Heterodyne laser interferometer has been widely used for displacement and velocity measurements in the fields of vibration calibration and acoustic metrology. When the displacement or velocity to be measured exceeds a certain limit, the phase of the heterodyne interference signal, corresponding to different displacement or velocity amplitudes, is wrapped within the range of (−π, +π], and various phase unwrapping methods have been proposed to obtain the corresponding true phase. However, these methods are computationally intensive and require large memory and expensive hardware resources, thus hindering the development of real-time, embedded solutions for wide bandwidth measurement applications. In this study, a one-dimensional dual-counters-based phase unwrapping (1D DCPU) algorithm is proposed to achieve real-time phase unwrapping while utilizing minimal resources, which can be easily implemented on a field programmable gate array (FPGA) in real-time. The performance of this 1D DCPU algorithm is simultaneously demonstrated by the simulated and experimental results on the FPGA in the frequency range of 10−500 kHz. The proposed 1D DCPU algorithm offers the advantages of small computation and high-efficiency operation on FPGA, which is especially suitable for the large bandwidth, embedded and real-time sensing applications.

Antenna arrays have become a popular choice for protecting global navigation satellite system (GNSS) receivers in strong radio frequency interference environments. Environmental nonstationarities, which are caused by fast variations in propagation channels and rapid motion of interfering sources or antenna arrays, cause performance degradation of antenna array receivers. The covariance matrix taper (CMT) has been used to widen the adaptive pattern nulls toward the interference sources in GNSS antenna array receivers to address highly dynamic environments. In this paper, it is verified that the loss of the signal to interference and noise ratio (SINR) caused by a CMT cannot be ignored, and a CMT with a fixed null width cannot work well when the dynamic range of the GNSS array receiver is large. To design the modest null width, a CMT with an adaptive null width for space-time adaptive processing (STAP) is introduced. In the adaptive CMT, the proposed subspace leakage in STAP is used to estimate the required null width. In the simulations, the SINRs of adaptive CMT are always close to the SINRs of the optimal fixed null width in different scenarios, and the average SINRs of the adaptive CMT are even more than 10 dB higher than that of CMT with fixed null width in some scenarios. The adaptive CMT can be widely applied to GNSS antenna array receivers with a large dynamic range.

This paper highlights a Tibetan syllable spelling check method based on memristive computing. Firstly, a rule-based syllable component recognization method is implemented to classify components of syllables into seven types, that is prefix letter, root letter, superscript letter, subscript letter, vowel letter, suffix letter, and farther suffix letter. In order to detect syllable formation errors in Tibetan text, a Tibetan spelling check method is presented based on component recognization by adding mapping principles. Then, a memristor crossbar array with an external control circuit is introduced to make memristive computing for Tibetan spelling check, by analyzing the total current value in a memristor crossbar array to implement rule-based matching and mapping processes. Finally, a stable memristive computing algorithm-based Tibetan spelling check is established. A simulation experiment on the memristor crossbar array demonstrates the stability of our memristive computing-based Tibetan spelling check method. Comparative experiments on 1274 of the most common Tibetan syllables show that the accuracy of our component recognization method is 97.19%, the F1-score of spelling check is 98.58%, beating other rule-based and statistics-based methods in precision and F1-score, demonstrating the accuracy of our method.

In target tracking stage, the integrated sensing and communication (ISAC) system needs to estimate various target parameters, such as the angle and range, while ensuring highly reliable communication. This paper proposes a two-stage constant-modulus waveform design. In the transmission waveform covariance matrix (TWCM) optimization stage, we minimize the Cramer-Rao bound (CRB) of joint estimation for two-dimensional direction of arrival (DOA) and complex reflection coefficients under total power constraint, obtaining an optimal TWCM. In the alternating optimization based on Riemannian steepest descent (RSD) stage, under constant modulus property and multi-user interference constraints, minimize the squared Euclidean distance between actual TWCM and optimal TWCM obtained from the TWCM optimization stage. To address the optimization problem, an alternate optimization algorithm based on the RSD method is proposed. Simulation results demonstrate that the proposed waveform design simultaneously achieve reliable communication and multi-target tracking.

A low-profile wideband multi-mode circularly polarized (CP) multiple-input-multiple-output (MIMO) antenna with the parasitic metasurface using characteristic mode analysis (CMA) is presented. The single monopole element of MIMO is composed of a transmission line and a branch after slotting on the ground. Structurally, the transmission line forms a 90° shape with the branch, providing favorable conditions for achieving circular polarization. Through CMA, the difference in characteristic angle (CA) at the intersection of the mode significance (MS) is about 90° with their almost consistent radiation directions. The analysis results show that four modes work together to achieve a circular polarization effect. Secondly, after rotating the monopole antenna at three angles (90°, 180°, 270°), a 4-port MIMO antenna is formed, and the results of their ports on the same level are consistent. Finally, the narrow axial ratio bandwidth (ARBW) of MIMO antennas is compensated by adding two rows of parasitic cross-circular metasurfaces. The measured results of the proposed antenna prototyped are in agreement with the simulated ones, demonstrating an impedance bandwidth (IBW) of 84.1% (3.65-8.95 GHz), a 3-dB ARBW of 52.0% (3.7-6.3 GHz), and a peak gain of 5.8 dBic at 4.9 GHz. The antenna achieves a total efficiency ranging from 72.8% to 98.8% across the bandwidth, and its envelope correlation coefficient (ECC) remains below 0.15 throughout the working frequency range.

In order to study the influence of the support device on the energy conversion of the electromagnetic coil launcher, five different material and structure Livermore Software Technology Corporation’s Dynamic Analyzer (LS-DYNA) finite element analysis models are established, including the insulation ring column structure model, non-magnetic ring column structure model, magnetic ring column structure model, non-magnetic half ring column structure model, and magnetic half ring column structure model, to investigate the effect of the support device on the energy conversion of the internal drive coil launcher. In order to obtain the circuit characteristics of the internal drive coil launcher, a mathematical model and Maxwell finite element analysis model of the internal drive coil launcher are established, and the circuit characteristics of the launcher are solved based on the established model. By inputting system circuit data into a supported LS-DYNA finite element analysis model, the internal driving three-dimensional non axisymmetric simulation is completed, which solves the problem of Maxwell simulation software being unable to achieve three-dimensional hollow motion domain. The simulation results show that the peak velocities of the five supported launchers are reduced by 2.5%, 61.9%, 42.1%, 29.9%, and 14.7% respectively compared to the unsupported coil launchers. The results show that the influence of insulation materials on the energy conversion of the launcher can be basically ignored, while the influence of metal materials on the energy conversion of the launcher is significant. The use of magnetic high resistance materials and anti eddy current structures can significantly weaken the influence of metal support devices on the energy conversion of the coil launcher, and improve the launch performance of the coil launcher.

To address the issue of insufficient radar detection capability caused by the low signal-to-noise ratio of target echoes, a coherent synthesis method for distributed aperture rotating array radar (DARAR) based on two-stage compensation is proposed. Firstly, a parallel control model is employed to uniformly manage the servo subsystems, ensuring consistent rotation across multiple radars. Next, by utilizing prior information such as the distributed system layout and beam pointing, the echo signal undergoes a first-level coarse compensation. Finally, a regional cross-correlation (RCC) method estimates the time and phase parameters within a small-scale range, completing the second-level fine compensation of time and phase and achieving coherent synthesis of multiple rotating array radar echo signals. A system comprising a three-node DARAR is designed, and joint detection experiments targeting unmanned aerial vehicles are conducted. Experimental results demonstrate that the efficiency of coherent synthesis increases by more than 20% under signal-to-noise ratios ranging from 8 dB to 13 dB. The proposed method enables DARAR operation and enhances the detection capability of rotating array radar.

Unsupervised methods for defect detection hold significant potential for automated optical inspection (AOI) due to their independence from the challenges of collecting defect data. However, their performance is often limited in scenarios where normal samples exhibit high variability and defect features are subtle. To address this limitation, an unsupervised defect detection method named feature isolation autoencoder (FIAE) is proposed. This method compresses images into two components: one that encodes the information necessary to reconstruct normal samples, and the other that encodes anomalous information. As a result, any non-zero value in the anomalous component indicates the presence of an anomaly in the input images. Additionally, we propose the dynamic defect variability guidance (DDVG) technique and the stabilization-compensation training strategy, which ensure that FIAE achieves the expected training outcomes in the training data and process levels, respectively. Extensive experiments on the transistor dataset compiled in this study and the benchmark dataset MVTec Anomaly Detection dataset demonstrate that FIAE outperforms the baseline methods in most cases, achieving an area under the curve (AUC) of 0.979 and a best F1 score of 0.958 on the transistor dataset. The necessity of the proposed DDVG and the stabilization-compensation training strategy through ablation studies are validated.

In electronic warfare, radar systems are confronted with significant threats within the complex electromagnetic environment, particularly from the prevalent method of blanket jamming that impairs radar’s detection capabilities. Frequency agility is recognized as an effective anti-jamming method against blanket jamming in practice. However, the traditional frequency agility selects the carrier frequency randomly, which highly limits the potential for robust anti-jamming capabilities, despite offering sufficient optimization space to improve the radar performance. This paper proposes an antijamming decision strategy based on the comb gradient-bandits algorithm (CGBA) to optimize decisions for frequency agility radar under blanket jamming conditions. In this strategy, the reward value is defined as the signal to interference plus noise ratio (SINR) of the radar in each training round, and the radar selects the frequency channel with the highest reward value as the new operational channel. The proposed method has dynamic adaptive ability according to the jamming status, which enables the radar to update its frequency channel in real-time. Theoretical analysis and experimental results show that the proposed model improves the radar’s ability to select anti-jamming strategies to a certain extent.

When detecting weak targets, radar systems often encounter the interference from complex clutter in the field scene, leading to a high false alarm rate and a low detection rate when the traditional radar methods are employed. To address this issue, this paper proposes a clutter suppression method based on generative adversarial networks (GANs) to improve the detection performance of weak targets. This method designs a clutter identification network based on the characteristics of clutter in the range-Doppler domain. By adopting a divide-and-conquer strategy, a clutter suppression network is selected based on the clutter identification results. This method transforms the clutter suppression process into a mapping problem from the clutter-affected data domain to a clutter-free data domain. Experiments results show that the proposed method achieves a 10% improvement in detection rate in rain clutter scenarios, and reduce the false alarm rate by 90% in ground clutter scenarios.

Compressive sensing (CS) has been successfully combined with deep-learning neural networks to generate models for image-depth CS. Many existing models face significant limitations, such as their inability to handle tasks with multiple CS ratios and suboptimal performance in image reconstruction. To overcome these challenges, an optimized deep network model, named global multi-attention condition module (GMCM) iterative shrinkage-thresholding algorithm (ISTA)-Net⁺⁺ is proposed. In this model, the CS ratio of an image is incorporated as a network parameter, enabling the network to effectively handle multiple CS ratio tasks. Additionally, a hybrid attention mechanism is employed to preprocess the reconstructed image for feature extraction and to calculate the image complexity score. These elements are then combined with the CS ratio to jointly predict the parameters and iterative reconstruction steps, which are subsequently broadcast to all iterative stages of the model. Extensive experiments on high- and low-resolution image datasets demonstrate that the proposed GMCM ISTA-Net⁺⁺ model achieves a superior performance compared to traditional models. Given its flexibility, enhanced reconstruction quality, and adaptability, this model holds significant potential for deployment in resource-constrained mobile image-processing terminals.

Traditional optimization algorithms from operations research are not well-suited for addressing the vehicle routing problem (VRP) with a large number of instances and high real-time requirements. Deep reinforcement learning (DRL) introduces a new approach to solving VRP, characterized by its rapid problem-solving capabilities and robust generalization abilities. In the domain of VRP variants, numerous benchmarks are utilized to evaluate the performance of the solving algorithms. However, the limited number of instances in these public benchmarks is insufficient for training DRL algorithms. To address this issue, an effective benchmark augmentation method using distribution derivation is proposed. By modeling the distributions of key VRP variables and estimating their parameters, the method derives distributions closely resembling the original benchmark and subsequently generates a large number of synthetic instances accordingly. The proposed method holds promise for widespread applications across various benchmarks. Experiments on VRP with time windows (VRPTW) and heterogeneous fleet VRP (HFVRP) verify the generalization abilities. The results show that DRL trained using the augmented benchmarks performs much better on public benchmarks than trained using the data customized by the algorithm authors. In addition, the DRL algorithms trained using the augmented benchmarks achieve performance comparable to state-of-the-art algorithms in public benchmarks for various VRP variants while significantly reducing the computation time.

It is a complex system problem to dispatch the robot swarm to complete exploration missions with high quality in uncertain environments. To solve the challenge, an algorithm of the robot swarm exploration mission planning supported by humans is proposed in this paper. Firstly, the target environment is decomposed into several areas, where it takes a price to explore these areas and each robot gets a reward after exploring. Secondly, the exploration problem is modeled as a multi-agent dynamic programming problem with the help of humans. Then, an extensible greedy based planning algorithm is proposed, where the searching policy of each robot is calculated based on the judgment index. Finally, the proposed algorithm is optimal through theoretical analysis, and simulation experiments show that the algorithm can provide high-quality solutions for the swarm exploration.

Supportability simulation analysis has emerged as a cost-effective alternative to physical experimentation for enhancing the logistics capacity of equipment support, especially in complex combat scenarios. However, key issues such as support scheme planning, resource optimization, and comprehensive analysis of the reliability, maintainability, supportability, and testability (RMST) indices are seldom explored in depth. This paper proposes mission-driven methods for equipment supportability analysis and optimization. Firstly, a supportability modeling and optimization framework, along with a mission-driven support activity simulation, is presented to describe the complex interaction between large-scale equipment and resources. Then, an optimization model for resource allocation based on simulation is developed, and a heuristic method is introduced to solve the problem. Fault uncertainty is incorporated into the support activities to integrate the RMST indices of the equipment. A Monte Carlo simulation is then used to comprehensively evaluate these RMST indices. Finally, a numerical example is used to validate the proposed methodology. The experimental results demonstrate that the presented approach successfully performs simulation, analysis, and optimization of the support activity, and the equipment deployment results in 93.29% availability, which can provide reliable decision support for equipment operation and logistics.

A reliable evaluation index system (EIS) is essential for assessing the operational capabilities of equipment system-of-systems (SoS), yet systematic methods for validating its effectiveness in complex scenarios remain lacking. To address this problem, this paper proposes a validation algorithm, which includes a validation framework, a data collection format tailored to operational contexts, and a grey clustering algorithm. First, the validation framework is designed at three levels, syntactic, semantic, and pragmatic, to systematically assess the normativity, logical coherence, and practical utility of the EIS. Second, a data collection format suited to operational backgrounds is introduced, enabling multi-domain coverage across the physical, informational, cognitive, and social dimensions, and accounting for both expected and abnormal scenarios. Third, a grey clustering algorithm is developed, in which a clustering weight calculation model based on the interval fuzzy preference relation matrix ensures consistency among indicators, while rough and fine clustering analyses are combined to achieve quantitative validation of EIS effectiveness. Case studies demonstrate that the proposed method effectively validates EIS performance under diverse operational conditions. Overall, this work provides a systematic basis for EIS validation and practical guidance for enhancing SoS evaluation.

To address issues such as model discretization, incomplete coverage of verification elements, and broken requirement traceability in the verification of combat system-of-systems (SoS) architectures, the paper proposes a verification method for combat systems based on model-based system engineering-multi-granularity system simulation verification, architecture data model verification and requirements model verification. To achieve consensus on architectural design elements and avoid deviations caused by terminology logic, a meta-model data set is constructed based on the Department of Defense Architecture Framework (DoDAF) meta-model method. The model-based system engineering methodology is introduced to clarify the design and verification framework of combat SoS from requirements analysis, design, analysis to verification. To address the issue of insufficient verification, a closed-loop system verification approach is proposed from the bottom up, encompassing multi-granularity system simulation verification, architectural data model verification, and requirements verification. An unmanned aircraft system cooperative combat SoS is constructed and verified. The verification method can systematically and comprehensively verify the domain, system, SoS architecture, and requirement models, reflecting the rationality, standardization, and effectiveness of the system verification method design. The verification method provides a systematic verification framework for combat SoS. It has engineering application value for improving the design reliability of combat systems.

To address the problem of inefficient visual information processing and high cognitive load in manned/unmanned aerial vehicle (M/UAV) cooperative command and control (C2) interfaces under complex battlefield environments, this paper proposes a phased dynamic information presentation strategy tailored to different mission phases. Based on visual attentional capture theory and stimulus characteristics, we designed four dynamic presentation methods—blinking, flowing border, masking, and size scaling—and evaluated them across four typical mission phases: early warning detection, target identification, interception strike, and damage assessment. We conduct a mixed-design experiment with 4 mission phases × 5 presentation methods (including a static baseline) and collected data through eye-tracking, behavioral metrics, and subjective evaluations. Results show that dynamic presentation methods significantly outperform static ones in terms of task completion time, time to first fixation, total fixation duration, and subjective ratings. The optimal dynamic method for each phase is identified as blinking for early warning detection, masking for target identification, size scaling for interception strike, and flowing border for damage assessment. The findings demonstrate that phase-adapted dynamic presentation strategies can effectively enhance visual search efficiency and reduce cognitive load in M/UAV cooperative C2 interfaces, providing practical guidance for human-computer interaction design in high-pressure operational scenarios.

Large-scale group decision-making (LSGDM), as a unique form of group decision-making, has been extensively studied. Its characteristic lies in the large number of decision-makers (DMs), resulting in a certain degree of complexity in the decision-making process. In the social network environment, DMs typically have some form of social relationship, influencing each other’s decision information. Considering the interactive connections among DMs, this paper proposes a model for LSGDM that takes into account community opinion leaders. Firstly, experts use linguistic distributed preference relations (LDPRs) to express assessment opinions. Different from the traditional approach that assigns the same semantic information to each expert, we construct a consistency-driven optimization model to obtain personalized individual semantics (PIS) of experts and language preferences are transformed into fuzzy numerical preferences. Secondly, combining social relationships and preference information among experts, the classic community detection algorithm is improved to partition large-scale groups into communities. Additionally, a feedback mechanism is designed to identify community opinion leaders and provide preference adjustment suggestions for experts. Finally, the feasibility and effectiveness of the proposed consensus decision-making model are verified through a specific case analysis.

The semiconductor chip industry has become a bottleneck in China’s industrial development. How to improve efficiency has attracted increasing attention. Effective and efficient scheduling decision is the key to improve production efficiency. Oriented towards the wafer manufacturing of micro-electro-mechanical system (MEMS), a reentrant flexible job-shop scheduling problem with assembly constraints (RFJSSP-AC) is modeled to minimize the makespan and the total energy consumption simultaneously. An improved multi-objective evolutionary algorithm based on decomposition (IMOEA/D) is developed to solve the problem. The proposed IMOEA/D incorporates three key innovations: a hybrid initialization strategy combining heuristic rules with random generation to produce high-quality initial populations; an objective- oriented search strategy for targeted solution optimization; and an energy consumption inverse optimization mechanism for further energy efficiency improvement. Comprehensive comparative experiments demonstrate the effectiveness of the proposed algorithmic components and the exceptional performance for the RFJSSP-AC problem.

In the study of complex networks, the identification of key nodes is crucial but challenging. Traditional centrality indicators (such as degree, betweenness, and closeness centrality) are difficult to achieve a balance between accuracy, effectiveness and time complexity, time-consuming and resource-consuming, or unable to accurately locate key nodes. In order to solve these problems, based on the commonly used Laplacian centrality (LC), a two-hop LC (TLC) index is proposed, which aims to improve the index recognition performance by introducing second-order neighbor node information. The two-hop degree centrality (TDC) is proposed and combined with LC to derive the TLC, whose superiority is verified through multi-dimensional comparative analysis. Meanwhile, the application of higher-order hop degree centrality in the derivation of LC is deduced, and it is proved that it is not feasible to apply three-hop and above degree centrality to the derivation of LC.

The application of Dempster-Shafer evidence theory in multi-source information fusion faces some problems including how to determine the weight of evidence in reliability multi-source information fusion, and how to solve the problem of conflict in Dempster-Shafer evidence fusion. In this paper, Jeffreys-Matusita distance function is introduced to establish the reliability index and consistency index, realize the difference measurement between evidence, determine the reliability degree of evidence, and evaluate the consistency difference of evidence. The distance measurement of multi-source information is achieved with the reassigned weight to avoid conflicts at the time of evidence fusion. In the end, the multi-source information fusion in the development of guided munitions is taken as an example to illustrate the conversion methods for expert experience information, component and subsystem information, reliability growth information, and similar product information, respectively. With the field test information, the multi-source reliability information is fused. The applicability and effectiveness of the proposed method are verified by comparing with the existing multi-source information fusion methods.

To address the issue of local oscillations in the slime mould algorithm (SMA) during path planning within complex environments, a multi-strategy improved hierarchical guidance SMA (MHGSMA) is introduced to enhance the search performance of SMA. Based on a priori analysis of the early-stage search behavior of slime moulds, a hierarchical method is developed to achieve a more rational division between elite and general populations. While adhering to the inherent search mechanism of SMA, the guidance functions for each population are refined to expand the search scope. Additionally, a greedy selection strategy is incorporated further to enhance the search efficiency and accuracy of the SMA. Simulation experiments for path planning are conducted in four distinct obstacle environments by comparing the MHGSMA, SMA, and hierarchical guided SMA (HG-SMA) for smooth path planning. The analysis of convergence speed, global path length, and accuracy reveals that MHGSMA excels in path planning. In complex environments, it shows improvements of 58.6% and 42% in convergence speed, 9.3% and 6% in path length, and 38.5% and 63.6% in accuracy, significantly enhancing SMA’s search performance. At the same time, under the same conditions, MHGSMA also demonstrates strong competitiveness compared to particle swarm optimization and grey wolf optimizer.

The flight plan of departing from the Earth and rendezvous near the Moon is a preferred strategy for future manned lunar exploration programs. However, significant decision-making challenges arise from diverse lunar insertion schemes during the Earth-Moon transfer process, where optimal energy consumption varies substantially across different orbital regimes. To address these challenges, an intelligent decision-making model employing deep convolutional neural networks (DCNNs) is proposed to select the optimal Earth-Moon transfer scheme, subsequently combine with the lunar rendezvous orbit to formulate an integrated mission profile. Results demonstrate that the intelligent decision-making approach achieves computational efficiency improvements of 3−4 orders of magnitude compared to conventional methods while maintaining comparable accuracy, confirming its practical viability for designing orbit of departing from the Earth and rendezvous near the Moon. This study offers valuable insights for future manned lunar missions and the application of intelligent methods in orbital design.

Guidance of advanced missiles is faced with new constraints such as side-window detection. Existing guidance laws fall short of fulfilling the constraints simultaneously with acceptable computational efforts. To tackle the issue, this paper proposes a blended optimal guidance law (BOGL). The non-zero lower bound of field-of-view (FOV) is satisfied by incorporating an artificial impact angle constraint to bend the missile trajectory. To reduce the control resource consumption, sufficient instead of overlarge weight coefficient on impact angle is determined based on the optimal control theory. Moreover, a modified hyperbolic tangent function is used to formulate a blended form of guidance law, thus further bypassing the unnecessary terminal load factor. Simulation results prove that the terminal load and total control resource consumption are reduced by over 72% and 20% respectivley compared to the pure optimal guidance law.

Few studies have sought to integrate the strengths of advanced control technologies with classical control approaches to address the stringent requirements of hyper-agile satellites for both ultra-high precision and ultra-fast maneuverability. The present work addresses this issue by combining active disturbance rejection control (ADRC) and proportional-integral-derivative (PID) control, with a focus on the use of reaction flywheels as control actuators in hyper-agile satellites. The ADRC algorithm enables the rapid correction of significant control errors and avoids excessive control torque by adopting a transition process in a tracking differentiator. In contrast, the PID algorithm applies direct error feedback to achieve ultra-high steady-state precision tracking. The results of numerical simulations demonstrate that the proposed integrated control method not only achieves ultra-fast maneuverability, with an angular displacement of 25° within 10 s, but also delivers ultra-high precision steady-state tracking with an angular error of 1×10⁻⁶° and an angular velocity variability of 2×10⁻⁸ °/s without the reaction flywheel actuator control errors. However, the reaction flywheel output torque errors reduce the velocity control stability by four orders of magnitude and the attitude control accuracy by a factor of two.

To address long-range impulsive orbital pursuit-defense game scenario, characterized by long decision cycles, sparse actions, and misaligned maneuvering times, a multi-agent asynchronous hierarchical reinforcement learning (MA-AHRL) method is proposed. The method considers practical constraints including space perturbations, thrust limitations, fuel consumption limitations, and perceptual capabilities. Firstly, a relative orbital dynamics model is established, and the pursuit-defense game environment is defined, which clarifies the success criteria for the tasks of both the pursuer and the defender. Secondly, an actor-critic network structure integrating long short-term memory (LSTM) networks with an adaptive attention mechanism is designed. The LSTM extracts features from historical interaction data to compensate for the critic network’s inability to access opponent action information during training. The attention mechanism adaptively adjusts the weights of different time steps, thereby enhancing the ability to extract key state features in the game environment. Then, the action network employs a dual-layer decision structure that integrates dynamics, where the high-level policy outputs the maneuver time interval and the low-level policy outputs impulsive actions, effectively reducing training difficulty and improving training efficiency. Finally, the pursuit-defense game is modeled as a partially observable Markov decision process (POMDP), where the action space, state space, observation space, and reward function are explicitly designed. Simulation comparison demonstrates that MA-AHRL can effectively coordinate the competitive relationship between the pursuer and defender while exhibiting excellent training stability. Testing the decision-making networks, both the pursuer and the defender display strong strategic adaptability and real-time decision-making capabilities, fully validating the effectiveness and reliability of the proposed method in this paper.

This paper presents a time-varying formation tracking strategy for multiagent quadrotor operating in hostile or challenging circumstances, where the position constraints become crucial. A decentralized cooperative control scheme is proposed based on barrier Lyapunov functions in conjunction with linear sliding mode control with directed communication among the quadrotors. The consequence of using the proposed control scheme is to provide practical finite time convergence and disturbance rejection while fulfilling position constraints. The practical finite time convergence of the proposed method is shown for multi-input and multi-output (MIMO) multiagent systems via Lyapunov analysis. The capability of the proposed algorithm is investigated for six quadrotors using MATLAB/Simulink. The simulation results demonstrate the intended performance of the proposed formation tracking controller under modeling uncertainties and external disturbances.