To address the challenges of environmental uncertainty， multi-objective conflicts， and computational complexity in multi-UAV time-coordinated path planning within unknown obstacle environments， this paper proposes a hierarchical reinforcement learning method incorporating a prediction-correction mechanism. By constructing a bilevel policy architecture consisting of upper-level obstacle avoidance and lower-level correction， the complex multi-objective optimization problem is decomposed into independently solvable sub-problems. The upper layer utilizes the soft actor-critic （SAC） algorithm to train an obstacle avoidance policy， establishing an end-to-end mapping between sensor information and control commands. A trajectory prediction module based on graph attention network-Informer （GAT-Informer） is introduced between the two layers. By modeling the spatial topological relationships of sensor data and trajectory temporal dependencies， this module predicts path perturbations caused by unknown obstacles. The lower layer employs the multi-agent soft actor-critic （MASAC） algorithm to train a coordination policy model， which applies bias correction to the upper-level commands based on the predicted coordination errors. Simulation results demonstrate that， across scenarios with varying obstacle densities， the proposed method reduces the average time coordination error by 62%， improves the task success rate by 21%， and accelerates training convergence speed by over 65% compared to end-to-end reinforcement learning（E2E-RL）methods. Furthermore， ablation studies verify the effectiveness of the trajectory prediction module， which reduces the coordination error by 43% on average. In conclusion， the proposed method effectively balances the multiple constraints of safety and time coordination， demonstrating the capability to address coordination challenges arising from environmental uncertainty.
Objective Normal hanging vibration testing of rail-mounted airborne external stores,presents sig-nificant engineering challenges due to large volume and heavy mass.With the increasing of user demand,the overall envelope dimensions of the system continue to expand.As a result,the structural size and weight of the cage-style fixtures required for normal vibration tests increase sharply.The fix-ture mass accounts for a larger proportion of the entire test system,often exceeding the capacity of the test equipment and making it difficult to complete the test.The inverted hanging system was adopted,where the external stores are positioned above the launcher.It can simplify fixture geometry,reduce its mass and the capacity requirement for test equipment.However,the fundamental validity hinges on whether the dynamic environment imposed on the specimen adequately replicates the normal hang-ing test environment.Core issues involve potential nonlinear dynamic behaviors arising from assembly clearances between the store's lugs and the launcher's rail,which may be influenced differently by the reversed direction of gravity field.Therefore,this study aims to conduct a comprehensive compara-tive analysis of the random vibration responses of a rail-mounted store-launcher system under normal hanging and inverted hanging configurations.The primary goal is to rigorously evaluate the technical feasibility and fidelity of the inverted hanging method by benchmarking its results against established testing standards. Methods To achieve this objective,a detailed finite element model of a representative rail-mounted airborne external stores system was constructed.The model incorporated the stores and the launcher,with their mechanical interface characterized by defined radial and axial clearances,as illustrated in Fig.3.The contact-interaction between the store's suspension lugs and the launcher's rail surfaces(and potential contact with web structures)was modeled to capture nonlinearities.The normal contact forces were calculated using a penalty function method,and tangential friction forces were represented by a Coulomb friction model.The system's equations of motion,incorporating these time-varying con-tact forces,were solved numerically using an explicit central difference integration scheme,chosen for its effectiveness in handling dynamic contact problems.The environmental test was simulated as a random vibration input.A standard flat acceleration power spectral density(PSD)profile,typical for such qualifications,was defined(Fig.5).This frequency-domain specification was converted into a corresponding time-domain acceleration signal using an inverse Fourier transform process with rando-mized phases,ensuring the signal possessed the required statistical properties(Fig.6).This gene-rated time-history was then applied as the base input motion.A parametric study was performed by scaling this base input to three distinct levels:1×,3×,and 5×the original PSD magnitude,encom-passing both lower-level exploratory and higher-level qualification-type environments.For each input level,two full nonlinear transient dynamic analyses were performed:one with gravity oriented to simu-late normal hanging (store weight pulling down on the launcher),and another with gravity reversed to simulate inverted hanging (store weight pulling the lugs upward against the rail). Key response met-rics,including acceleration time histories and derived power spectral densities,were extracted from critical locations:the forward and aft sections of the external store and the launcher structure. Results and Discussions The simulation results provided a detailed,quantitative comparison of the system's behavior under the two mounting configurations.At the lowest(1×)input level,distinct dif-ferences were observed.The root mean square(RMS)acceleration responses were consistently higher for the inverted hanging case,with the maximum relative difference reaching approximately 18％at the store's forward section(Table 1).Furthermore,the acceleration PSD curves for both the store and launcher showed more pronounced deviations in the low-frequency region(20～100 Hz)for the 1×input case(Figs.7,10).The initial clearance between the lug and rail,combined with the different effective contact stiffnesses between the lug's upper and lower surfaces and their respective contact partners,leads to an asymmetric contact condition.Under small excitation levels,the vibration ampli-tude may be insufficient consistent bilateral contact.Consequently,the system experiences different effective stiffness and damping depending on which surface is in contact,which is directly influenced by the gravity direction.In normal hanging,gravity biases the lug against the lower rail surface;in inverted hanging,it biases the lug against the upper structure.This bias results in the observed low-frequency response differences at low excitation.The RMS response values for the two configurations converged,showing that the differences reduced to about 8％for the 3×input and within 4％for the 5×input(Table 1).The corresponding PSD curves(Figs.8～9 for the store,Figs.11～12 for the launcher)exhibited remarkable similarity in shape,peak frequencies,and peak magnitudes.The higher energy input causes larger relative displacements,ensuring that the lug interacts with both con-tact surfaces regularly during a vibration cycle,thereby"averaging out"the directional bias intro-duced by gravity.From a standards compliance perspective,the most significant finding is that across the vast majority of the frequency spectrum and for all practical input levels,the difference in the acceleration PSD between the normal hanging and inverted hanging conditions was within 3 dB.The cumulative frequency bandwidth where the deviation exceeded 3 dB threshold was calculated to be less than 5％of the frequency range.This performance meets the tolerance requirements outlined in military testing standards such as GJB 150A.The results strongly suggest that while the micro-dynamics of the lug-rail contact differ initially,the macro-response,in terms of vibrational energy dis-tribution and dominant frequencies,is effectively equivalent under representative test conditions. Conclusions This investigation substantiates the inverted hanging method as a technically sound and practical alternative for conducting random vibration tests on large,rail-mounted airborne external stores.The computational analysis demonstrates that the dynamic responses elicited under the inverted configuration are acceptably equivalent to those from the traditional normal hanging method,with discrepancies falling within accepted engineering tolerances,particularly at qualification-level excitation.The inverted hanging condition is physically realistic,corresponding to aircraft maneuver states,and preserves the essential structural load path.Its principal advantage lies in drastically reducing fixture complexity and mass,which lowers shaker force requirements,improves control sta-bility,and enhances overall test feasibility for large specimens.Thus,the inverted hanging approach represents a valid and efficient strategy to overcome the escalating challenges associated with normal vibration testing of modern,sizable external store systems.
To address the problem that existing cooperative guidance laws are mostly limited to 2D planes without fully considering input constraints， and are prone to guidance command distortion and reduced cooperative accuracy when extended to 3D space， this paper proposes a 3D spatiotemporal cooperative guidance law with line-of-sight （LOS） angle constraints. Firstly， the relative motion equations between missiles and the target in three-dimensional space are established， and a complete cooperative guidance model is constructed by incorporating terminal LOS angle constraints. On this basis， the cooperative guidance law is designed in a dimension-wise manner. In the LOS angle direction， the range-velocity errors among missiles are converted into the desired velocities of each missile to ensure simultaneous engagement of the target by multiple missiles. In the LOS angle normal direction， the hyperbolic tangent function is introduced to constrain the control input， and based on the homogeneity theory， it is rigorously proven that the LOS angle can converge to the desired angle within finite time. This guidance strategy effectively balances temporal cooperation and angular coope-ration， enabling missiles to simultaneously meet the requirements of engagement timing and attitude accuracy， and an extended state observer is constructed to compensate for target maneuvering disturbances. Finally， simulation experiments are conducted to verify the effectiveness and feasibility of the proposed cooperative guidance algorithm. The simulation results show that the miss distance of each missile is less than 0.1 m， the LOS angle error is less than 0.3°， and the total saturation time of the three missiles is reduced by 32.13% compared with the existing schemes. This provides a reliable technical solution for the multi-missile cooperative interception mission in three-dimensional space.
For the unclear combustion heat release characteristics and flame stability mechanisms of the combustion chamber of air heaters under different structural parameters， high-speed photography and mid-wave infrared thermal imagers are used to simultaneously collect and analyze the spray combustion of the air heater. Window ignition tests are conducted on three-component direct-scale air heaters with oxygen/alcohol/air as the fuel. The influence of structural parameter changes on the spray combustion process of the air heater is explored， and the instantaneous spray combustion flame structures and heat release characteristics at key positions of the combustion chamber under 5 sets of opera-ting conditions are obtained. The results show that the variation of air injection area has the greatest impact on the combustion of the air heater， followed by the alcohol injection area， and the oxygen injection area has the least influence. The main impact is through the shear effect between gas and li-quid， which affects the mixing efficiency and thereby influences the combustion heat release and combustion efficiency. At the same time， the combustion heat release position in the combustion chamber has a crucial impact on the stable maintenance of the flame.
An integrated control scheme， based on backstepping adaptive nonsingular fast terminal sliding mode， is proposed for trajectory tracking control of quadrotor UAVs under varying load mass and center of gravity offset conditions. Considering the unbalanced load of quadrotor UAVs， a dynamic model is established. For position control， an adaptive law is designed to estimate the load mass and unknown disturbances， and a backstepping method is used to design the position controller to compute the required total lift and desired attitude angles. For attitude control， a RBF neural network is used to approximate the disturbance torque generated by the load， and a nonsingular fast terminal sliding mode controller is designed to ensure the stable and finite-time convergence of attitude angles. The global asymptotic stability of the closed-loop system is proved via Lyapunov stability analysis. The numerical simulation and real-time flight experiment results show that the proposed algorithm exhibits good tracking performance and robustness.
Common demand forecast methods for aviation ammunition universally need a large amount of data,the considered factors of these methods are not comprehensive and the relationship to combat missions are not close.Aiming at the complex problem of the demand forecast for aviation ammunition in the process of strike operations planning,this paper proposes a scientific and practical demand forecast method for aviation ammunition.According to relevant factors influencing aviation ammunition demand during wartime,it designs the modeling process of demand forecast,and based on target count,damaged rate of carrier-based aircraft in battle and ammunition damage effectiveness,it constructs the demand prediction model for aviation ammunition under the condition of typical mis-sion,then provides the determination methods of model parameters.The calculation example shows that according to relevant parameters of ships,aircrafts,ammunition,this model can calculate quickly aviation ammunition demand,and can provide methodologies for developing aviation ammuni-tion support plans.
Aiming at lithium batteries'intensified temperature rise and thermal runaway under high-load and extreme conditions,conventional cooling technologies lack efficiency and struggle to ba-lance pressure drop with heat dissipation,demanding lightweight and efficient liquid-cooling solutions.This study combines numerical simulations with experimental validation to establish a Bernardi heat generation equation and fluid-solid-thermal coupling model,which is verified with an average relative error of 0.58%.Systematic parametric analysis is conducted to explore the effects of channel configu-ration,mass flow rate,and water droplet-shaped spoilers on cooling performance.Results show that the serpentine channel delivers the best cooling with a maximum temperature of 30.9℃and tempera-ture difference of 1.3℃,but it incurs a significant pressure drop of 5 740 Pa.In contrast,linear and annular channels have lower pressure drops of 272 Pa and 404 Pa respectively,while their maxi-mum temperatures reach 35.3℃and 35.2℃.When the mass flow rate exceeds 60 mL/min,the cooling rate decreases from 0.24℃/(mL/min)to 0.04℃/(mL/min),and the range of 30～45 mL/min is optimal for balancing heat dissipation and energy consumption.Water droplet-shaped spoilers offer limited cooling enhancement,reducing the maximum temperature by 0.6℃for serpentine channels and 0.5℃for annular channels,while increasing the pressure drop by 49.3%and 40.5%respec-tively.Increasing the angle of attack does not alter the maximum temperature.This study clarifies the limitations of smooth spoilers under laminar flow and provides parametric guidance for the lightweight,low-resistance,and high-efficiency design of lithium battery liquid-cooling systems.
In order to solve the problems of WOA in UAV 3D route planning， such as falling into local optimal， low convergence accuracy and failing to solve multi-objective optimization， this paper constructs a multi-objective urban street battle route planning model with path length cost， complexity cost and concealment cost as the objective function， and proposes an improved multi-objective whale optimization algorithm. Firstly， the good point set strategy is used to improve the diversity of the initialization population. Secondly， an adaptive spiral search strategy is designed to dynamically balance the global exploration and local exploitation capabilities of the algorithm. Thirdly， Lévy flight is introduced to interfere in the later stage of iteration to enhance the ability of the algorithm to jump out of local optimum. Then， the sorting method is used by replacing the crowding distance with widely distributed reference points to enhance the convergence accuracy and distribution uniformity of Pareto frontier. Compared with MOWOA， NSGA-II and MOPSO on test functions， IMOWOA is about 4.2 times， 145 times and 730 times higher than MOWOA， NSGA-II and MOPSO on average. In the three-dimensional environment， compared to MOWOA， NSGA-II and MOPSO， IMOWOA increases by 88.7%， 106.7% and 35.8%， respectively， when the path length cost is optimal. When the path complexity cost is optimal， it increases by 98.6%， 97.5% and 10.5% respectively. When the concealment cost is optimal， it achieves 100% performance improvement. It is proved that the proposed algorithm has significant advantages in terms of convergence accuracy and solution set distribution， and can plan a flight path with better comprehensive performance for UAV.
To address issues such as poor amplitude-phase consistency in multi-channel Butler matrix networks and challenges in integration with antennas， this paper proposes a novel layout of Butler matrix for operation at X-band. This work consists entirely of microstrips without any electronic component. While avoiding crossovers， the phasic sequentiality of outputs is ensured by printing a shared metal ground and slotted quadrature couplers on the double-layer substrate. This paper achieves better performance than traditional Schiffman phase shifter as the number of microstrip sections of phase-shift branch is added. The output phase error and the amplitude imbalance of the feeding network are controlled within ±10° and 2 dB， respectively， while the port isolation is maintained below -10 dB. Utilizing microstrip dipole antennas as radiating element， an integrated design of the antenna and feeding network is achieved without the need for coaxial transitions.
This paper proposes a target sorting and localization method based on multi-scale grid division and weight measurement for multi-radiator passive radar systems. The receiving unit receives signals reflected by targets from radiation sources and calculates the time difference of arrival （TDOA） between any two radiation sources to the same target through correlation operations， thereby calculating the corresponding distance difference. According to any two radiation sources and any one target， a hyperboloid can be determined. Based on this， target sorting and localization are achieved by statistical analysis of the number of hyperboloids passing through space， thus avoiding complex equation solving. Exploiting multi-target spatial sparsity， a multi-scale grid cascaded processing approach is used. Firstly， panoramic perception of the detection area is formed through coarse grid division and grid weight measurement. Then， fine grid division is performed on key areas to achieve high-resolution detection. By adopting a hierarchical processing method with coarse and fine grids， computational complexity can be effectively controlled while ensuring target sorting and localization accuracy. Based on theoretical analysis and simulations， the effectiveness of the proposed method is verified.
Significance Microwave plasma control has emerged as a promising route for actively regulating the combustion of composite solid propellants because it provides externally imposed,electrically control-lable energy deposition without fundamentally changing motor architecture or grain geometry.Com-pared with mechanically complex variable-thrust approaches,microwave excitation offers the advan-tages of rapid response,flexible modulation,and lower structural penalty,making it attractive for future propulsion systems requiring thrust adaptability,enhanced ignition reliability,and improved combustion stability.However,when microwave plasma is applied to composite solid propellant com-bustion rather than single-phase gaseous flames,the governing physics becomes substantially more complicated.Condensed-phase decomposition,gas-phase reaction,metal-particle ignition and oxida-tion,agglomeration,radiative transfer,and electromagnetic loading coexist in a strongly coupled,high-pressure,particle-laden environment.Existing studies have therefore demonstrated both clear potential and persistent challenges,including inconsistent definitions of incident,reflected,and absorbed power,strong dependence on platform-specific boundary conditions,and incomplete mecha-nistic criteria for distinguishing thermal,chemical,and multiphase effects.A systematic review is thus required to clarify the current state of this field and to identify the key scientific and technical issues governing its future development. Progress This review summarizes recent progress in microwave plasma regulation of composite solid propellant combustion from four closely linked perspectives:coupling concepts and power accounting,typical coupling configurations and key experimental techniques,combustion-response characteristics and mechanistic evidence chains,and predictive models together with engineering evaluation approaches.Representative coupling architectures include waveguide/resonant-cavity systems,coaxial or re-entrant resonators,remote plasma injection by microwave torches,combustor-integrated resonant cavities,and structure/material-assisted coupling schemes(Figs.1,4～8).These architectures differ markedly in field localization,impedance sensitivity,contamination tolerance,and matching strategy,which directly affects effective energy deposition and cross-platform comparability.One major advance has been the recognition that nominal generator power is not an adequate descriptor of microwave action intensity;meaningful comparison instead requires a traceable power-chain description involving reference-plane definitions,incident/reflected power measurement,and,where possible,absorbed-power estimation,together with time-dependent matching information.Another important line of pro-gress concerns combustion-response regularities.Under atmospheric conditions,microwave plasma has been shown to produce substantial and composition-dependent burning-rate enhancement.In represen-tative alkali-doped systems,the burning-rate gain ranges from only 3.0%for neat ammonium perchlo-rate to 62.3%for a heavily NaNO3-doped aluminized formulation,accompanied by large changes in calculated electron mole fraction and baseline specific impulse,indicating that ionization-promoting additives and microwave-absorbing multiphase products strongly influence the net response(Table 1).Under elevated pressures,the response remains significant but becomes increasingly shaped by coupling-state evolution and multiphase effects.For a propellant containing w(NaNO3)=3.5%under 1 000 W excitation,the burning-rate gain was reported as 38.2%at 1 atm and 10.3%at 7 MPa,while the characteristic agglomerate size at 7 MPa decreased by 67%,showing that macroscopic burning-rate enhancement and condensed-phase/particle-process modification do not necessarily scale in the same way under motor-relevant pressures(Fig.11).More recent work has further revealed the coupled evolution of combustion response and electromagnetic loading over 0.1～7 MPa and 0～1 400 W.Under atmospheric conditions,the burning-rate gain increased from about 21.23%at 800 W to 40.55%at 1 400 W,whereas reflectivity rose from about 13%to about 36%,demonstrating that larger apparent gains at high power may coexist with lower coupling efficiency and a narrower stability window;at fixed power,the gain decreased systematically with increasing pressure,for example from 17.5%at 1 MPa to 9.7%at 7 MPa under 1 250 W(Fig.12).These results indicate that microwave regulation should be interpreted through the coupled sequence of nominal input,reflection/matching evolution,effective deposition,and combustion response rather than through nominal power alone.Mechanistically,the reviewed literature increasingly converges on three interacting pathways.The first is enhanced thermal feedback,in which microwave energy is absorbed and thermalized in the flame zone,condensed-phase/product layer,or particle cloud,thereby increasing local temperature and heat feedback to the burning surface.The second is chemical activation,in which non-equilibrium plasma generates energetic electrons,radicals,and excited species that accelerate key reaction pathways and reinforce gas-phase reaction intensity.The third is multiphase coupling and feedback,where particle heating,ignition acceleration,agglomeration-path modification,and particle-cloud loading simultaneously affect combustion and the electromagnetic load itself(Figs.2～3,13).Although macro-level evidence based on burning rate,luminosity,and pressure response is abundant,closure of the mechanistic evidence chain remains incomplete because synchronized diagnostics of near-surface temperature/heat flux,plasma parameters,radical signatures,and time-resolved particle statistics are still limited.Considerable progress has also been made in model development.Existing predictive approaches can be grouped into empirical/semi-empirical correlations,energy-balance or thermal-feedback formulations,and higher-order coupled electromagnetic-plasma-combustion models.Of particular significance for engineering use,a semi-empirical model incorporating microwave and pressure-coupling terms has achieved a coefficient of determination of 0.977 for burning-rate predic-tion over 0.1～7 MPa and 0～1 400 W,indicating that experimentally observed regularities can already be translated into useful quantitative tools when supported by a consistent power-accounting framework(Fig.14). Conclusions and Prospects Overall,the reviewed studies demonstrate that microwave plasma has strong potential for burning-rate enhancement and multiphase-process regulation in composite solid pro-pellants.Nevertheless,its transition from laboratory demonstration to engineering application depends on whether two issues can be resolved simultaneously:establishing a predictive relationship between effective energy deposition and combustion response,and defining a practical operating envelope con-strained by efficiency,dynamic response,and reliability.Future work should therefore prioritize four directions:standardization of power accounting and reference-plane reporting;synchronized diagnos-tics linking deposition,flame structure,heat feedback,and particle evolution;development of trans-ferable models and absorbed-energy-based evaluation metrics;and advancement toward high-pressure long-duration operation together with closed-loop control strategies.If these issues are systematically addressed,microwave plasma regulation is expected to evolve into a viable enabling technology for controllable thrust,ignition enhancement,and combustion-stability management in next-generation solid propulsion systems.
In a complex and dynamic environment， achieving high-precision navigation and control for unmanned aerial vehicles （UAVs） largely depends on the accurate estimation of system states and key parameters. This paper systematically studies the mainstream Gaussian filtering techniques based on multi-source information fusion， traces the evolution process from linear Kalman filtering to modern nonlinear filtering， and deeply analyzes the mathematical principles and applicable boundaries of various filtering algorithms such as linear optimal， nonlinear， adaptive， and robust filtering algorithms within the Bayesian estimation framework. Through multi-dimensional horizontal comparative analysis， this paper reveals the core performance differences of various filters in terms of calculation accuracy， algorithm complexity， and environmental robustness， and clarifies that combining lightweight and intelligent fusion algorithms is the cutting-edge trend and main challenge in addressing the current UAV navigation technology bottlenecks. Finally， this paper summarizes and refines the comprehensive selection suggestions for Gaussian filters in different application scenarios.
As the tactical and technical requirements for air-to-air missiles （AAMs） continue to increase in modern warfare， the design and verification of their general quality characteristics face the dual challenges of complexity and efficiency. This paper addresses the limitations of traditional design methods in the design and verification of general quality characteristics， and proposes a digital transformation technology route based on model. It conducts research on technologies such as data refinement and knowledge management， model-based collaborative design platform support， and AI-empowered application of general quality characteristics， then establishes an overall architecture for the digital transformation of AAMs general quality characteristics design， comprising three layers： foundation layer， tools layer and application layer. The paper shares the technical implementation path of an integrated design process and the development of general quality characteristics domain-specific large models. Finally， taking a model-based thermal design process for air-to-air missile actuator as the example， the effectiveness of digital transformation technologies in supporting demand management， design analysis， and virtual validation is verified.
Synthetic aperture radar （SAR） possesses all-weather and day-and-night imaging capabilities， but SAR image interpretation is complex， which limits its application range. Translating SAR images into optical images that better conform to human visual perception has emerged as an effective approach to facilitate SAR image interpretation. Diffusion models can implement SAR-to-optical image translation step by step， generating high-quality images. However， the inference sampling efficiency of diffusion model is low， which consumes too much time and affects the efficient execution of downstream tasks. To address this issue， this paper proposes a fast SAR-to-optical image translation algorithm S2C2DM. Based on the consistency diffusion model， SAR image conditional guidance is added to the optical image consistency reconstruction process， and deep feature extraction is used to mitigate modality differences between SAR and optical images. By improving the fitting accuracy of the conditional consistency function， it achieves high-quality single-iteration conversion from SAR to optical images. The evaluation of S2C2DM is conducted on two datasets， SEN1-2 and SAR2Opt. Experimental results show that compared with the existing advanced model S2ODPM， the sampling efficiency is increased by 271 times， and the color inconsistency issue is improved.
In response to the challenges posed by intercepting highly maneuverable targets， including difficulties in obtaining target acceleration information， insufficient adaptability of traditional guidance laws to high-dynamic environments， low exploration efficiency of reinforcement learning algorithms in wide spatial-temporal contexts， and poor training stability， this paper proposes a terminal guidance law based on a double random distillation network and real proximal policy optimization algorithm. Firstly， a Markov decision process is designed for the three-dimensional terminal guidance scenario. Subsequently， deep reinforcement learning algorithms are employed to train high-speed vehicles. To enhance the exploration efficiency of these vehicles within wide spatial-temporal contexts， a double random distillation network is introduced that incentivizes the exploration of unknown states through internal rewards. To accelerate training speed and ensure training stability， the trust region rollback objective function of real proximal policy optimization algorithm is used to replace the clipping objective function for effectively improving the convergence rate while enhancing training stability. Simulation results demonstrate that the proposed terminal guidance law exhibits generalization and robustness when faced with highly maneuverable targets， and can achieve efficient interception without requiring target acceleration information.
For fixed-wing unmanned aerial vehicle （UAV） formation flight systems subjected to internal uncertainties and time-varying external dynamic disturbances， a distributed trajectory tracking control protocol problem is investigated under a directed communication topology in this work. Firstly， for overcoming the strong nonlinearity and strong coupling characteristics of fixed-wing UAVs， the feedback linearization is employed to equivalently transform the fixed-wing UAV nonlinear model into an affine nonlinear system with strict-feedback form. Secondly， radial basis function neural networks are utilized to online approximate internal uncertainties in the fixed-wing UAV formation flight system， and nonlinear disturbance observers are designed to online estimate the compounded disturbances caused by external dynamic disturbances and neural network approximation errors. On this basis， state predictors are constructed based on the outputs of the neural networks and nonlinear disturbance observers， and the prediction errors of state predictors are taken as decision variables for the online updating of the neural networks and disturbance observers. It can further overcome the problems of poor interpretability and low transparency arising from the black-box nature of neural network approximation. Moreover， a distributed anti-disturbance composite learning trajectory tracking cooperative control protocols is developed for the fixed-wing UAV formation flight system based on dynamic surface control and the framework of multi-agent consensus theory. The semi-globally uniformly ultimately bounded stability of the closed-loop formation system is rigorously proved based on Lyapunov stability theory. Finally， simulation experiments verify the feasibility and effectiveness of the proposed control protocol. The designed distributed anti-disturbance composite learning trajectory tracking cooperative control protocols enable formation keeping of fixed-wing UAVs while achieving higher control accuracy and more continuous， smoother control signals than traditional backstepping control protocols.
To explore and address the cross-influence issues arising from the coexistence of two distinct flight and control logics during the transition between high-altitude flight and ground-effect flight modes for aircraft， a novel morphing airfoil suitable for this environment is introduced. By reasonably adjusting the deflection angles of the leading and trailing edges of the wing section during the transition between normal flight and ground-effect flight modes， the aircraft's wings can better adapt to both flight modes and achieve a smoother transition. A neural network algorithm is employed for automatic optimization.The results demonstrate that the morphing airfoil utilizing this algorithm achieves a higher lift-to-drag ratio compared to conventional airfoils in adapting to the aforementioned two flight modes， the maximum increase in lift-to-drag ratio is 130%.
In 2025， the development status of the next-generation manned fighter for the USAF and USN exhibits significant divergence. This paper aims to thoroughly analysis the factors influencing the distinct trajectories of the USAF F-47 and the USN F/A-XX programs， and assesses their respective capabilities and development prospects. Firstly， this paper outlines the differences in progress and funding between the two programs， compares their current development statuses， and explores the underlying reasons. Subsequently， based on the official conceptual renderings of the F-47 and F/A-XX platforms， this paper analyzes their aerodynamic configurations and potential performance characteristics， interprets the explicit information， deliberately obscured details， and implied capabilities depicted in the images. Finally， it summarizes the common characteristics in the development of the US next-generation fighter， and discusses the shifts and potential advantages in acquisition and management models. The analysis indicates that there are fundamental differences in operational requirements and development pathways between the next-generation manned fighters of the USAF and USN. Moreover， basic capability of the national defense industrial and the evolution of warfare forms are the key factors shaping the selection and development direction of these programs.
Objective Urban offensive operations serve as a critical form of modern warfare， yet they are severely restricted by factors such as multi-dimensional space， fragmented combat forces， and complex electromagnetic environments， resulting in a significant decline in the effectiveness of traditional equipment in urban environments. To break through the bottlenecks of urban offensive operations， this study aims to reveal the coupling mechanism between unmanned swarms and urban offensive operations， providing theoretical support and practical tactical solutions for the application of unmanned swarms in typical urban scenarios. Methods This paper adopts a method that combines theoretical analysis， induction and summary with scenario-based design. Based on the urban battlefield environment and typical combat cases， it summarizes the characteristics and dilemmas of urban offensive operations， and sorts out the operational process and winning mechanisms of unmanned swarms in urban offensive operations systematically. A three-level coupling framework is constructed from the dimensions of perception， decision-making， and action. Targeted designs of combat formations， command modes， and operational actions are carried out for the scenarios of high-rise building raids and underground space clearance. Results and Discussions The results show that urban offensive operations are confronted with three core predicaments： multi-dimensional space， fragmented command， and inefficient equipment （Table 1）. The complete operational process of unmanned swarms includes six stages： planning， deployment， formation， penetration， mission execution， and recovery （Fig.1）. This study establishes a three-level coupling model： blind compensation， tracking and positioning， and communication relay at the perception layer； distributed intelligence and human-machine hybrid decision-making at the decision-making layer； decoy deception， maneuver and area control， precision strike， and cooperative jamming at the action layer （Fig.2）. Heterogeneous functional grouping and hierarchical adaptive command architecture are designed for two typical scenarios， and four-stage combat actions are adopted to effectively enhance the efficiency of unmanned swarms penetration， containment and strike. Conclusions This paper reveals the internal coupling mechanism between unmanned swarms and urban offensive operations， constructs a multi-level and multi-dimensional coupling framework， and provides scenario-based tactical designs for typical urban operations. The research results can support the tactical innovation of unmanned swarms in complex urban environments， optimize combat formations and command modes， and offer theoretical references and application guidance for breaking through the bottlenecks of urban operations. In the future， with the continuous development of swarm intelligence， autonomous control， human-machine interaction and other technologies， unmanned swarm systems will further promote the transformation of operational modes and force structures in urban warfare.
With the continuous enhancement of target maneuverability in the combat environment， the characteristics of target maneuverability often exhibit irregular and nonlinear especially during terminal engagements， which make it difficult for missiles to accurately predict the changes of target motion， resulting in increased miss distance. Conventional guidance laws including proportional navigation （PN）， adaptive sliding-mode control and guidance methods based on classical control theory are formulated solely on real-time state measurements. Consequently， those strategies lack the ability to predict the future motion trends of the target. For addressing this issue， this paper proposes a guidance law for variable maneuvering targets using trajectory prediction based on diffusion models. Proposed diffusion model including an encoder-decoder architecture takes the target’s historical trajectory， predicted noise， and diffusion-step embedding as inputs to predict future trajectories based on past motion of the target. During the real-time guidance process， the model predicts the target’s future trajectory online， and estimates the future line-of-sight （LOS） angle variation from the predicted trajectory. Then introduces an accumulation mechanism based on the predicted LOS angle differences. The compensation term for the proportional guidance law is constructed by accumulating the angle deviations over successive time steps and a compensation coefficient. Experimental results show that the proposed guidance method reduces the average interception time by more than 4.5 s under various target maneuvering flight scenarios compared with the conventional guidance law without trajectory prediction. This method still achieves effective target engagement under complex conditions such as high-speed maneuvers.
