Aiming at the obstacle avoidance of multiple quadrotor UAVs in complex and variable wind fields， a aimed trajectory planning algorithm based on Gauss pseudo-spectral method （GPM） is proposed to solve the obstacle avoidance of multiple quadrotors in mixed wind fields. Firstly， a dynamic model for a single quadrotor in a mixed wind field is established. On the basis of this， the constraints on the wingmen in the formation are transformed into constraints on the leader， thereby a multi-constraint trajectory planning model is established. Under the premise of ensuring that the formation can reach the target location on time， this model minimizes the energy consumption required for formation flight is minimized by using this model regarding the optimization objective. The GPM is utilized to optimize the trajectory of the leader， which is then formulated as a nonlinear programming problem （NLP）. The sequential quadratic programming （SQP） algorithm is adopted to solve this problem， which yields an optimal flight trajectory that satisfies both the constraints and the optimization objective. Simulation results demonstrate that the proposed method enables multiple UAVs to safely navigate around obstacles within mixed wind field environments and generate optimal flight trajectories while satisfying the constraints.

Regarding the significant impact of code generation techniques based on large language models （LLMs） on software productivity and their broad application prospects to the aerospace field，the latest research progress on this kind of technology is reviewed from three aspects: problem background and definition，typical technologies and their potential application scenarios in the aerospace domain，and application evaluation methods，with the aim of providing guidance and insights for related research on code generation techniques in the aerospace domain. Firstly，the basic capabilities of LLMs are discussed on code generation according to the features of code generation problem definition and LLMs structures. Then，the main methods for code generation are elaborated，including pre-training，instruction fine-tuning，prompt engineering and retrieval-augmented generation as well as their potential application scenarios to the aerospace field. Next，due to the perspectives of semantic similarity and validation datasets，the popular methods are reviewed for evaluating the results of LLM-based code generation techniques and their characteristics and limitations are analyzed. Finally，the challenges are presented and future improvements are proposed.

The requirements of the flexibility of aerospace control system are extensively analyzed in this paper，and the challenges are deeply discussed，which are concentrated on the system level and the device level. The design concept of the flexibility of the system level and the device level is proposed in innovation，and the key technical path of the flexibility of the system level and the device level is further elaborated，such as rapid assembly and interoperability management of system software and hardware，multi-level communication interconnection，flexible integrated conformal layout and analysis，and verification of the adaptability of flexible electronics to the aerospace environment，and the development trend of the flexibility of the aerospace control system is finally visioned in terms of integrating innovative technologies，strengthening cross domain cooperation，establishing standards and specifications，and expanding and deepening applications.

Under the background of the vigorous development of theory and application of embodied intelligence， in order to further improve the intelligence level of low altitude UAVs and expand their application boundaries， the development requirements of low altitude UAVs are focused on the context of low altitude economy， and the researches are implemented and systematically analyzed， regarding three domains in terms of the theoretical basis， key technology paths and application challenges. Firstly， the advantages and research value of embodied navigation compared with traditional navigation on semantic understanding， environment interaction and group collaboration are determined. Secondly， derived from the development of traditional simultaneous location and mapping（SLAM） to the technology evolution of visual language navigation（VLN） and visual language action（VLA） model， one special navigation technology framework for low altitude UAVs is established. Thirdly， through analysis of two typical application cases， the practical application trend of low altitude UAV in complex environment is discussed. Finally， future development directions of low altitude UAV embodied navigation are overviewed， the proposed achievement can serve as valuable references for theoretical research and industrial application to intelligent autonomous navigation.

The transfer strategies for multi-objective missions around the Earth-Moon Lagrange points are studied and a two-impulse transfer strategy that allows spacecraft to move between near-rectilinear halo orbits and distant retrograde orbits is designed. Firstly，the spacecraft's orbital dynamics model is established in the synodic frame. Then，the overall design of the transfer strategy is implemented，the relevant optimization variables are analyzed，the objective function and constraints are determined，and the transfer strategy design problem is transfered into a trajectory optimization problem. Furthermore，the feasibility of genetic algorithms and particle swarm optimization algorithms for this specific problem is verified for solving the transfer trajectory. A numerical method is presented through this research by applying genetic and particle swarm optimization algorithms to transfer strategies design of halo orbits，which focuses on recent interest in near-rectilinear halo orbits and distant retrograde orbits. The proposed optimization method has ability of effectively resolving the orbital transfer design problem without prior information and can be applied to various transfer scenarios.

Aiming at the orbit determination of non-cooperative continuous low-thrust maneuvering spacecraft，a rapid pre-identification method of thrust acceleration based on single-arc orbit determination is proposed. Based on the relationship between satellite orbit parameters and acceleration，the single-arc orbit determination results of two radar observations with a certain time interval are used to inversely solve the tangential acceleration of the spacecraft under continuous tangential thrust，and fairly smaller solution error is remaining under the condition of sparse data in a short time by using this method Which is applied respectively to the orbital climb of Starlink，OneWeb and the Qianfan constellation. The results show that when the observation interval is greater than 11 h and 15 h separately，the proposed method can realize the rapid pre-identification of the tangential thrust acceleration of Starlink and OneWeb satellites，and the solving error is less than 2%. The calculation results can be used as initial values for precise orbit determination of continuous low-thrust maneuvering spacecraft.

Aiming at the stability control of a slender air defense missile under low elastic frequency and multiple constraints， a linear time-varying model predictive control method which is considered under elasticity and multiple constraints is proposed. Firstly， in order to ensure the amplitude attenuation in the mid-frequency band， the overload error integral output is used to augment the missile's linear state space equation. Under the model predictive control full state feedback strategy， the controller inherits the structure of the traditional three-loop control method. Then， a cost function is established， which is based on the criterion of ensuring the rapidity and smoothness of the missile's overload and attitude response， and the missile's stability control problem is transformed into a rolling optimization solution of model predictive control. Finally， in order to ensure the stability and elastic suppression capability of the system under elastic working conditions， a notch is inserted into series at the output end. The simulation results show that the proposed method has faster response speed， higher control accuracy and better elastic suppression ability， compared with the traditional three-loop control method.

A guidance law with variable control weight is proposed，which is based on the optimal control method for intercept different types of maneuvering targets by using air-defense missile. A state equation of relative motion between missile and target in the longitudinal plane is set up，and an indicator function based on variable control weight and line of sight angular velocity constraint is designed，furthermore，the optimal guidance law related to weight coefficients is derived by using the minimum numerical principle. Simulation experiments are implemented for different types of maneuvering targets，and the results indicate that the performance of the optimal guidance approach proposed in this paper is superior to traditional proportional guidance method. The line-of-sight angular velocity convergence is fairly quick，which can form preferable reverse orbit situation. Furthermore，regarding targets under different maneuvering modes，the weight of the optimal guidance law can be adjusted to match the optimal control parameters of each maneuvering target.

Aiming at the current situation of tight software-hardware coupling， large scale and high complexity in the test and launch control system of aerospace equipment， a layered decoupled software-defined test and launch control system architecture is proposed. Through taking advantage of middleware technology， the hardware resources of test and launch control system are highly integrated and abstracted， and application-oriented software is enabled to dynamically load and reconstruct components based on users' requirements. The proposed architecture has the features of hardware-on-demand expansion and flexible high extensibility of software， which improves the utilization rate of test and launch control system resources， task flexibility and system reliability.

Aiming at the shortage of existing fault diagnosis methods for in-orbit satellites， a satellite fault diagnosis system based on Clips expert system is proposed in this paper. The satellite fault is diagnosed in real time by taking advantage of existing satellite fault knowledge and experience， combining expert experience with real-time telemetry data and using inference machine technology. At the same time， expert knowledge is input and edited by visualization technology including graphics and knowledge expression. The expert knowledge base is established to transform the simple logic statements that is easy to be understood by users into complete and complex Clips statements to realize complex fault diagnosis programs， and it is going on to realize the automation and intelligence of real-time fault diagnosis of spacecraft in orbit， accurately locate faults and improve the reliability and safety of satellite systems. Through multi-domain simulation of multiple fault scenarios and project practice， the diagnosis results of the satellite fault diagnosis system are the same as the actual fault data received by the satellite， which verifies the effectiveness of the designed system.

In order to achieve formation flight for unpowered aerial vehicles and ensure the convergence of formation errors in three-dimensional space， a formation control method based on consensus theory is proposed. Firstly， by taking into account the formation group's characteristics of only having negative acceleration， the consensus theory is modified， and a longitudinal control method suitable for unpowered aerial vehicles is introduced. Next， a bank angle distribution and flipping strategy is involved， and under the condition that the lift is prioritized to satisfy high-directional control requirements， the bank angle flips by the lateral error corridor， thereby both high-directional and lateral errors are simultaneously reduced. Finally， simulations are conducted to verify the adaptability of the proposed method to various formation sizes and communication topologies. The simulation results demonstrate that， within the given range of initial conditions， the proposed method can effectively control the formation in different initial states. Moreover， the initial state of the cluster has significantly impacts on formation time and cluster speed loss and average formation time and lower average cluster speed have less loss while there are clusters with more directed edges in the topology structure.

As a critical factor affecting safety-critical systems， it has been drawn increasingly attention to software safety issues. Based on engineering practices in aerospace embedded software testing and verification， the software safety requirements are taken as a clue and typical safety problems are focused in this paper. The two dimensions are introduced by the formal verification of source code safety properties and the automated testing of software safety requirements which analyze and summary key technologies， including special safety analysis， source code static analysis， source code model checking， fault-model-based safety testing and keyword-driven automated testing. A comprehensive technical solution for aerospace embedded software safety verification is proposed， and an independent controllable software assurance support platform and tools are developed to systematically enhance the trustworthy assurance capabilities of aerospace embedded software.

On account of the issue that the communication network of multi-flight vehicles is vulnerable to be attacked by false data injection in the disturbance environment， a collaborative guidance strategy against high-speed targets that can autonomously identify and suppress the fault link is proposed. By designing the coordinated guidance method based on relative distance， when only the leader flight vehicle can get the target movement information， the flight vehicles can hit the maneuvering target at the specified time. The radial basis network is used to estimate the unknown items for the design of cooperative guidance law. The trust coefficient is introduced to identify the false data injection attacks on the communication network， which guarantees the information assurance for self-suppression of the multiple flight vehicle systems under the network attacks.The numerical simulation result indicates that the leader and followers multi-vehicle collaborative impact on the high-speed target simultaneously can be controlled and implemented by using the proposed method under the false data injection attacks.

A design method for an active disturbance rejection controller based on complementary sliding mode is proposed to address the position and attitude control problem during the point cloud data collection process of quadrotor unmanned aerial vehicle （UAV） that carries a light detection and ranging （LiDAR） for power transmission lines. Regarding the problems of strong coupling，non-linearity，and external disturbance effects under the complex environment of quadrotor UAV，a finite time convergent extended state observer is adopted to estimate the state and lumped disturbances of the quadrotor UAV dynamic system，which introduces the observed lumped disturbances into the active disturbance rejection controller for feedforward compensation. At the same time，complementary sliding mode manifolds are established，and the exponential power function and the integral form of sign function are used to ensure the continuity of the active disturbance rejection controller. Finally，a rigorous proof of the asymptotic convergence of position and attitude tracking errors is provided，which is based on the Lyapunov analysis method，and the simulation results show that the proposed method has higher control precision and can effectively suppress different forms of external time-varying disturbances.

A hybrid optimization method based on deep neural networks is proposed in this paper for the spacecraft pursuit-evasion game problem with free terminal time and {\mathrm{J}}_{\mathrm{2}} perturbation considerations. Firstly，the data set is generated by solving the two-point boundary value problem without {\mathrm{J}}_{\mathrm{2}} perturbation using traditional optimization algorithms. Then， on the basis of that， a deep neural network is established to fit the relationship between the initial state and the solution， and the initial guess solution is generated. Finally， the solution is further optimized by using a local optimization algorithm. Through simulation and verification， it is demonstrated that this method not only performs good feasibility and robustness but also significantly improves computational efficiency， which is compared with traditional hybrid optimization algorithms.

The algorithm of the integrated celestial navigation combined accelerometers with X-ray pulsar sensor is researched. Firstly，the inertial navigation algorithm is provided for spacecraft based on the heliocentric inertial reference frame，and the accelerometer drift is estimated by the phase difference of the inertial navigation and X-ray pulsar sensor in pulsar direction based on PI filter. Then，the algorithm of the integrated celestial navigation combined X-ray pulsar sensor with accelerometers based on the time of arrival （TOA） of the pulsar pulse is proved. Lastly，simulation results show that the algorithm of the integrated celestial navigation is effective.

According to the position and attitude control problem of quadrotor unmanned aerial vehicle （UAV） with external disturbances and time-varying loads，a parameter adaptive sliding mode control method is proposed，which is based on the parameter adaptive method and sliding mode control theory.The external disturbances and time-varying loads of quadrotor UAV are estimated and compensated by using parameter adaptive method so that the quadrotor UAV can be enabled to resist the disturbances caused by changes in time-varying loads and the capability of high-precision position and attitude control can be achieved. In the end，the Lyapunov stability theory and simulation results fully verify that the proposed method can effectively degrade the adverse effects of unknown disturbances and time-varying loads on the control system.

Regarding the ascent trajectory of horizontal take-off and landing reusable launch vehicles，a simplex-pseudospectral loop optimization algorithm is proposed. By optimizing the control parameters of the horizontal run section，the initial value of the ascent section is adaptively generated，and taking a certain horizontal take-off and landing reusable launch vehicle as an example，the trajectory optimization design simulation of the launch vehicle from taking off to climbing to the handover point is completed by using the simplex-pseudospectral double-loop optimization algorithm，which verifies the feasibility and effectiveness of the proposed method. On this basis，the influences of engine thrust and sensitive parameters of the flight trajectory on air-breathing mode flight profile and remaining mass are studied. At the same time，a design method of air-breathing mode power compensation is proposed. By comparing with the conventional design method，fuel consumption is further reduced and the carrying capacity of the reusable launch vehicle is improved.

A combined navigation positioning method based on improved radial basis function neural network （RBF） assisted volume Kalman filter （CKF） is proposed to resolve reduced precision of integrated navigation positioning caused by global navigation positioning system （GNSS） signal interruption in complex environments such as tunnels， urban roads and canyons. Firstly， the integrated navigation fusion data is preprocessed by using kernel principal component analysis （KPCA） combined with K-means++ clustering model to make its distribution representative； Secondly， the orthogonal least squares （OLS） method is used to determine the number and center values of hidden layer neurons in the RBF neural network， and the trust region constrained Gaussian-Newton （TR-CGN） algorithm is used to optimize its parameters； Finally， when the GNSS signal loses lock， the trained improved RBF neural network is used to assist in nonlinear CKF filtering for error compensation. The experimental results show that the average positioning error is reduced by 17.87% through application of this method without increasing hardware costs which is compared to the way of using the autonomous driving collaborative positioning system； Compared with the average positioning error assisted by KPCA-RBF， the reduction based on the proposed method takes advantage of 54.37%， which indicates that the adaptability and robustness of the integrated navigation positioning system are effectively enhanced within complex environments.

A predictive control method for electromagnetic switching valve is proposed for attitude control of spacecraft. Based on the predicted control time and nozzle layout，the ignition logic of electromagnetic valve is designed and the force of the nozzle is utilized to cancel out each other，so that the control torque action time can be adjusted arbitrarily during the control cycle. The simulation analysis shows that compared with conventional pulse width modulation techniques，the predictive control method allows the solenoid valve to achieve precise attitude control without being limited to the minimum continuous opening time.

Traditional scene matching methods for unmanned aerial vehicles （UAVs） in low-altitude environments often suffer from ineffective outlier rejection， leading to degraded positioning precision. To address this issue， an improved scene matching localization algorithm is proposed in this paper. Firstly， initial data are generated by using triple relationships in this algorithm. Subsequently， a ternary matching optimization method is introduced by combining triangular feature similarity measurement and maximum Euclidean distance screening to reduce computational costs and enhance matching correctness. Furthermore， a data refinement strategy is adopted to improve the sampling performance of the algorithm. Simulation results demonstrate that the proposed algorithm achieves superior accuracy and real-time performance for UAV scene matching localization in complex low-altitude environments， which significantly improves computational efficiency and positioning precision.

Aiming at full system test data and career information， environmental data， etc of multiple source and huge heterogeneous data of aerospace equipment， a data-based health management system is developed，which achieves the integration and storage management of test data，product career data，environmental data， model analysis and health assessment.The system is developed through service-oriented architecture design， which is based on information gathering and processing for realizing integration and management of equipment test data and takes advantage of high-reliability test data storage and management foundation using lightweight distributed column storage mechanism.Regarding the health status of the system and key units， the trend analysis，life prediction， maintenance decision information and hierarchical situation presentation， which are based on algorithm models，are introduced. Realization of the health status evaluation，health trend analysis，life prediction，maintenance decision support，etc for the full system and key units can be served as a reference of equipment data health management solution.

The anti-peak circuit of the timing attitude control system of the launch vehicle adopts a "diode+resistor" circuit and a "resistor+diode+voltage regulator diode" circuit. MULTISIM is used for simulation and experimental verification. It is found that the theoretical solenoid valve shutdown time becomes shorter when anti-peak resistance grows larger，and the voltage regulator diode voltage goes larger，separately. Regarding the same electromagnetic valve shutdown time index requirements，a lower reverse voltage produced by using the method of increasing the voltage regulator than using the method of increasing the resistance； Due to the precisely controlling of the flows of fuel and gas that are ensured to be delivered to the engine in the predetermined ratio and time，the boost valve，relief valve and auxiliary power solenoid valve.play an important role of adjusting regulating pressure and protecting safety in the attitude control system，which maintain the stability of the launch vehicle's attitude and flight safety. Therefore，their shutdown time and anti-peak voltage need to be precisely controlled.

To address the challenges of insufficient control stability， limited navigation precision and poor generalization ability encountered by unmanned aerial vehicles during implementation of autonomous visual navigation tasks， a brain-inspired convolutional neural network-spiking recurrent neural network （CNN-SRNN） is proposed to achieve robust end-to-end stable flight navigation strategies with strong generalization capabilities. This network architecture simulates the flight control circuit of the fruit fly brain， which uses CNN network by extracting visual features to form high-level state representations and integrates with an attention mechanism for precise target recognition and localization. A spiking recurrent neural network （SRNN） serves as the flight navigation controller， which realizes time sequence motion information integration and flight control. Additionally， a regularization strategy based on Gershgorin disc theory is designed to enhance the stability of navigation control. Evaluations of UAV navigation performance through diverse simulation environments demonstrate that CNN-SRNN network has the outstanding scene-generalization capability， robustness against noise and decision-making stability. The encoding and decoding relationship between neural activation patterns in SRNN and UAV flight trajectories is further analyzed， and the navigation control mechanisms of brain-inspired neural network are revealed and model interpretability is significantly improved.

Based on the upper wind benchmark of the sounding balloon，the upper winds and the precision of corresponding maximum aerodynamic load from the wind profile radar and numerical weather prediction model forecasts from 1st day to 4th day are compared and analyzed. The results show that: the precision of the upper wind from low to high is presented by the wind profile radar and the forecast of the 4th day to the forecast of the 1st day. The precision of the wind profile radar at the altitude of 7.6 km and above is obviously low，and absolute difference is more than 5 m/s； The precision of maximum aerodynamic load from low to high is wind profile radar and the forecast of the 4th day to the forecast of the 1st day. The average absolute differences of the maximum aerodynamic load from wind profile radar and the the forecast of the 1st day to the forecast of the 4th day is respectively showed by 326.72 Pa∙rad，126.53 Pa∙rad，162.26 Pa∙rad，183.15 Pa∙rad and 212.59 Pa∙rad，and the correlation coefficient values are separately recorded by 0.76，0.98，0.96，0.95 and 0.92. Therefore，the precision of the maximum aerodynamic load from the wind profile radar is low，which cannot be used for the safety guarantee of rocket flight and needs to be further improved.

To address the issues of complex modeling procedures of conventional methods for quadrotor UAVs carrying time-varying slung loads and the poor adaptive capability of traditional PID controllers under complex wind disturbances， a Kane's method-based dynamic modeling approach and a model reference adaptive control （MRAC） method with adaptive learning rates are proposed. Kane's method takes advantage of combination of forces and partial velocities， which eliminates the need for explicit analysis of cable constraint forces required by Newton-Euler formulations and bypassing the Lagrangian function established with second-order derivative computations， thereby the calculation process is simplified. The adaptive learning rate's MRAC method enables quadrotor UAVs to resist composite disturbances from wind and time-varying load variations through adaptive learning rates application and variable parameter control， which achieves precise position and attitude control. Simulation results show that under composite disturbances from time-varying loads and complex wind fields， the adaptive learning rate's MRAC demonstrates superior performance in both overshoot suppression and convergence rate compared with conventional MRAC.

The evolutionary trajectory of aerospace control technology is focused from classical control and modern control to agent-featured intelligent control technology 3.0. The agent-featured intelligent control technology 3.0 is represented and known as the key indicators of future aerospace control systems. The key attributes of control technology 3.0 labelled by "learning while flying"， "lifelong learning" and the "new-generation system architecture" are pointed to elucidate. The critical technologies of "intelligence empowerment"， "functional augmentation" and "information capability enhancement" are subjected to in-depth analysis. On this basis， the exploration of future aerospace intelligent control development is oriented to typical scenarios such as large model empowerment and software factories. Consequently， prospective thoughts on the development of advanced intelligent aerospace control are expanded upon the matter.

To address the issue of safety limitations of traditional path planning algorithms in dense obstacle environments， a Voronoi diagram-based safe obstacle avoidance algorithm is proposed for polygonal obstacle regions. Firstly， a circular coverage model with minimal overflow rate is established to optimally encapsulate obstacle areas. Subsequently， a Voronoi diagram construction algorithm is designed， which is based on the circular coverage to generate a navigable skeleton of the free space. Furthermore， a path generation method integrating unmanned aerial vehicle （UAV） kinematic constraints is developed by using the skeleton. Finally， cubic B-spline interpolation is applied for ensuring path smoothness. The results of simulations demonstrate that， compared with paths generated by an improved A* algorithm， the proposed method achieves smoother trajectories while maintaining comparable path length and significantly increasing the minimum distance to obstacles， and highlighting its superior safety performance in obstacle avoidance. The research can serve as a practical solution for ensuring safe UAV navigation in complex urban environments.

To address the challenge of significant attitude deviations and persistent oscillations in coaxial unmanned aerial vehicles （UAVs） under wind disturbances， an adaptive backstepping attitude control method is proposed. Firstly， a nonlinear attitude dynamics model integrating with wind disturbances is established through mechanical modeling for coaxial UAVs. Subsequently， a neural network is employed to estimate real-time disturbance amplitudes in the pitch and roll channels for the torque and model uncertainties caused by wind disturbances， while an adaptive backstepping controller dynamically adjusts control parameters for precise stabilization. Finally， the performances of attitude tracking and disturbance resistance of control system are tested through simulations. Comparative simulations demonstrate the adaptive backstepping based method has superior performance over PID control in attitude tracking accuracy and disturbance rejection robustness and significant improvements in overshoot suppression and oscillation attenuation. These results validate this solution in complex disturbance environments for coaxial UAV attitude control.

Regarding the focused security requirements of test launch and control system in authentication and authorization， data security protection and traceability of abnormal operations， researches are conducted on the application of blockchain technology in authentication and authorization management. The design of network architecture， account information model， consensus mechanism and differentiated authorization smart contract algorithm are elaborated in this paper， which is verified through the simulation by using self-developed measurement and control network blockchain authentication and authorization platform.

In this paper， the federated contrastive learning method for SOH estimation of lithium batteries is proposed. Firstly， federated learning is utilized to jointly train a model across multiple clients， which can enable knowledge sharing among clients while data privacy is protected. Next， in the federated learning framework， the concept of contrastive learning is introduced to achieve feature alignment among multiple clients， thereby the data distribution differences among clients can be reduced. Further considering the differences in data quality among different clients， a dynamic weighted aggregation algorithm is proposed to reduce the impact of low-quality data on the global model. Finally， the effectiveness of the model is validated on the 18650 lithium battery data set and data privacy protection and the fairly lower SOH estimation error are both guaranteed.

To address the issues of challenges posed by the resource constraints faced by emergency UAVs equipped with communication base stations， a heterogeneous integration architecture and system-level resource constraints optimization research is proposed. Firstly， based on the characteristics of UAV subsystems， a heterogeneous air-space-ground integrated emergency communication framework is constructed. Secondly， specific performance indexes and functional requirements are formulated under resource constraints， and targeted interference mitigation solutions are developed. Finally， through link budget analysis and field experimental validation， the effective coverage range at the recommended transmission rate is empirically determined. By ensuring UAVs which can "fly farther， see clearer， be reliably controlled and be effectively utilized" during critical moments and supporting the sustainable development of the low-altitude economy， the proposed research has points to achieve the goals.

An adaptive sliding mode control strategy based on the randomized feedforward neural network is proposed for the path tracking control problem of heavy-load quadrotor unmanned aerial vehicle （UAV） affected by model uncertainties and external unknown disturbances. The dynamics system of heavy-load quadrotor UAV is experienced by the model uncertainties and external unknown disturbances during flight which are defined as the lumped disturbance term. By taking advantage of randomized feedforward neural network， the lumped disturbance term in each channel is adaptively estimated and then used to compensate the adaptive sliding mode controller. Based on the Lyapunov stability analysis method， a rigorous proof of the convergence of path tracking errors for heavy-duty quadrotor UAV is presented. The effectiveness of the proposed control method is fully verified by the simulation results.

To address the issue of challenge of rapid and accurate prediction of the trajectory terminal velocity by using offline trajectory optimization methods during unmanned vehicles operation under complex and strong interference， an integrated velocity prediction and control algorithm is proposed， which is based on the improved gated recurrent Unit neural network algorithm. The velocity prediction is based on the parameters of the neural network model trained by a trajectory data sample library， which takes an eleven-dimensional feature sequence including the target position， current altitude， velocity， ballistic angle and other relevant parameters as input of the network and the velocity at the final moment as the output， and yields a neural network model capable of predicting terminal velocity. Based on the velocity prediction results， a decoupling control scheme for velocity and position is employed for terminal velocity control. The predicted terminal state deviations are used to correct and generate closed-loop control inputs for terminal velocity regulation. In final stage， the designed velocity prediction and control method are validated through six-degree-of-freedom （6-DOF） ballistic simulations. The simulation results demonstrate that accurate and effective velocity prediction and control can be relatively achieved by using proposed algorithm applied to the terminal velocity under 6-DOF closed-loop state.

A positioning method is proposed for high Earth orbit （HEO） spacecraft based on Chebyshev orthogonal domain transformation. By transforming the time-varying receiver coordinates over a period into an invariant Chebyshev coefficient domain， this approach is based on effective combination with sparse ranging observations obtained by the spacecraft across different epochs， which enables joint resolution of multi-epoch measurements. Under conditions of sparse satellite visibility， the historical observation data is leveraged to provide effective constraints for positioning on the current epoch and achieve continuous and reliable positioning for medium-high Earth orbit （MHEO） spacecraft. Furthermore， the Chebyshev based positioning method demonstrates robustness against random measurement errors during observation， which enhances GNSS positioning precision in high-altitude environments. The experimental and simulation results demonstrate that the method is superior to the Extended Kalman Filter （EKF） by handling random noise and surpasses traditional least squares algorithms in both computational speed and positioning precision， which is capable of continuous positioning for spacecraft by pseudorange-level in high Earth orbit scenarios.

Aiming at comprehensive optimization of the robust disturbance rejection capability， convergence time， and control accuracy of traditional UAV cooperative formations， a particle swarm optimization-based fast robust cooperative control method for multiple UAVs is proposed in this paper. A finite-time cooperative formation controller is designed to accelerate the response speed of traditional distributed formations. A fast disturbance observer is developed to compensate for the control system， which is capable of accurately estimating composite disturbances within a finite time， thereby the formation control precision and robust disturbance rejection are enhanced. On this basis， by considering both the convergence time and control error， a penalty-based particle swarm optimization algorithm is employed to optimize the design parameters of the formation system， which comprehensively improves the flight performance of multiple UAVs robust cooperative control.

Regarding the online evaluation of the flight capability of aerospace transportation spacecraft， a customized online reachable domain calculation method is proposed. Firstly， a reachable orbit envelope calculation scheme is designed around the current orbit plane under given constraints and initial conditions. Secondly， customized methods and hardware product modules for heterogeneous acceleration are designed， which can quickly achieve online calculation of reachable orbit envelopes and provide fuel optimization planning guidance for specific orbits. Finally， the proposed customized calculation method for reachable domain is validated from the aspects of reachable domain calculation analysis and planning guidance. The results show that the proposed algorithm has good convergence， can quickly calculate the reachable domain envelope and the control variables can smoothly adapt to changes in orbital parameters.

The utilization of drones and other devices for low-altitude inspection is recognized as a typical application scenario in the future low-altitude economy. Infrared imaging， as a critical tool for low-altitude environmental perception， faces challenges in acquiring reliable datasets due to high costs and difficulties in ensuring confidentiality. A style transfer-based target implantation method is proposed to generate simulated infrared images. On the basis of this method， which is initiated by solving the target temperature field through finite element analysis， atmospheric transmission effects are incorporated to render preliminary simulated images. A convolutional neural network-based style transfer technology is then utilized to achieve high-quality implantation between targets and real infrared background images. Comparisons are conducted against traditional methods through three different scenarios. Quantitative evaluations are performed by using objective metrics， including information entropy， peak signal-to-noise ratio， standard deviation， average gradient and spatial frequency. Experimental results demonstrate average improvements by 5.82%， 1.03%， 4.24%， 10.5% and 33.58% of these metrics， respectively. The proposed method is proven to significantly outperform traditional approaches in high-frequency information reconstruction and detail preservation.

The generation of abnormal time-series data for weapon systems is one of the core challenges in the field of industrial intelligent operation and maintenance. Traditional equipment data generation method suffers from several limitations in abnormal data generation， including strong coupling in the latent space， violation of physical laws and a lack of diverse abnormal patterns. To address these issues， a causal disentangled VAE method is proposed in this paper， which involves latent variable independence through causal graph constraints， incorporates physical equations taken as prior knowledge in the decoder and a dynamic perturbation strategy employed to generate diverse abnormal data and realize equipment abnormal time-series datasets intelligentized generation. Experiments demonstrate that this method takes advantage of fairly good adaptability by single factor accuracy control， fit time dimension fault propagation and physical constraint controllability， and the CD-VAE significantly outperforms existing methods in terms of physical plausibility， interpretability and the training effectiveness of anomaly detection models on multiple cross-domain public benchmark datasets.

Regarding the geosynchronous orbit （GSO） multi-target flyby imaging services mission， a two-layer mission planning method based on a novel orbital configuration is proposed. Firstly， orbital configurations and maneuver strategies applied to existing research are analyzed， and a frozen-high elliptical orbit （F-HEO） is proposed for the mission. Based on J ₂ dynamics model， considering with temporal and positional consistency constraints， the perigee double-impulse maneuver strategy is studied. On the basis of this， the mathematical model for the inner-layer single-target maneuver planning and the outer-layer multi-target sequence planning are established. A dynamic neighborhood search （DNS） is adopted in the inner-layer to accelerate search， and the outer-layer employs a genetic algorithm to obtain the global optimum. Finally， a simulation case is implemented by considering with typical GSO targets.Results verify the feasibility， rationality and accuracy of the mission planning method and show that DNS can effectively reduce the time consumption of inner-layer optimization computations and the F-HEO takes advantage of lower fuel and time costs.