Aiming at the problem of error accumulation and low long-term positioning precision of the Strapdown inertial navigation system (SINS), an SINS/ADS error correction model is proposed by using the atmospheric data system (ADS) to correct the state amount of SINS, which is based on SINS, literature-based SINS/ADS and SINS/ADS based on error correction model. SINS performed navigation simulation analysis and the simulation results verified that SINS/ADS based on error correction model can effectively suppress error divergence, reduce navigation error and improve positioning precision.

Aiming at the space mission scenarios in the complex space environment, space mission planning and control methods of small satellite swarm system are proposed. The space mission modes and main process of the small satellite swarm system are studied. Swarm intelligent command and decision methods based on different modes are introduced. By optimizing the transfer orbit and mission configuration, the consumption of orbit transfer can be reduced while the effectiveness of detecting targets is improved. The swarm individuals are enabled to cooperate in space missions by introducing automatic coordinated control, which improve the accuracy of target detection and maximize performance. The simulation results show that the space mission planning and control method of the small satellite swarm system can achieve multiple rounds of precision detection on space targets. It can also solve the problem of small satellite cluster space coordinated operations, and provide the theoretical foundation and technical support for the development of larger-scale swarm design.

Aiming at the problems of weak adaptability and unstable penetration effect of programmed maneuvering penetration methods commonly used by high speed maneuvering aircraft, a maneuvering game guidance method based on deep reinforcement learning is proposed. By using this method, the encounter escape quantity is taken as the mission objective, deep neural network is used to fit the guidance law, reinforcement learning is used to train the network parameters and a intelligent maneuvering game guidance law is obtained with the position and velocity of both sides as the input and the demand overload of the aircraft as the output. The simulations show that the appropriate guidance instruction can be chosen by the aircraft according to the current situation in the continuous state space and action space. Compared with the traditional penetration methods, the encounter escape quantity can be significantly improved by using this guidance law, and its penetration effect is more stable.

In this paper, the fault modes of liquid rocket engine are reviewed, and the latest achievements of liquid rocket engine fault diagnosis technology at home and abroad are analyzed in recent years, including the fault diagnosis based on the models, the signal analysis and the artificial intelligence. The application progress and the diagnosis effect of the different kinds of fault diagnosis methods are compared, and the future development tendency of fault diagnosis methods for liquid rocket engine is prospected.

Firstly, the functional requirements and performance requirements of fly control software on rockets or missiles are analyzed. Based on the requirements results, multi-core distributed software architecture based on the MARS real-time operating system (RTOS) is presented. The architecture is divided into five layers that are known as RTOS layer, hardware excuse layer, frame layer, application layer, reuse component layer from top to bottom. This architecture is used in a couple of programs, which has good portability, high reliability and can segment in real time. It also can reduce development time and cost, and improve fly control software reliability and security.

A multi-channel and end-to-end attitude control method based on deep reinforcement learning is proposed for hypersonic morphing vehicle in the presence of situations by external disturbance and model uncertainty. Firstly, the attitude control model of hypersonic morphing vehicle is established. Secondly, the problem of vehicle attitude control is transformed into a Markov decision process. Furthermore, the training of the agent is implemented, which is based on the twin delayed deep deterministic policy gradient algorithm, and the end-to-end generated flight control instructions are deployed online. Finally, the proposed method's effectiveness and generalizability are confirmed through both basic performance and adaptive simulations.

By considering the attitude control system of the spacecraft with large flexible appendages, the fuzzy model predictive attitude control strategy for large flexible spacecraft is designed by combining adaptive fuzzy control with model predictive control. Based on the dynamic model of the large flexible spacecraft, a nonlinear model predictive control law is derived by using Taylor equation, which avoids the huge calculation burden in the online optimization process of predictive control. The computational complexity is effectively reduced. Aiming at the uncertain disturbance caused by large flexible appendages vibration, the adaptive fuzzy control is used to approach the uncertain disturbance. Based on the Lyapunov theory, the stability of the designed attitude control system is proven, and the adaptive law of fuzzy controller parameters is derived. The simulation results show that the designed attitude control strategy can effectively suppress the vibration of large flexible appendages, the expected value of the attitude angle of spacecraft can be reached quickly and accurately, which presents better control performance.

Based on the application of interactive sensing space teleoperation in space on-orbit service, the engineering cases of interactive teleoperation in space technology at home and abroad are firstly described, and the main technical components, core difficulties, current mainstream technology and research progress of interactive sensing space teleoperation robot are investigated. The necessity and urgency of interactive perception teleoperation are discussed, and the application of interactive perception teleoperation on-orbit service and the future research direction are discussed on the basis of that, and technical route based on the visual feedback, force feedback and local autonomous smart + auxiliary mechanical arm assembly+ path planning and motion planning is proposed.

In this paper, the environment security and autonomous collision avoidance method of huge constellation is studied in the view of orbital motion character. By focusing on the relative motion character, collision formulation derivation in forms of orbital elements is developed, which is based on spherical trigonometry. The collision risk caused by inter-constellation spacecraft and out-constellation spacecraft are analyzed. Regarding the different orbit-plane space targets with big relative velocity and short-time rendezvous-pass-by character, multiple rendezvous-pass-by observations and autonomous collision avoidance are discussed. By taking the collision between Cosmos-2251 and Iridium-33 for consideration, the relative motion character of potential collision-risk target is analyzed. Simulations are implemented by deploying the method of multiple rendezvous-pass-by observations and autonomous collision avoidance.

In order to improve the deficiency of transient performance and robustness of ADRC in the application of drag-free satellite, an improved form of nonlinear fal function, known as faln function, is proposed. This new function can enhance robustness of extended state observer and the performance of control system through the adjustment of gain in the large error section. The comparation and analysis of simulation result shows that the controller has the rapidest and stablest transient process by comparing with other drag-free control system. In a nutshell, the control performance of drag-free control system can be improved.

By starting from the model of a three-loop overload pilot, a unified open-loop transmission function, closed-loop transmission function and closed-loop characteristic equation of the system are derived, and then the relationship between the system control gain and the closed-loop pole, open-loop cut-off frequency is determined according to the closed-loop transmission function. The objective function of minimizing the equivalent time constant of the closed-loop system is given by introducing the rudder yaw rate constraint, the open-loop cut-off frequency constraint, the stability margin index constraint and the high-frequency component constraint. The typical parameter optimization paradigm under constraint conditions is formed. Through using the optimal solution of the system, the precise design of the open-loop cut-off frequency and phase margin of the system is realized, and an engineering design method of pole placement is formed. The effectiveness of the method is verified by an example.

Regarding the stringent requirements of information confidentiality review in the aerospace field, current manual screening methods are suffering from high costs and insufficient accuracy of keyword matching. An enhanced review framework integrated with large language models is proposed to improve the efficiency and accuracy of confidential information screening. Initially, the characteristics of confidential information are analyzed in the aerospace sector, an architecture that enhances the auditing performance of large language models is introduced in this study, which is combined with dynamic domain-specific expert system prompts to enhance the granularity and accuracy of reviews among multiple perspectives including technical and business confidentiality. By introducing a dynamic system prompt mechanism, the framework is effectively combined the semantic understanding capabilities of large language models with the real-time updating of keywords. Additionally, in order to prevent excessive auditing by the large language model, a hybrid cross-training strategy is developed, which significantly improves the recall rate of confidential information that reaches by 96%. Experiments on a self-developed high quality test set of 1000 entries demonstrates that the proposed method outperforms global open-source large language models by 18% in aerospace classified information inspection tasks.

Aiming at the attack-defense game between high speed aircraft and the interceptor, an improved DDQN is researched for high speed aircraft. The algorithm is aimed at the low utilization efficiency of sample in classical DDQN, by setting up multi- experience replay buffer,and combining accumulate Q-value temporal difference error (TD-error) with accumulate reward, the samples by fuzzy reasoning are classified and stored. Then, according to the training process, integral sampler and sampling form different experience replay buffer are designed The design principle of reward function is to reduce its mechanical energy consumption on the basis of successful penetration. The results show that the utilization efficiency of samples is improved by using this algorithm which provides a new idea to solve high speed aircraft maneuver penetration problem.

In order to improve the applicability and robustness of the fault detection method based on the principle of equivalent space in redundant strap-down inertial navigation system (SINS) fault detection, a t test-optimal parity vector technology (t-OPT) is proposed in this paper. The t-test for establishing a new fault detection function based on the optimal parity vector technology (OPT) is introduced by using this mehtod, and the Kalman filter algorithm is used to compensate the random noise of the fault detection function in the t-OPT method, so that the faulty device can be still accurately located, which is under the condition that the statistical characteristics of the parity residual are unknown, by applying this detection method, and there is noise interference and the fault amplitude is small. The simulation results show that the method can detect the constant drift of low fault amplitude in the redundant SINS, and effectively reduce the false alarm rate of the constant drift fault detection and the detection delay of the linear drift fault.

The ASIC chip architecture is adopted in the current GNSS receiver. After the design is completed, the algorithm and software are not easy to change, and the scalability is low and flexibility is poor. A new kind of architecture based on FPGA is proposed, also the hardware and software are designed based on it. A series of operations are completed, such as signal capture, tracking and calculation on the basis of FPGA in the receiver, which greatly improve the integration, expansibility and compatibility of the receiver. Moreover, aiming at the disadvantage of large altitude error caused by the principle of satellite positioning, a high-precision atmospheric pressure sensor is introduced into the receiver to correct the altitude information output by the receiver. The experimental results show that the receiver can reliably achieve the expected functions and has certain engineering practicability.

In order to solve the problems of less measurement data and insufficient fault samples for the same type of rocket control system,a digital twin-driven rocket control system health management framework is proposed in this paper. Firstly, the system-level analysis of the control system is implemented from aspects of the system composition level and the flight mission profile; on the basis of analyzing the typical failure modes of the control system, the functional requirements and scheme formulation of the health management system are proposed; the five-dimensional system composition and four-dimensional system composition of the control system are established. Model fusion is required to realize the model establishment of its digital twin; then, it proposes a six-dimensional system architecture for fault diagnosis driven by digital twin, and expounds its third-order operating mechanism from the control system operation and maintenance strategy entity verification, virtual model verification and virtual-real real-time mapping; this framework can be served as theoretical basis and model reference for digital twin fault diagnosis technology of typical stand-alone equipment of control system.

In order to cope with the problem of anti-jamming communication for UAV swarm, an intelligent fast frequency hopping algorithm is proposed with the goal of maximizing the information transmission rate and minimizing the frequency hopping overhead simultaneously. Firstly, based on the traditional Q-learning, a myopic value of perfect information is used to select the transmission channel, and then according to the observation information of the environment, the moment update method is adopted to correct the Q value based on the Gauss-Gama distribution model, which thus improves the performance of the fast frequency hopping strategy for UAV swarm. The simulation results show that compared with the random fast frequency hopping algorithm and the Q-learning based intelligent fast frequency hopping algorithm, the proposed scheme can learn fast frequency hopping strategy with faster convergence.

Regarding the transfer time in the Lambert rendezvous with a fixed speed increment, a Lambert rendezvous algorithm based on transverse eccentricity is used to solve the Lambert rendezvous problem by applying classical universal variable method, and the transfer time is expressed as a function of lateral eccentricity; The adaptive particle swarm optimization algorithm is used to optimize the transfer time in the orbit transfer, and the adaptive inertial weight and the adaptive variation mechanism are used to realize the Lambert rendezvous with the fixed speed increment in the shortest time. Finally, the simulation results show that the new algorithm has high solution efficiency under the condition of fixed speed increment, and the optimal transfer time is obtained, which can be applied to spacecraft space rendezvous.

In order to solve the problem that GNSS receiver on satellite is susceptible to single-event upsets and ground interference, the correcting single-event upsets and interference method of GNSS receiver on satellite is studied. Based on the on-orbit test data of satellite GNSS receiver, the cause and mechanism of satellite GNSS receiver anomaly are analyzed. The method is applied to a satellite-borne GNSS receiver, and the application results fairly show that the normal operation of GNSS receiver can be effectively guaranteed and the satellite-borne mission can not be affected.

Aiming at the characteristics of strongly nonlinear and large uncertainty occurring in the dynamics model of morphing aircraft, as well as the problem that the dynamics model changes widely, which is caused by deformation, a deep reinforcement learning attitude control method based on twin delayed deep deterministic policy gradient algorithm is proposed. First of all, the dynamics model of morphing aircraft based on the multi-rigid system is established. After that, the state space, action space and reward function are designed in the framework of Markov decision process. The control precision is further improved by introducing the history information of attitude tracking error into the state space. Furthermore, the strategy network is combined with the traditional PD controller to form a composite controller, which improves the algorithm training efficiency. Ultimately, the simulation results show that the control algorithm proposed is robust to the large uncertainty of morphing aircraft model and complex external disturbances during the deformation process, and also has strong adaptability for different deformation instructions.

Aiming at the problem of multi-target assignment of missile swarms under communication denied environments, a distributed multi-target assignment method is proposed. Through implementation of the consistency and auction stages, the target allocation scheme is optimized. Inspired by the swarm cooperative movement mechanism of social creatures such as flocks of birds and fishes in nature, the kinematic model of the missile swarm is established by using the separation alignment cohesion offense (SACO) swarm behavior rules and the swarm dynamic topology interaction mechanism. The missile swarm emerge macroscopic combat behaviors are being caused in different levels of communication denied environments. The simulation results show that the method designed in this paper can work out the target allocation decisions under different communication denial levels and has better optimization performance than the other methods.

Aiming at adapting to the digital and intelligent development of the electrical equipments of new generation launch vehicle, the multi-task design method of fly control software is introduced in this paper. The tasks are designed such as fly control task, data management task and correct cutoff task, and the corresponding execution priorities are set, and the synchronous operation method of fly control task in three redundant processors is proposed. The constraint requirement for multi-task sharing critical resource in one control cycle is proposed as a reliable margin index to control synchronization operation of fly control task in the three processors. The multi-tasks designed have been validated in flight experiments of the new medium launch vehicles.

An autonomous navigation method based on inertial measurement unit(IMU) is designed for Chinese geostationary orbit (GEO) satellites, which can be devoted to the satellite during the station keeping. The IMU is used to measure the attitude and acceleration information of the satellite, and a recursive equation of autonomous navigation during orbit control is designed based on the orbit dynamics equation. The simulation results show that this method can greatly improve the autonomous navigation precision of GEO satellite during Station Keeping and accordingly ensures the continuous business of GEO satellite. The method is very simple in architecture and reliable in performance, which shows the feasibility in actual project application.

Aiming at the path optimization problem when the formation of multi-quadrotor UAV remains unchanged, the multi-quadrotor UAV formation is regarded as a virtual rigid body based on the graph theory method in the flight process, and the UAVs in the formation are treated as the constraints of the leader quadrotor UAV. In the optimization process, only the maximum turning radius constraints, formation transformation time constraints and transformed constraints of the leader quadrotor UAV are considered, and the minimum energy required for formation flight is taken as the optimization goal, the track optimization problem of multi quadrotor UAV formation arriving at the target at fixed time is resolved. In this paper, the flight path optimization of the leader quadrotor UAV is implemented based on Gauss pseudo spectrum method, which is transformed into a nonlinear programming problem. The simulation results show that a smooth flight path can be obtained by using the algorithm proposed in this paper, which meets the constraints of timing arrival at the destination, multi UAV performance and environment; the obstacles can be safely bypassed by using obtained track which can improve the survivability of UAV formation and has certain engineering application value.

Regarding the attitude tracking control of spacecraft,a second-order mathematical model of attitude control system based on error quaternion is established. Based on modern control theory,a state feedback controller based on error quaternion is designed to configure the poles of the attitude control system and realize the attitude stable tracking of spacecraft. The simulation results show that the state feedback controller can achieve high precision attitude tracking control of micro spacecraft.

The UAV formation control method based on consistency algorithm and pilot following method is applied to study the multi-dimensional cooperation of time, space and attitude in the process of multi UAV formation flying mission. Based on the pilot following method, the UAV is divided into pilot aircraft and follower random, and the pilot aircraft and follower random together form a UAV formation. On this basis, the UAV kinematics model and formation controller based on consistency algorithm are designed. The UAV formation flight control system is deployed to control multiple UAVs for performing cooperative detection formation flight mission. Finally, three UAVs are used to carry out flight experiments, which verifies the effectiveness of the formation flight control method and provides support for the use of UAV simulation satellites to test spaceborne payload equipment.

A reliability analysis method of liquid rocket engine using a new adaptive Kriging based on voronoi-region-partition important sampling is proposed to solve the problem that the external load and its own attributes are uncertain and can lead to failure during the operation of liquid rocket engine. The method is based on global voronoi-region-partition, which is combined with improved leave-one-out method to select the most sensitive regions through active learning function selection of sample points and then iterative updating liquid rocket engine kriging model. The method can use fewer sample data to establish the mathematical model for liquid rocket engine to realize the failure probability of accurate prediction. The simulationresults show that the proposed method can effectively realize the reliability analysis of liquid rocket engine and reduce the calculation cost.

The stability analysis and parameters design of the adaptive augmentation control (AAC) framework of a launch vehicle are investigated in this paper. Firstly, a simplified stability criterion is proposed that the AAC controller can not destabilize the system if the PD controller is stable. Furthermore, an algorithm is proposed, which is based on the magnitude calculation principle to efficiently determine the parameters of the AAC controller. The proposed algorithms are then verified by the simulations, and the results show that the AAC controller is able to realize the attitude control of a flexible launch vehicle in complex mission environments.

In view of pressure change caused by air bag expansion during liftoff of non-conformal stratospheric airship, the launching phase of the airship is divided into free expansion rising phase and shaping rising phase. On the basis of this, the thermodynamic coupling model is established during the airship liftoff. The model can clearly be used to reflect the gas volume change and pressure change in the airship, which provides a model basis for studying the pressure difference control of the stratospheric airship. By considering that the internal states of airship can not be acquired in real time, a pressure difference control method based on fuzzy observer is proposed. In order to ensure the real-time learning ability, online sequential fuzzy extreme learning machine is used to train the fuzzy system parameters. The motion simulation of non-conformal stratospheric airship during its liftoff is firstly analyzed, and the necessity of differential pressure control for airship is expounded. The simulation results from different differential pressure targets show that the proposed differential pressure controller has good tracking ability. This achievement represents that it can serve as the important reference value for the design of airship differential pressure control.

By taking advantage of periodic phase modulation, PWM control and differential amplification technology, the positive-negative alternating and adjustable duty ratio signals required for liquid crystal membrane driving are generated. A reflectivity control device (RCD) based on liquid crystal membrane is designed and developed to continuously control sail reflectivity. The result of ground test shows that reflectivity control or simple on-off control can be achieved by using the developed RCD, and control flexibility and adaptability are both increased.

Regarding the controller design of the variant vehicle attitude control system, an integrated controller design method based on structure search and parameter optimization is proposed, which can realize the automation of the controller structure design work, searches for the most suitable controller structure and completes the parameter rectification in a shorter time. The simulation study on the attitude control of the pitch channel of a certain type of variant vehicle shows that the method can significantly improve the bandwidth of the system, and the number of iterations and the consumed time are both promoted by only about 38% than those of the traditional method.

Aiming at the problem of coordinated trajectory optimization between the boost and return sections of the RLV, a trajectory evaluation and decision-making program and a new optimized guidance law are proposed. Firstly, based on the numerical integration of ballistics and the mechanical energy, the carrying and return capability evaluation and decision-making plan for the thrust loss fault of the RLV booster stage are designed, and the trajectory decision interval with the best benefit is obtained. Then, PSO is used to optimize the ZEM/ZEV guidance law. After the thrust of the booster phase is reduced, the shutdown and landing time are optimized to promote the orbiting and return capabilities. After the payload and booster phases are separated, the secondary ignition height and the guidance law parameters are optimized to meet the maximum heat flow constraint. After the thrust is further reduced, the landing time and guidance parameters are finely optimized to achieve a precise soft landing. The simulation results show that the scheme proposed in this paper improves the RLV's orbiting and landing ability under the condition of engine thrust loss and the algorithm has benefits for simple structure and low burden computation.

Aiming at the large deviations of the thrust, mass flow rate and burnout time under different temperatures, a velocity increment based successive convexification method is proposed for solid rocket ascent guidance in this paper. The velocity increment is applied to this method as an independent variable for the dynamic model, and the successive convexification method is adopted by solving the ascent depleted shutdown guidance under multiple constraints. Moreover, the temperature conditions are recognized online by comparing the velocity increment for the purpose of parameters correction of interior ballistics. The precision and robustness of the successive convexification method based on the velocity increment are higher than that based on the flight time. Finally, the proposed guidance is verified through the numerical simulations of the ascent guidance for the solid rocket.

In order to solve the problem of too long optimization time of TVM compiler in the aerospace application process, a new cost model based on LightGBM algorithm is proposed to predict the running time of the code for the automatic search framework of TVM Ansor. The number of split points, the number of samples and the number of features are reduced by using the LightGBM algorithm, so that the new cost model can effectively reduce the amount of calculation, speeds up the training speed of the model without compromising accuracy and shortens the optimization time. The experimental results show that under the same optimization task, the optimization time and the overall reasoning time of Ansor with the new cost model can be improved by 1.6 times and 6.1% respectively.

As bad block management and data backup methods can not ensure data reliability for accessing NAND Flash data in missile borne information application, a set of highly reliable data management method for the NAND Flash whole life cycle is proposed: the new data management method is based on the bad block management by DSP software, and is combined with an improved BCH ECC algorithm based on FPGA. Moreover, the improved BCH ECC algorithm which uses PE Unit multiplexing to realize the real-time detection of checking 9 bit data error, checking and correcting 8 bit data error for reading NAND Flash data. By comparing with the traditional BCH ECC algorithm, the new algorithm not only can realize the data accuracy detection, but also can realize less FPGA logic resources and more fast detection speed. Thus, the new BCH ECC algorithm is very suitable for missile borne information application, which has high reliability and strong real-time performance. Simulation results and experimental results show that the reliability of NAND Flash data management can be improved by using this method.

In this paper, the problem of three-dimensional guidance for intercepting high velocity maneuvering target is studied and analyzed. Firstly, According to the head pursuit, the three-dimensional guidance system model is introduced. Secondly, an adaptive integral sliding mode guidance law is designed, which is based on the integral sliding mode and adaptive algorithm. Thirdly, an anti-saturation adaptive integral sliding mode based three dimensional guidance law is designed by introducing an auxiliary system. Finally, the stability of the designed guidance strategy is proved by using Lyapunov stability theory, and numerical simulations show the effectiveness of the proposed guidance scheme.

Aiming at the problems of single navigation source navigation information, easy interference and low reliability, a multi-source fusion positioning and navigation algorithm is designed in this paper, and a fusion navigation model of SINS/BDS/odometer/barometric altimeter/electronic compass is established. And in the process of multi-source navigation information fusion, the dynamic model may be easily perturbed and the measurement information is asynchronous and heterogeneous, a robust adaptive filter is designed to control the abnormal prediction information of the dynamic model and the abnormal measurement information.In addition, the judgement and processing methods of filter divergence in the process of information fusion are further studied, and the means of filtering correction are described. Finally, based on different information fusion models and filtering methods, a hardware-in-the-loop simulation experiment is implemented, and the filtering results are analyzed and summarized.The experimental results show that the multi-source fusion navigation algorithm designed in this paper can be integrated with the advantages of multi-sensors and can serve by providing high-precision and high-reliability positioning and navigation.

In order to determine the initial orbit of the pointed target after separation in space test, the impulse thrust maneuver model is proposed for approximating the target separating process in orbit, and the Gauss perturbation motion equation whose singularity has been eliminated by using the vernal equinox orbital elements, is established to quantitatively describe the relationship between the increment of instantaneous velocity caused by the thrust and the change of orbital elements. Under the main perturbation forces having a marked impact on the dynamic differential equation of the target and the consideration of the obit prediction precision and modeling complexity, the method of high-precision orbit prediction for the target is developed by each selected perturbation acceleration formula. Simulation results show that the relative error of the initial values of the target orbit compared with the data of a certain spacecraft simulation software is not more than 1.576%. The decay characteristics of the target orbit and the variation of the visible distance of the target relative to the satellite are analyzed, which have important value for reference to orbit design of the accompanying satellite, separation window scheme and optimal initial orbit determination for the target in space test.

In this paper, a nonlinear dynamic system of missile target relative motion is established, and the missile target interception problem is transformed into a nonlinear system differential game problem. Regarding the fact that the Hamilton-Jacobi-Isaacs (HJI) equation is difficult to find analytical solution, the policy iteration (PI) algorithm is proposed to establish an online adaptive dynamic programming scheme. The policy evaluation and policy improvement in PI algorithm are realized by critic neural network and actor neural network. In the policy evaluation stage, the weights of the neural network are updated through online data sample, therefore the implementation of the algorithm does not use the explicit knowledge of the internal dynamic of nonlinear system. By repeating the two steps of policy evaluation and policy improvement, the approximate optimal solution of the missile target interception differential game problem can be obtained. This method is applied to the missile target interception system, and the effectiveness of the guidance law is verified through numerical simulation.

In order to solve the problem of low tracking precision of a single observer for a near space vehicle over long distance, a multi-observer cooperative tracking method is applied. In order to accurately fit the motion state information of high maneuvering target in near space, an interactive multi-model quadratic filtering method is proposed. Based on the target tracking information, a differential auto regressive integrated moving average (ARIMA) model is used to predict the target trajectory. The simulation results show that the tracking and prediction precision of the target motion information is effectively improved by using the quadratic filtering trajectory tracking and prediction method, compared with the traditional tracking and prediction methods. The angular velocity tracking error is reduced from 30% to less than 5%, and the velocity tracking error is reduced from 20% to less than 2%. Moreover, it can effectively predict the trajectory of the target in the future 100s without a lot of prior information, and the error is less than 0.5km.