural, social, and technical systems, particularly when decentralized, slow-speed pursuers attempt to capture a high-speed evader with limited observation. Most existing contributions place the focus on the greedy pursuit of the evader, overlooking potential collaborations among pursuers. To tackle this issue, a decision-making framework of multi-agent coordinated reciprocity formation pursuit (MACRFP) via deep reinforcement learning is introduced. This framework integrates the actor-critic algorithm with the coordinated reciprocity mechanism to enhance the capability of capturing a faster evader. Initially, a local perception model is created by utilizing a cellular network to simulate limitations caused by obstacles. Next, the formation coalition of pursuit is guided by the Cartesian Oval, enabling dispersed pursuers to create a siege against the faster evader. Furthermore, a coordinated reciprocity model based on the coordination graph and the attention-based graph neural networks is developed, addressing the global coordination problem by estimating a reciprocity coefficient to adjust agents’ rewards. Numerical simulations demonstrate the emergence of cooperative behaviors in cooperative besiegement, target tracking, and intelligent interception during the pursuit, indicating that the proposed algorithm enhances the feasibility and effectiveness of capturing a fast-escaping target by integrating coordinated reciprocity and coalition formation.

For mission-oriented unmanned aerial vehicle (UAV) swarms, mission capability assessment provides an important reference in the design and development process, and is a precondition for mission success. For this multi-criteria decision-making (MCDM) problem, the current literature lacks a way to unambiguously present criteria and the popular fuzzy analytic network process (ANP) approaches neglect the hesitancy of subjective judgments. To fill these research gaps, an MCDM method based on unified architecture framework (UAF) and interval-valued spherical fuzzy ANP (IVSF-ANP) is proposed in this paper. Firstly, selected viewpoints in UAF are extended to construct criteria models with standardized representation. Secondly, interval-valued spherical fuzzy sets are introduced to ANP to weight interdependent criteria, handling fuzziness and hesitancy in pairwise comparisons. A method of adjusting weights of experts based on their decision similarities is also included in this process to reduce ambiguity brought by multiple experts. Next, performance characteristics are non-linearly transformed regarding to expectations to get final results. This proposition is applied to assess the mission capability of UAV swarms to search and strike surface vessels. Comparative analysis shows that the proposed method is valid and reasonable.

A performance improvement model of research and development (R&D) institutions based on evolutionary game and Bayesian network is proposed. First, the nature and performance factors of new R&D institutions are systematically analyzed, the appropriate factor model is found, and the sharing of performance benefits between institutions and employees, the change in distribution proportion, and the risk of institutional improvement and employee cooperation are considered. Second, based on the mechanism improvement and employee cooperation, the payment matrix is given and evolutionary game analysis is carried out to obtain a stable and balanced institutional improvement probability and employee cooperation probability. These two probability values are substituted into the Bayesian network model of performance improvement of new R&D institutions, and the posterior probability of performance improvement is predicted by Bayesian network reasoning and diagnosis to find effective improvement measures. Finally, practical case analysis is given to verify the effectiveness and practicability of the proposed method.

Efficient multiple unmanned aerial vehicles (UAVs) path planning is crucial for improving mission completion efficiency in UAV operations. However, during the actual flight of UAVs, the flight time between nodes is always influenced by external factors, making the original path planning solution ineffective. In this paper, the multi-depot multi-UAV path planning problem with uncertain flight time is modeled as a robust optimization model with a budget uncertainty set. Then, the robust optimization model is transformed into a mixed integer linear programming model by the strong duality theorem, which makes the problem easy to solve. To effectively solve large-scale instances, a simulated annealing algorithm with a robust feasibility check (SA-RFC) is developed. The numerical experiment shows that the SA-RFC can find high-quality solutions within a few seconds. Moreover, the effect of the task location distribution, depot counts, and variations in robustness parameters on the robust optimization solution is analyzed by using Monte Carlo experiments. The results demonstrate that the proposed robust model can effectively reduce the risk of the UAV failing to return to the depot without significantly compromising the profit.

The proliferation of carrier aircraft and the integration of unmanned aerial vehicles (UAVs) on aircraft carriers present new challenges to the automation of launch and recovery operations. This paper investigates a collaborative scheduling problem inherent to the operational processes of carrier aircraft, where launch and recovery tasks are conducted concurrently on the flight deck. The objective is to minimize the cumulative weighted waiting time in the air for recovering aircraft and the cumulative weighted delay time for launching aircraft. To tackle this challenge, a multiple population self-adaptive differential evolution (MPSADE) algorithm is proposed. This method features a self-adaptive parameter updating mechanism that is contingent upon population diversity, an asynchronous updating scheme, an individual migration operator, and a global crossover mechanism. Additionally, comprehensive experiments are conducted to validate the effectiveness of the proposed model and algorithm. Ultimately, a comparative analysis with existing operation modes confirms the enhanced efficiency of the collaborative operation mode.

Manned aerial vehicle-unmanned aerial vehicle (MAV-UAV) combat organization is a MAV-UAV combat collective formed from the perspective of organization design theory and methodology, and the generation of force formation plan is a key step in the organizational planning. Based on the description of the problem and the definition of organizational elements, the matching model of platform-target attack wave is constructed to minimize the redundancy of command and decision-making capability, resource capability and the number of platforms used. Based on the non-dominated sorting genetic algorithm III (NSGA-III) framework, which includes encoding/decoding method and constraint handling method, the generation model of organizational force formation plan is solved, and the effectiveness and superiority of the algorithm are verified by simulation experiments.

The comprehensive evaluation of six properties for equipment product is an important basis for their quality control, and their correlative relationship among six properties will affect their quality level. To understand their correlative relationship among six properties, this paper firstly combines group evaluation with decision-making trial and evaluation laboratory (DEMATEL) model, and develops the optimization model based on group consensus to form six influent relationship matrices. Secondly, group consensus matrix is used to design super network hierarchy matrix, and the weights of six properties with relevant environment is also proposed. Thirdly, the elimination and choice translating reality (ELECTRE) model is used to make comprehensive evaluation, and an example is used to compare the results under two kinds of conditions, and illustrate the effect of the weights of six properties on the priority of equipment products.

In strategic decision-making tasks, determining how to assign limited costly resource towards the defender and the attacker is a central problem. However, it is hard for pre-allocated resource assignment to adapt to dynamic fighting scenarios, and exists situations where the scenario and rule of the Colonel Blotto (CB) game are too restrictive in real world. To address these issues, a support stage is added as supplementary for pre-allocated results, in which a novel two-stage competitive resource assignment problem is formulated based on CB game and stochastic Lanchester equation (SLE). Further, the force attrition in these two stages is formulated as a stochastic progress to consider the complex fighting progress, including the case that the player with fewer resources defeats the player with more resources and wins the battlefield. For solving this two-stage resource assignment problem, nested solving and no-regret learning are proposed to search the optimal resource assignment strategies. Numerical experiments are taken to analyze the effectiveness of the proposed model and study the assignment strategies in various cases.