In air-ground collaborative operation scenarios,UAVs are required to have the capabilities of rapid path planning and accurate autonomous landing at the same time,which poses higher requirements on the real-time performance and accuracy of the algorithm under limited computing resources of the UAVs.This paper mainly solves the key technological problem of integrated realization of UAV rapid path planning and autonomous landing function.Firstly,to solve the problems of the traditional A* algorithm of bad real-time performance and abundant resource consumption,the traditional eight-neighborhood search approach is expanded to the sixteen-direction search approach,a high-matching-degree heuristic function is designed,and an adaptive weight coefficient is introduced to comprehensively enhance the performance of the path planning algorithm.Secondly,oriented to the scenario of air-ground collaborative operations,a UAV autonomous landing strategy based on visual feedback is designed,which improves the accuracy of autonomous landing.Finally,simulation and experiments are performed to verify the performance of the proposed enhanced A* algorithm,autonomous landing and the integrated realization strategy.The results show that the enhanced A* algorithm shortens the length of the planned path and improves the smoothness of the planned path,and the real-time performance of the algorithm is improved to a great extent,which meets the requirements of real-time obstacle avoidance and path planning in complex environments.Meanwhile,the UAV can quickly and steadily land at the target point on the ground platform.

To improve the tracking performance in complex scenarios,an object tracking algorithm with feature enhancement and dual-template updating is proposed.Firstly,an improved feature extraction network is proposed,and deep-layer features and shallow-layer features are fused,and then the spatial-channel attention module is utilized to enhance the fused features,so that better capability of feature representation is obtained.Secondly,a dual-template updating strategy is proposed,and the adjacent high-confidence frame of the image is reserved as the backup template,which is utilized to implement weighted fusion with the initial template to obtain new templates for conducting tracking prediction again when the confidence of the tracking response map is low.Finally,the tracking performance is evaluated on the datasets of OTB-100 and VOT-2017,and the experimental results show that the proposed algorithm improves the tracking success rate and accuracy in complex scenarios such as occlusion,illumination change and background clutter.

Finite-time distributed cooperative control of a second-order multi-agent system in the form of double integral with unknown leader control inputs is studied.Firstly,in order to estimate the leaders control input information quickly and accurately,a new finite-time tracking differentiator based on hyperbolic tangent function is proposed,and a terminal attractor function is introduced to eliminate chattering.Secondly,a finite-time tracking control law based on a novel variable speed reaching law with deceleration points is designed.The system has faster convergence speed when it is far from the deceleration point,and can effectively reduce the chattering when it is close to the deceleration point.Under the action of this control law,the followers can track the leaders speed and position better by using the obtained control input information.Finally,the simulation results verify the effectiveness of the proposed algorithm.

Traditional algorithms rely on the precise separation and information extraction of infrared ship targets and environmental backgrounds,and it is difficult to meet the needs of ship target detection under complex background and environments with noise and other interference.To solve the problem,an infrared ship target detection algorithm based on the improved YOLOv5 is proposed.The Reasoning layer is added to YOLOv5 network,and a new architecture is used to extract the semantic relationship between image regions to predict the bounding box and class probability,which improves model detection accuracy.At the same time,the loss function of the YOLOv5 target detection network is improved to further improve the detection accuracy of the model.The verification results show that the model trained by the improved YOLOv5 algorithm has significantly improved detection accuracy and speed in comparison with several target detection algorithms listed in the experiment.The mean Average Precision (mAP) can reach 94.65%.The model has been verified that it can meet the real-time requirements and ensure high accuracy.

Passive location systems are susceptible to the Earths curvature when detecting long-range targets,and the process of unifying the coordinate systems of multiple observation stations introduces additional errors into the system,resulting in large deviations between the target position estimated by the system and the real target position,which in turn causes serious consequences on the subsequent data correlation and information fusion.In view of this,this paper proposes a passive location algorithm using Time Difference of Arrival (TDOA),which is based on two-station coordination considering the Earths curvature.Firstly,all geographical coordinates are converted to geodetic Cartesian coordinate system.Then,the information of TDOA of the target is received by the two stations.Finally,the geodetic Cartesian coordinates of the target position are solved based on the elliptic equation of the Earth and the method of Taylor series expansion.Theoretical analysis and simulation experiments show that the proposed algorithm can overcome the influence of the Earths curvature,realize the standardization of the coordinate systems of the location system,and effectively reduce the location error,which provides a new technical idea for the passive detection of long-range targets.

As for general linear second-order multi-agent systems,a distributed impulse consensus control algorithm with sampled displacement information is designed.By using the methods of spectral analysis and quasi-periodic homogeneous Lyapunov function,the criteria for achieving second-order consensus are given, which reveal the mechanism of quantitative effects of the sampling period,the controller gain and the eigenvalue of Laplace matrix on achieving second-order consensus.The influence of the sampling period and uncertainties of system parameters on the distributed impulse consensus algorithm is further analyzed,and the sufficient conditions for achieving robust second-order consensus are given in the form of low-dimensional linear matrix inequalities.Finally,numerical examples have verified the proposed theoretical results.

In order to enhance the assault and penetration capability of airborne forces,the multi-objective planning problem of heavy equipment airdrop formation of dual transport aircraft is studied.Firstly,the motion process of the cargo platform of heavy equipment airdrop is analyzed based on the separation method, and the dynamics model of the cargo platform is constructed on the basis of Euler iteration idea.Then,the spatial position model of the dual transport aircraft formation is constructed by analyzing the influence of the front aircrafts tail vortex on the rear aircraft.In order to determine the optimal solution of the relative position of the dual aircraft,the improved NSGA-Ⅱ algorithm is adopted,the dominance strength,the optimized elite strategy,the crossing operator based on individual grades in the population and the mutation operator based on Gaussian distribution are introduced,and the congestion degree is calculated based on variance,so as to obtain the final results.The simulation experimental results show that the optimized algorithm can improve the flight safety of the dual transport aircraft formation,and the airdrop points are denser with higher airdrop accuracy.

To tackle the problem of complex maneuvering target tracking,an algorithm of Adaptive Random Weighted Cubature Kalman Filter (ARWCKF) is proposed.As the preprocessing process of Interactive Multiple Model (IMM) algorithm,this algorithm conducts filtering on different motion models.In order to improve the stability of the algorithm,a random weighted factor is introduced.A time-varying factor is used to adjust Markov probability transfer matrix,which improves the probability conversion accuracy of IMM algorithm.Compared with IMM-CKF algorithm,the proposed IMM-ARWCKF algorithm has higher tracking accuracy and better stability in dealing with complex maneuvering targets.

In the future,manned fighter aircraft will be an important command and control node for manned/unmanned collaborative operations and cross-domain operations.The information exchange between the pilot and the aircraft will increase explosively,which brings great challenges to the pilots situational awareness,information acquisition,flight control,and operation command and control.Using large-size display can effectively enhance the man-machine interaction ability.High-resolution,large-size,touch-control display has become the direction of research in the airborne cockpit display and control area.In order to meet the needs of future operations,from the perspective of requirements,this paper analyzes the future development of cockpit display of manned fighter aircraft,and gives the corresponding judgment and theoretical basis.

A distributed state estimation algorithm based on sensor network is proposed for networked systems with random transmission delays.In order to save communication bandwidth,an adaptive probability quantization mechanism is introduced.Considering the existence of random transmission delays,an improved Unscented Kalman Filter (UKF) algorithm is designed by using the data in the buffer to compensate for the delays.The fast covariance crossing method is used for fusion estimation after local estimation is completed, so as to obtain more accurate state estimation results.In addition,the boundedness of fusion estimation errors is analyzed.The effectiveness of the algorithm is verified by a numerical simulation example.

In complex electromagnetic environments,it is difficult to identify jamming signals due to small sample size.To solve the problem,a jamming signal recognition method based on meta-learning is proposed.Firstly,the Holder coefficient of the frequency response of the jamming signal is calculated.Then,the time-frequency diagram of the jamming signal is input into the residual network,and the output is the eigenvector.Multi-modal fusion of the eigenvector with the above Holder coefficient is conducted to form a new multi-dimensional eigenvector.Finally,through meta-learning,the outputted multi-dimensional eigenvector is split into a coding vector and a covariance matrix related to the time-frequency diagram of the jamming signal to calculate the predicted value of the jamming signal,and the shortest Euclidean distance between the actual value and the predicted value is calculated to identify and classify the jamming signal.The simulation results show that the jamming signal recognition method can effectively improve the recognition rate on the 1-shot and 5-shot data sets of small sample size.

In order to provide commanders with more ideal assault route options,a multi-attribute decision making method based on intuitive fuzzy GRA and TOPSIS is proposed.Firstly,an evaluation index system of the assault route that conforms to the actual operation is analyzed and constructed.Then,the intuitive fuzzy weighted average operator in the intuitive fuzzy set theory is used to assemble the decision expert evaluation matrix,and the attribute weights are calculated by using intuitive fuzzy cross entropy.Then,the GRA-TOPSIS model is used to rank the assault routes to be selected by merit.Finally,through example calculation and comparative analysis,the validity and reliability of the proposed method are verified,which provides a new method for the evaluation of assault routes.

Automatic Dependent Surveillance-Broadcast (ADS-B) has such features as open protocols and no authentication measures,which makes it vulnerable to message modification attacks.Firstly,the correlation between attacker intention and attack path selection is analyzed.Then,the TriLSTM-SVDD model with feature processing is proposed for spoofing data detection,so as to defend against message modification attacks implemented through 1090ES,onboard network or other attack paths.The experimental results show that the prediction error of TriLSTM model is significantly smaller than that of LSTM model,and the TriLSTM-SVDD model can identify latitude value deviations of 0.01°,longitude value deviations of 0.02° and height value deviations of 50 m with 98.9% accuracy.

Keystone Transform (KT) is a classical target range migration correction tool in radar.The existing KT implementation methods suffer from high computational complexity and dissatisfactory anti-noise performance.To solve the problems,a KT implementation method based on Fast Fourier Transform (FFT) interpolation is proposed.The basic idea is to convert the slow time echo to the frequency domain first,and then calculate the scaled echo through the continuous operations of frequency domain zero padding,Inverse Fast Fourier Transform (IFFT) and time domain decimation,etc.to complete KT decoupling.The simulation results show that the anti-noise performance of the proposed method is better than that of the existing KT implementation methods,and the computational complexity is significantly decreased.

To solve the problems of complex and diverse background,dense targets and big scale differences in remote sensing images,which are prone to cause missed detection and false detection of small targets,a target detection algorithm of remote sensing images based on context information is proposed by taking YOLOv5s algorithm as the basic framework of the network.Firstly,Context Module (CM) is designed and added to the backbone network to enlarge the perception range of the target area features,obtain more context information,and improve the ability of the model for small-scale target detection.Secondly,Coordinate Attention (CA) module is introduced into the feature backbone network to strengthen the models ability to recognize the target position information in the shallow network.Finally,the Spatial Pyramid Pooling (SPP) module is replaced with the Atrous Spatial Pyramid Pooling (ASPP) module to realize the fusion of global information and local information,and further enhance the semantic information of small targets.The experimental results show that the mAP50 of the improved algorithm is 97.9% on the RSOD dataset, which is 1.7 percentage point higher than that of the original YOLOv5s algorithm.The FPS reaches 71 frames per second,which meets the requirements of real-time detection.Compared with other detection algorithms,the improved algorithm has a lower missed detection rate and false detection rate,and the detection performance is better.

In the view of UAVs,target scales have large differences,the detection scenarios are complex, and the targets are small and dense,which will lead to low detection accuracy.To solve the problems,a real-time target detection algorithm based on the improved YOLOv5n is proposed.Firstly,the capability of Convolutional Neural Network (CNN) for extracting effective information from the feature map is improved by introducing the Efficient Channel Attention (ECA) module.Secondly,the Adaptively Spatial Feature Fusion (ASFF) module is added after the output of the Feature Pyramid Network (FPN) to improve recognition accuracy of feature maps at different scales.Then,EIoU loss function is used to calculate the difference value between the prediction frame and the target frame to speed up convergence and improve detection accuracy.Finally,improvements are made to the detection head of YOLOv5n to optimize the models detection performance on small targets.Training and testing are carried out on VisDrone dataset.In comparison with the basic YOLOv5n model,the enhanced model improves mAP50 by 6.1 percentage points at 640×640 resolution,and improves mAP50 by 7.1 percentage points at 1504×1504 resolution.Meanwhile,the detection speed of the improved model is higher than 22 frames per second on hardware.The proposed model ensures a high enough detection speed while improving the accuracy,which is more suitable for real-time detection of small targets by UAVs.

Multi-UAV cooperative task assignment is broadly applied in many fields and thus widely studied.However,the application scenarios of communication constraints and resource consumption allocation methods are often ignored.Therefore,this paper proposes a distributed multi-UAV task assignment method based on CNP to solve the above problems.This method can be divided into four steps,that is,task release,bid application,coalition formation and contract signing.In the task release stage,the adaptive setting method of maximum number of information transfer times is proposed to limit the range of recipients and reduce communication load.In the process of coalition formation,the resource consumption allocation algorithm based on the Gini coefficient is proposed to reasonably adjust the resource difference between UAVs in the coalition.The simulation results prove the effectiveness of multi-UAV task assignment method based on CNP and resource consumption allocation algorithm based on the Gini coefficient.

In the existing quality-related fault detection methods,the fault signal is prone to be covered up by the random trend of non-stationary variables,which leads to low fault detection rate.To solve the problem,a quality-related fault detection method based on Cointegration Analysis and Modified Projection to Latent Structures (CA-MPLS) is proposed.Firstly,the stationary feature trend among non-stationary variables is extracted by using cointegration analysis.Then,the stationary feature information and stationary variables are fused to construct orthogonal projection space,and the process variables are orthogonally projected into quality-related subspace and quality-unrelated subspace.Finally,the corresponding statistical indicators are designed in the two subspaces for online monitoring.Simulation results from Tennessee Eastman demonstrate the effectiveness of the proposed method in significantly reducing the influence of non-stationary features and enhancing fault detection accuracy.

As for the non-uniformity of infrared images,the causes of occurrence of non-uniformity are analyzed theoretically.The EM algorithm of model-free image correction based on SG filter pretreatment is improved,and a single-frame image correction method of improving LSNR is proposed.A comparative experiment is designed,under the scenes of infrared target,dim target and cloud background,the method can reduce the non-uniformity (NU) of the infrared image by 56.869 3%,85.938 4% and 87.886 3% respectively, and increase the LSNR by 3.687 7 dB,0.256 9 dB and 3.553 1 dB respectively.Finally,the Gaussian-distributed target is superimposed on the non-uniformity background for simulation,and the relationship between the size of the target and the filtering window is discussed.The approximate relationship between the filtering window and the small infrared target is obtained as follows:H≈4.6*ST+0.3,and the selection of the initial value of the methods filtering window is determined from practical engineering applications.The results show that the proposed method can effectively correct the non-uniformity by using the scene information of a single-frame infrared image on the premise of maximum LSNR.