Under the context of the explosion of the traffic data in 5G network, two software-defined networking-driven (SDN) routing searching optimization algorithms are proposed and evaluated to solve the traffic load balancing problems of the traffic data in 5G network. Firstly, a multi-constraints data transmission routing selection model of SDN is established. Secondly, a traffic load balancing scheme with breadth-first-search(LBB) routing optimization algorithm is proposed to solve the model. In the process of the breadth-first-search, LBB sets a dynamic traffic threshold to monitor the link in real time, which aims to find the optimal data transmission routing from the source node to the target node. To further reduce or even avoid the space overhead caused by the unnecessary searching, an iterative deepening search with depth-first-search(IDDFS) routing optimization algorithm is proposed, which selects the link with the highest available bandwidth in the searching process. During the routing searching, IDDFS optimizes the iterative optimization with the policy of depth-first-search. Simulation results show that the proposed algorithms have the excellent performance in resource utilization and network throughput.

In response to the problem of poor robustness of traditional visual odometers that only rely on point features, a monocular visual inertial odometry that utilize point features, line features and structural features are deduced. The structural feature is an orthonormal basis that satisfy the Manhattan world hypothesis. This feature contains the parallel and orthogonal information of the structural object. The introduction of structural feature provide an overall constrain for visual inertial odometry and can reduce the accumulative error of rotation effectively. Meanwhile, to ensure the real-time performance of the system after the introduction of structural features, a robust and efficient algorithm is proposed to extract structural feature, which is based on the matching relationship of line features and the direction of gravity vector. The experimental results demonstrate that, compared with the existing open source systems, the proposed system can effectively reduce the accumulated rotation error while maintaining the real-time performance.

In the genetic programming algorithm, population diversity plays an important role in avoiding premature convergence. Improving algorithms by controlling population diversity is a hot spot in genetic programming. Therefore, this paper improves the selection mechanism of the algorithm from the perspective of diversity, and proposes a genetic programming algorithm based on clustering tournament mechanism and parent generation matching. The algorithm divides the population into multiple sub populations through clustering so as to adjust the selection pressure of the algorithm to maintain population diversity and to improve the search ability of the algorithm. In addition, the algorithm extracts individual binary features and uses local matching to cross operate the parent generation. From the perspective of parent pair diversity, the algorithm achieves a better balance between the exploration and the exploitation. Multiple comparative experiments on different benchmark problems verify that the population diversity of the proposed algorithm is greatly improved and the optimization ability and convergence speed are better improved.

A deep learning-based impact time and angle control guidance algorithm is proposed to solve the impact time and angle control problem of missile guidance under aerodynamic force. A predictor module is designed for estimating the exact impact time under optimal impact angle guidance law with realistic aerodynamic characteristics, and a feedforward loop for fusing the theoretical model and deep learning method is presented. An approximate time-to-go is predicted by the theoretical model and the prediction error is estimated by deep neural networks, so that the accuracy of prediction is improved. The impact time error could converge to zero after introducing the exact impact time into guidance law, then the joint control of the impact time and the terminal impact angle is achieved. The simulation results show that the designed guidance algorithm can realize the impact time and angle control more accurately compared with that based on a constant velocity assumption.

The problem of serious speckle noise and poor visibility in polarimetric synthetic aperture radar (PolSAR) directly affect the accuracy of target recognition. A ship recognition algorithm based on the multi-level cooperative fusion of multi-source remote sensing images is proposed. The method of multi-level cooperative fusion is adopted to enrich the image features and improve the accuracy of ship recognition. Firstly, the multi-source remote sensing data is fused at the pixel level. Then, the feature-level fusion is carried out on the basis of the previous step. Finally, new target features are obtained. This method gives full play to the information complementarity advantage of PolSAR and multispectral images in different frequency bands. This method retains the polarization scattering characteristics of the target in different frequency bands of PolSAR. Meantime, the spectral-spatial information of multispectral data is also retained. Compared with the traditional single remote sensing data, the proposed method performs better in visibility and detection accuracy. The recognition accuracy of the proposed method is improved by 5.12% at least.

The field calibration of strapdown inertial navigation system usually uses the system-level calibration method. Since the state variables which have small observability degrees cannot identify the estimation error parameters by Kalman filter, a calculation method which is observability degree based on the orthogonal basis of state quantity is proposed, and a transposition scheme to improve the observability of all state variables is designed according to this method. In the process of rotation, the velocity error of the static base inertial navigation is measured at each time, and the Kalman filter is used to identify the error parameters of all quantities of state. The experimental results show that compared with the field calibration results without transposition scheme, the field calibration method can not only improve the observability degrees of the all quantities of state, but also improve the results of system-level calibration, which has good engineering practicability.

Although there are some traditional cache placement strategies in named data network (NDN), most of them cannot utilize the limited cache resources in the mobile ad hoc network efficiently. In order to improve the cache space utilization rate and reduce the data transmission time delay, an assisted edge cache strategy based on content popularity is proposed. In the mobile ad hoc network, NDN link state routing protocol is used for routing in mobile ad hoc network. Each of the node in the network decides whether to cache data independently, and all the popularity of the data, the location information of the node and the cache flag information are used to make the caching decision. As a result, data packets with higher popularity are pushed to edge nodes which are closer to users. The simulation experimental results show that, compared with the traditional cache placement strategy, this cache strategy has made progress in cache hit rate and data request time delay.

Missile-borne bistatic forward-looking synthetic aperture radar (MBFL-SAR), can achieve efficiently two-dimensional high-resolution imaging during the whole terminal diving period of the receive platform. Compared to the airborne/spaceborne bistatic configuration, the missile-missile bistatic configuration has advantages of not easily attacked, the trajectory is flexible and the synthetic aperture time is short. However, due to the missile's strong maneuverability and high requirements for imaging time, the traditional airborne/spaceborne bistatic configuration design method is not suitable for MBFL-SAR. Based on the geometric model of MBFL-SAR imaging, this paper first analyzes the resolution calculation method based on gradient method. Then, combined with the characteristics of the two-dimensional resolution, the influence of the configuration parameters on the two-dimensional resolution is analyzed, and the configuration design method by reducing the dimensions is proposed, while improving the efficiency of configuration parameter design, it can effectively shorten the imaging time. Finally, the effectiveness of the proposed method is verified by simulation experiments based on typical working parameters.

A method for analyzing the anti-damage network of launch platforms based on damage flow infor-mation transmission is proposed to address the relative importance weighting of each subsystem in the evaluation of anti-damage capability of launch platforms. Taking a certain launch platform as the research object, this paper analyze the network topology properties by constructing its anti-damage network topology model. A mutual information model is proposed to calculate the weights of key components based on the functional operation characteristics of the launch platform, and a method for characterizing the anti-damage ability of the launch platform based on node connectivity is proposed. The research results indicate that the anti-damage analysis method based on complex network comprehensively considers the relationship of the overall layout and structural functional, making the evaluation results more systematic and comprehensive. The use of the mutual information model can objectively calculate the importance of key nodes in the network, reducing the impact of subjective weighting. The use of node connectivity can characterize the difficulty of transmitting damage flow information in the transmitting platform. The research results can provide reference for the evaluation and optimization of the anti-damage ability of launch platforms.

When deep learning methods are used for specific emitter identification, the existing algorithms are insufficient under the low signal to noise ratio. Meanwhile, they all focus on the inter-class distance but ignore the intra-class compactness. To solve this problem, a residual prototype network is proposed to recognize the differential constellation trace figure of input signals by combining the residual network and prototype learning. In addition, prototype loss is combinedwith the distance-based cross entropy loss to further amplify the inter-class distanceby improving the intra-class compactness. The results show that the proposed algorithm has better recognition performance under the same signal to noise ratio condition through experiments on five ZigBee devices. And the accuracy can reach more than 99% when the signal to noise ratio is higher than 8 dB.

An absolute decoupled robust sliding mode observer design method for fault reconstruction is proposed, which is used for a class of nonlinear system with multiple faults(actuator faults and sensor faults existing simultaneously) and measurement noise. Firstly, the sensor faults and measurement noise are taken as the parts of augmented state vector to build the descriptor system, and the robust descriptor sliding mode observer is designed. Then considering the case that Lipschitz coefficient and upper bound of faults are unknown, the design algorithm with the adaptive law is proposed. Besides, the auxiliary matrix and descriptor constrained inverse matrix are introduced to design the coefficient matrix of observer, which is used for decoupling the faults from disturbances. Meanwhile, the gain matrix design method of observer is proposed through linear matrix inequality. Finally, the sliding mode differentiator is provided, and the actuator and sensor fault reconstructed results are given. The simulation is given to demonstrate the effectiveness of the proposed method.

In view of the problem of network damage with limited resources when the attack cost is equal, the definition of network damage maximization is given. In order to improve the defect that the approximate algorithm suffers high computation complexity when solving network damage maximization problem, the algorithm based on topology potential and cost-effective lazy-forward (TPCELF) is proposed. Experiments with simulated scale-free networks and real network show that the TPCELF algorithm has a greater improvement in calculation speed, and the average damage effect of the network is close to the approximate algorithm. What's more, the TPCELF algorithm is better than other commonly used importance metric ranking algorithms in average network damage effect. The proposed approach can provide a reference for mining key nodes in complex networks from the perspective of network damage.

Inter-satellite links play a very important role in satellite network data transmission, which can solve the problem of limited layout of ground stations in China. However, the satellite network topology is time-varying, and the inter-satellite links have intermittent connectivity, which make inter-satellite data transmission a great challenge. In order to overcome this difficulty, firstly, the satellite network is modeled by using the storage time aggregation graph. Under the condition of considering the resource constraints on the satellite, an integer programming model of data transmission is constructed. Then, a knowledge-guided hybrid evolutionary algorithm (KGHEA), which integrated local search algorithm, path traffic allocation algorithm and a variety of knowledge operators, is designed to solve the model. Finally, a simulation experiment is designed to verify the performance of the KGHEA algorithm, and the impact of various parameters on the data transmission performance is analyzed to provide a reference for the construction of inter-satellite network.

The modern air combat environment tends to be complicated, and it is difficult for traditional methods to obtain the ideal guidance effects for maneuvering targets. Therefore, an air combat guidance method based on top-level rolling optimization and bottom-level tracking is proposed for the modern unmanned aerial vehicle (UAV) air combat guidance problem. Among them, the top-level optimization based on the Gaussian pseudo-spectral method uses time as the performance index to quickly guide the UAV to the target area. When the target maneuvers, its trajectory is predicted and the guidance trajectory re-optimized. In addition, the underlying tracking method for the guide line of sight angle is introduced, so as to complete the rapid correction after the target yaw. Finally, the simulation of rolling optimal guidance and line of sight angle tracking in the air combat environment is carried out for a certain type of combat UAV, which verifies the interception ability of uncertain maneuvering targets of the proposed method.

The grey correlation analysis method is widely used in the field of simulation model validation due to its advantages such as simple calculation and not requiring large sample size. However, the Deng correlation degree model has the drawback of being unable to determine the sequence distance, which can easily lead to misjudgment of simulation validation results. In order to improve the accuracy of using grey correlation analysis to verify the reliability of simulation results, a distance analysis method is proposed to determine the distance between data sequences. When verifying the consistency of simulation results for aerospace product performance prototypes, the distance analysis method is combined with the grey correlation analysis method to form a grey correlation analysis method that considers sequence distance. It can judge the consistency of two sets of sequences from both the shape and position of the curve, improving the accuracy and reliability of the verification results.

Aiming at the problem that the spectral sensing algorithm based on eigenvalue has poor sensing performance in the environment of impulse noise. All eigenvalues of the matrix are analyzed and the geometric mean of the eigenvalues of the matrix is introduced. A spectrum sensing algorithm based on the difference between maximum eigenvalue and geometric mean of eigenvalue (DMGM) of the fractional low-order covariance matrix is proposed. Alpha stable distribution noise is selected to simulate the impulse noise environment. Theoretical analysis and simulation results show that DMGM has better perceptual performance than other algorithms in low signal to noise ratio environments, and has better perceptual performance under low signal to noise ratio conditions.

The amplitude interpolation azimuth estimation method can obtain direction of arrival (DOA) by sampling and interpolating a few beams, which avoids intensive DOA search and is widely used in engineering. At present, most of the methods are based on one dimension, and use symmetric function for interpolation, which leads to high estimation error. In this paper, the method is extended from one-dimension to two-dimension arbitrary array by decoupling, and considering the asymmetry of left and right beams when the DOA deviates from 0°, the asymmetric function is used for amplitude interpolation with the solution condition of the slope ratio of left and right beams. Furthermore, in order to solve the problem of angle mismatch in two-dimensional asymmetric interpolation, a method of interative interpolation is proposed. Finally, the effects of the interpolation method, angle mismatch and signal to noise ratio (SNR) on the interpolation accuracy are simulated, the interpolation accuracy of the proposed method is better than that of the traditional symmetric interpolation method. Through the experimental data of Qiandao Lake, the estimation accuracy obtained by the proposed method with 10° interval interpolation can be compared with that obtained by conventional beamforming with 0.01° intensive scanning. The method presented can meet the real-time requirements of some engineering applications and achieve high accuracy of azimuth estimation at the same time.

When the deception jamming is implemented on synthetic aperture radar (SAR), the jammer usually needs to obtain some key parameters in advance, such as the speed of carrier platform of SAR, the slow time of azimuth, the nearest oblique distance from the jammer to the carrier platform of SAR, and the self-position of the jammer, etc. However, the acquisition of these key parameters above usually needs complex reconnaissance equipments. In response to this problem, a passive direction-finding and SAR deception jamming method based on multi-receiver coondination is proposed in this paper. This method uses the arrangement of jammers and receivers and the time difference of arrival (TDOA) information between the receivers, and effectively solves the problem of the requirement for complex equipment to reconnoitre the above key parameters, and greatly simplifies the configuration of the interference system. According to the TDOA information, the pitch angle and horizontal azimuth of jammer relative to the carrier platform of SAR can be calculated, so as to realize the passive direction-finding of carrier platform of SAR. On this basis, a deception jamming algorithm against SAR is proposed and compared with the existing deception jamming algorithm against SAR based on multi-receiver coondination. The result shows that, with the increase of distance between the generated false target and the jammer, the false target focusing effect and real-time performance of the proposed method are better than that of the existing methods.

In a pulse, jammers precisely replicate and rapidly forward to form interrupted-sampling repeater jamming (ISRJ), which seriously affects the performance of radar. Aiming at this problem, this paper starts from waveform design and filtering and then proposes a method of frequency agility waveform combined time-frequency filter to suppress ISRJ. Firstly, the conventional linear frequency modulation (LFM) signal is divided into several segments and randomly rearranged to construct the intra pulse frequency agility waveform. Secondly, the time domain projection component is obtained by time-frequency analysis of the echo signal, and the optimal segmentation threshold is calculated by OTSU algorithm to extract the undisturbed signal segment. Finally, a filter is constructed by convolution operation to perform band-pass filtering on pulse pressure output to preserve the real target and achieve the interference suppression. Simulation results show that the proposed method can effectively suppress the ISRJ under the condition of high jamming power and low signal-to-jamming-and-noise ratio, which greatly improves the radar anti-jamming performance.

Aiming at the problem of data governance caused by the complexity of data in the field of equipment testing and indentification, the technical idea of assisting relevant test and indentification institutions to carry out data governance by constructing knowledge graph is put forward. Based on the analysis of the characteristics of test identification tasks and domain data, an ontology construction method with universality in the field of testing and identification is designed on the basis of Stanford ontology construction process. The knowledge graph ontology model constructed based on this method has clear business pertinence and system expansion. Finally, through typical examples, the knowledge graph constructed by this method is verified to support the data governance in the field of equipment testing and indentification.

Aiming at the problem of anti-jamming performance degradation caused by time-varying statistical characteristics of linear frequency modulated (LFM) interference in satellite navigation application, a space-frequency adaptive processing (SFAP) algorithm based on data grouping by space-time-frequency three-dimensional feature grouping is proposed. Firstly, the time-frequency analysis method is used to obtain the joint distribution of sampling data in time domain and frequency domain, and the spatial correlation coefficient is used to analyze the spatial correlation of the same frequency interference at different times. Secondly, the sampling data of SFAP are grouped, and the sampling points with the same frequency and direction of arrival (DOA) parameters at different times are divided into the same group. Finally, the covariance matrix estimation, weight calculation, and adaptive filtering are carried out by using the grouped data to improve the interference eigenvalue value, the null depth, and the anti-jamming ability. Simulation results show that the algorithm can effectively improve the suppression ability of satellite navigation receivers'LFM interference, and can be applied to the scene both with single LFM interference or multiple LFM interference.

The frame-inconsistent problem exists in conventional error equation of strapdown inertial navigation system (SINS). In this study, a more rigorous error equation of SINS attitude, velocity and position is derived by projecting the velocity error in the real navigation frame. Based on the newly defined velocity error, a modified Kalman filter model, including state model and measurement model, is constructed for Doppler velocity logger (DVL) /SINS integrated navigation. Compared with conventional Kalman model for DVL/SINS, simulations and field experiments indicate that the proposed algorithm is efficient in solving the frame-inconsistent problem in conventional DVL/SINS system, which can improve the navigation accuracy for the integrated navigation system.

In light of the high dynamic property and the complex running environment of the Internet of Things, this paper sets up a store-carry-forward information transmission mode based on the opportunistic network. Then, it proposes a theoretical model to evaluate the performance of the information transmission process with uncertainty. According to the model, considering both the energy consumption and transmission performance, this paper proposes the optimal transmission policy by the Pontryagin's maximum principle, and proves that the policy conforms to the threshold structure. Finally, several experiments signify that the model is accurate with the error below 4.08%. At the same time, the experiments show that the optimal policy is better than the traditional static policy, and gets better with the increasing of the information's lifetime.

Pythagorean fuzzy set expands the information representation content of intuitionistic fuzzy sets and has the advantage of reflecting larger information space. Constructing the distance measure of Pythagorean fuzzy sets is the core link of implementing Pythagorean fuzzy decision-making. Firstly, this paper deeply analyzes the defects of the existing four distance measures of Pythagorean fuzzy sets, establishes a spatial information representation model of Pythagorean fuzzy sets integrating the membership degree and nonmembership degree, hesitation degree and strength of commitment, and then puts forward a novel distance measure based on the spatial centroid of Pythagorean fuzzy sets. Secondly, the specific properties of the proposed distance measure are extended, and the advantages of this distance measure model in reflecting hesitation and solving counterintuitive situations are proven and compared. Finally, the application research of pattern recognition and medical diagnosis verifies the accuracy and superiority of this distance measure between Pythagorean fuzzy sets.

Estimating camera pose in a scene with known 3D information is important in several fields such as autonomous driving, augmented reality and virtual reality. There are many methods to directly return to the camera pose from the input image, and then calculate the camera pose through the 3D coordinates of the returning pixels. However, the problem of these methods is that they are seriously coupled with the training scene and lack the generalization ability in the new environment. The convolutional neural network should focus on learning the robust and invariant visual features. Therefore, a direct alignment optimal method is proposed based on multi-scale features. It takes the feature similarity as the measurement form and the camera pose as the optimization quantity, and estimates the accurate 6 degree of freedom (6DOF) pose of the camera through end-to-end training from pixel to pose. The system separates the model parameters from the scene, has strong generalization ability for new scenes, and has good pose location accuracy.

In view of the integration ossification problem, traditional airborne networks have difficulty in allocating network resources to recover quickly from node failure and complete corresponding missions. Therefore, a node reliability-aware protection-differentiated virtual airborne network embedding (NRPD-VANE) algorithm is proposed for airborne networks in the wireless network virtualization environment. Firstly, node mapping adopts a novel node ranking method that comprehensively considers node failure possibility, wireless interference and network resources in order to map virtual nodes onto reliable substrate nodes. Secondly, link mapping adopts the efficient P-Cycle protection technology to protect the unreliable substrate nodes on the path that maps corresponding virtual links. Simulation results show that comparing with other latest node protection embedding algorithms, the proposed algorithm improves the embedding success ratio while maintaining a low average recovery delay.

Aiming at the needs of wind turbine clutter (WTC) suppression in passive radar, the modulation model of the wind turbine acting on the radiation signal is built firstly. Secondly, the similar characteristics between the modulation effect of the wind turbine blade and the periodic structure of the radar emission signal on the range Doppler (RD) spectrum is studied. Thirdly, based on the perspective of periodic component suppression, this paper proposes a new method to suppress the extended Doppler clutter in matched filter processing with the modified reference signal. Finally, simulation test and experimental data are used to analysis verify the proposed algorithm along mith measured data of wind turbine based on passive radar. It shows that this method can improve the effect of WTC suppression, and it has high computational efficiency and less resource occupation. In addition, this method has been verified with the experiment data of self-developed digital television based on passive radar.

A synergistic guidance method is proposed for interceptor missiles in the later part of midcourse. For cooperative detection, time-to-go, range-to-go, terminal velocity and terminal angle should be controlled independently. The semi-analytic prediction method is used to predict the velocity, needed overload and supplied overload; convert the required overload into the lift coefficient requirements. The cooperative trajectory planning problem is transformed into a nonlinear optimization problem through the selection of trajectory shaping variables, and the reference trajectory is generated online. To improve the particle swarm optimization algorithm, adaptive inertia weight selection and invalid particle substitution strategy were adopted, which can improve the probability of the population and escaping from the local optimal point. Trajectory that meets the terminal constrains can be planning fast on consider of the flight dispersion and trajectory deviation. Finally the effectiveness is verified based on mathematical simulation.

The performance of the sparse recovery-based parameter estimation methods for moving target in space-time adaptive processing (STAP) degrades significantly in the case of dictionary mismatch. A space-time moving target parameter estimation based on non-convex relaxation of atomic norm is proposed. According to continuous compressed sensing and low rank matrix recovery theorems, the estimation of azimuth and velocity for moving target are obtained with high accuracy and super resolution, which utilizes the intrinsically sparse characteristic of moving target echo in the angle-Doppler domain. The proposed method can avoid the dictionary mismatch problem in the sparse recovery, and improves the performance of parameter estimation effectively. Numerical results show that compared with the existing grid-based approaches and atomic norm minimization approach, this approach has higher estimation accuracy and allows closely spaced targets to be more easily distinguished.

The Gaussian-mixture-model-based model predictive controller is proposed for the precise operation requirement and task-space control problem of space robots. Based on the nominal model, the Gaussian mixture model is utilized to analyze and compensate the model uncertainties accurately and efficiently, which are caused by the joint friction, measurement error, etc. Then, considering the physical constraints, such as joint limitations and input saturations, the nonlinear model predictive control method incorporated with the augmented model is proposed to realize the direct and accurate tracking for both the robot base and end-effectors pose. Besides, the thrust allocation algorithm is presented for the thruster's redundant configuration. Finally, the effectiveness of the proposed method is verified by the simulation results.

In order to improve the scheduling efficiency of material distribution vehicles in naval aviation stations, an optimization model for material distribution vehicle scheduling is established based on the characteristics of material distribution tasks in naval aviation stations, and a hybrid genetic algorithm (HGA) is proposed to solve the model. The HGA introduces the simulated annealing (SA) algorithm operation to improve the classical genetic algorithm (GA): choosing the coding method and crossover operator suitable for the model; using a method similar to path construction to construct the initial population; after the genetic operation generating the sub-population, the SA operation is used to find the potential outstanding individuals in the sub-population neighbourhood to improve the local search ability of the algorithm. Finally, the effectiveness and reliability of the proposed algorithm are verified by comparing the experiments with the classical GA.

A composite platform high precision control method composed of satellite platform and two-dimensional turntable is proposed for space moving target high precision attitude tracking. Firstly, the coupled dynamics model of the composite platform is established. Secondly, the backstepping controller is designed for the satellite to achieve coarse tracking. When the attitude error meets certain switching requirement, the two-dimensional turntable adopts the model predictive controller based on the loading observer to assist the satellite to perform the attitude precision tracking. In addition, the nonlinear disturbance observer is designed to estimate the disturbance torque generated by the coupled motion on the satellite body. Finally, the numerical simulation results show that the attitude accuracy of the proposed method can be improved by an order of magnitude, which can realize high-precision attitude tracking control, and provide research basis for engineering practice.

In the prediction of rolling bearing degradation trend, traditional feature selection mainly relies on manual experience and a single evaluation algorithm, which is easy to cause under-selection or misselection of features. Moreover, a single deep learning network cannot fully mine the performance degradation information contained in the data, which results in low prediction accuracy of the model. To solve the above problem, a prediction method of rolling bearing degradation trend based on adaptive Mahalanobis space (AMS) and fusion deep learning network is proposed. Firstly, the original signals are decomposed and the correlated kurtosis coefficient criterion is used to screen the intrinsic mode function (IMF) to reconstruct the new signals, and the features are extracted from the multi-domain perspective. Secondly, the multi-objective feature selection algorithm based on AMS is built to optimize characteristic automatically. With the aim of reducing manual dependencies, and serengthening the adaptability and generalization, Mahalanobis distance (MD) is combined with the exponential weighted moving average (EWMA) method to well characterize the degradation trend of bearing performance. Finally, the fusion model of sparse auto encoder and gated recurrent unit (SAE-GRU) is used for prediction. The experimental results show that the proposed method can effectively screen the optimal features and significantly improve the prediction accuracy.

The detection, tracking and recognition performance of monopulse radar will be severely degraded due to the mainlobe cover jamming. How to estimate the target angle in this scene is a classic problem faced by monopulse radar. In order to solve this problem, a target angle estimation algorithm for monopulse radar is proposed. The algorithm uses residual matrix F-norm minimization criterion to extract target echo signal and estimate the direction of the target echo and jamming. The simulation results show that the proposed algorithm is more accurate and stable than the mainlobe jamming cancellation (MJC) algorithm.

In order to solve the problem of rendezvous control for manned/unmanned aerial vehicle (MAV/UAV) formation, a rendezvous control strategy for MAV/UAV formation based on path planning-following is designed. Firstly, considering formation rendezvous boundary constraint and anti-collision constrain conditions, a formation rendezvous path planning algorithm based on Dubins curve combined with cooperative simulated annealing particle swarm optimization (CSAPSO) is designed. Then, the rendezvous path generated is set as the desired path, a nonlinear path following control method using dual-loop structure is designed to track the desired path. Lastly, the simulation experiments of path planning and following control method are carried out by Matlab and results demonstrate the effectiveness and superiority of the proposed methods.

To obtain the effectiveness evaluation results of weapon equipment system quickly and accurately, a method is proposed that can simplify the effectiveness evaluation index of weapon equipment systems. Based on the importance of evaluation index measured by random forest (RF) algorithm, the importance of evaluation index can be sorted in the case of missing index value, and then the specified number of important evaluation index can be refined by improving the decision tree partition attribute selection, leaf node value determination, and decision tree output value calculation methods in RF algorithm. Based on the simulation data set with missing index value, it is verified that the improved RF algorithm is more accurate than the traditional RF algorithm in the application scenario of air defense system effectiveness evaluation index simplification.

In order to ensure the safe operation of unmanned aerial vehicle (UAV) in the low-altitude airspace, this paper proposes a demarcation method of safety separation for multi-rotor UAV in free airspace in combination with the characteristics of multi-rotor UAV. Firstly, the smallest circumscribed sphere of the UAV body cylinder is used as the collision box of the UAV. A UAV mid-air collision probability model is established based on the position information, speed information, position error information and three-dimensional Gaussian distribution assumption. Secondly, according to the characteristics of free airspace, a demarcation method of safety separation based on the maximum collision probability is proposed. Taking the Matrice 600 Pro UAV as an example, when the safety target level is 5% of the maximum collision probability, the longitudinal, lateral, and vertical safety separations are 16 m, 11 m and 6 m respectively according to the proposed method. The proposed method can provide a theoretical basis for the formulation of safety separation standards for multi-rotor UAV.

Aiming at the problem of extracting different levels of roads in synthetic aperture radar (SAR) images, an effective multi-scale extraction method for SAR roads from coarse to fine is proposed. Based on Duda operator, the method improves the extraction probability of different levels of roads and increases the link probability of potential road segments through multi-scale extraction and results, integration of road in SAR images. On this basis, the effective extraction of roads in SAR images is realized by combing the hierarchical design of scale filtering, morphological filtering, and fragment linking. The proposed method is tested on real SAR images, and the effectiveness of the method is verified by continuous and complete extraction results of road in SAR images.