In the planning of UAV safe collision avoidance path,Deep Deterministic Policy Gradient (DDPG) algorithm suffers from slow convergence rate and reward function setting difficulties.To solve the problems,based on reverse reinforcement learning,a UAV path planning algorithm that integrates expert demonstration trajectories is proposed.Firstly,based on the simulator software,the demostration trajectory dataset of the expert manipulating the UAV to avoid obstacles is collected.Secondly,the hybrid sampling mechanism is used to update the network parameters by integrating high-quality expert demonstration trajectory data in the self-exploration data to reduce the cost of algorithm exploration.Finally,according to the maximum entropy reverse reinforcement learning algorithm,the optimal reward function implied in the experience of experts is calculated,which solves the problem that the reward function is difficult to design in complex tasks.Comparative experimental results show that the improved algorithm can effectively improve the efficiency of algorithm training and the obstacle avoidance performance is better.

In order to optimize the problem of tracking failure during the operation of the UAV SLAM system,which leads to the failure of local mapping,and optimize the problem of large trajectory errors of the binocular camera in far and near complex scenes,this paper proposes a binocular SLAM.Firstly,new constraints are added to the system to optimize the disparity and depth of binocular matching,then the optical flow method is applied to remove the mismatches based on homography motion compensation and PROSAC algorithm,and finally new data associations are obtained by projection according to the homography compensation results.A new method is applied to the key frames and map points to improve the utilization rate of compensation results and ensure the real-time performance and robustness of the system.The proposed algorithm uses binocular parallax to optimize the calculation of the depth of key points to improve the positioning accuracy and robustness of the system,and achieves excellent accuracy improvement on the EuRoC data set of indoor UAVs.

With the development of industrial automation,infrared image recognition technology is applied more frequently to the field of automated production.Infrared images are characterized by high noise,poor image quality and lack of color information.In view of the above characteristics,a detection method,Infrared Image Frequency Domain Detection Method (IFDM),is proposed based on infrared image frequency information to identify infrared images.Firstly,different from traditional image processing,this method starts from the frequency domain and transforms the image information into the frequency domain through discrete Fourier transform,which is beneficial to better grasp the unique structural features of infrared images.Secondly,the learnable screening of frequency information in the frequency domain enhances the feature extraction capability of the model.Finally,the Transformer structure is introduced,which can better fuse the global information in the image than the CNN structure.Three unique infrared image datasets are used to verify the feasibility of the method in comparison with other algorithms in terms of accuracy and model convergence rate.

Building the parameterized turbulence model of a specific turbulence wind field is of great significance for the study of flight dynamics and flight safety under atmospheric disturbance.A parameterized modeling method of atmospheric turbulence is proposed based on the flight data of civil aviation aircraft.Firstly,based on the original flight data,the three-axis components of turbulent wind along the flight trajectory are derived,and the non-turbulence components in the turbulent wind are separated by Gaussian process regression.The turbulent components conforming to stationarity and normality are extracted through statistical tests.After that,the frequency-domain turbulence model is built based on Maximum Likelihood Estimation (MLE),and the time-domain turbulence model is built based on an auto regression process.The test results based on real-world flight data show that the customized modeling of turbulence wind field can be realized by the established parameterized turbulence model,which can be further applied to flight safety analysis and flight parameter estimation.

In wartime,the multi-destination material transportation task of Unmanned Aerial Vehicle (UAV) has the problems of difficulty in determining the destination and high real-time requirements for path planning due to its complex environment and changeable material requirements.In order to solve the problems,a multi-destination path planning method based on ant colony decision-making and receding control is proposed.This method establishes a multi-destination decision-making function with the minimum loss as the objective through the optimization mechanism of ant colony algorithm.In addition,the Receding Horizon Control with extended Solution (RHC_eS) is adopted to carry out the multi-destination path planning by adopting the strategy of decision-making while moving at the same time.Finally,in order to verify the effectiveness of the method,experiments are carried out in a multi-destination environment with obstacles.The results show that compared with other methods,the proposed method has the advantages of shorter path and less loss.

A sliding mode control method is proposed to track the trajectory of a tilted quadrotor UAV in the rotor mode and the flight mode after the change of the nacelle angle.Firstly,the dynamic system of the tilted quadrotor UAV is considered as the combination of a fully-actuated subsystem and an underactuated subsystem.Secondly,by using the Extended State Observer (ESO),the control laws of the two subsystems are designed respectively.For the underactuated subsystem,the composite sliding surface is introduced,and the Hurwitz stability is used to solve the sliding surface coefficients.Then,based on Lyapunov stability theory,all subsystems can reach the sliding surface.Finally,the effectiveness of the proposed method is verified by comparative simulation experiments.

Considering the nonlinear and fast time-varying characteristics of hypersonic vehicle,the wavelet-based predictive function control method is studied,aiming at improving the control performance of traditional step-signal-based predictive function control and realizing real-time control of hypersonic vehicle.Firstly,the nonlinear model of the hypersonic vehicle is converted into a state-dependent linear model,and then the control law is represented as a linear combination of wavelet functions,and the calculation of the control law is converted into the calculation of the basis function coefficients,which greatly reduces the dimensionality of the optimization calculation,so that the real-time control of hypersonic vehicle can be realized.In addition,different from the traditional predictive function control method,this paper uses the wavelet function as the basis function,which can make full use of the multi-scale analysis and compact local characteristics of the wavelet.By flexibly setting the number and position distribution of wavelet basis functions,the approximation requirements of the fitting points and the overall control performance are ensured.The simulation results show that the wavelet based predictive function control has better tracking performance than the traditional step-signal-based predictive function control does.

A finite-time flight control strategy based on the sliding mode method is proposed for the hovering problem under uncertainty and external disturbance in the attitude and position control of the unmanned helicopters.Firstly,the 6 Degrees of Freedom (6-DOF) model of the unmanned helicopter is analyzed,and the model is divided into a position model and an attitude model.Secondly,the model is simplified in the hovering state,and according to the characteristics between position and attitude and the control requirements,non-singular terminal sliding mode and integral sliding mode are used to design the position controller and the attitude controller respectively,and it is proved that the system error can converge to the equilibrium point in finite time under the framework of Lyapunov theory.Finally,the simulation analysis shows that the control strategy has good control performance.

In view of the shortcomings of traditional path planning methods,such as excessive dependence on environmental prior information,weak real-time performance of path planning and difficulty in adapting to complex obstacle environment,an improved Dynamic Window Approach (DWA) based on the theory of electric potential energy is proposed to expand the subfunction space of trajectory evaluation.Firstly,the design of predictive trajectory selection function is improved to guide the jump out of local optimum and get away from the concave obstacle environment.At the same time,the dynamic adjustment mechanism of the weight of the velocity evaluation function is designed to enhance the ability to cross dense obstacle areas.Finally,the design of target navigation function is added to improve the guidance ability of the algorithm towards the target point.Simulation experiments show that the improved algorithm can effectively overcome the problems of traditional DWA algorithm,such as prone to fall into local optimum and unable to deal with complex obstacles,and effectively improve the path planning ability and real-time performance of the algorithm.

To solve the problem that the number of UAVs increases sharply and the collision risk increases sharply,the position and velocity information provided by Automatic Dependent Surveillance-Broadcast (ADS-B) is used to monitor the collision risk,and the genetic algorithm is used to realize the autonomous collision avoidance and collision avoidance path optimization of UAVs.The collision avoidance strategy comprehensively considers the UAV performance constraints,the confidence of ADS-B parameters and the UAV flight environment,the fitness function with multi-constraint parameters is constructed for multi-dimensional optimal collision avoidance path calculation.As for the situation of two and multiple UAVs collision avoidance,the UAV collision avoidance simulation is carried out.The simulation results show that the genetic algorithm can effectively optimize the collision avoidance path while achieving autonomous collision avoidance for UAVs.

To solve the problem of low detection rate of small targets and high false alarm rate due to the small size of aircraft in SAR images,an improved method based on YOLOv5 is proposed.First,the K-means clustering algorithm is used to optimize the anchor frame for the size of the small aircraft target,and the Swin Transformer module is integrated into the backbone network.At the same time,the multi-scale feature fusion mechanism of adaptive learning weights and the Global Attention Mechanism (GAM) are introduced to make the network span the space channel dimension and amplify the global dimension interaction,so as to improve the models ability to capture information in different dimensions.A small target detection layer is added to improve the networks ability to detect small aircraft targets in SAR images.The experimental results show that,compared with the original YOLOv5 method,the improved method has stronger feature extraction ability and higher detection accuracy in the detection of small-size aircraft targets in SAR images.

For infrared multi-scale target detection,a lightweight detection network based on single-stage receptive field enhancement is proposed.The detection network uses the simplified MobileNet V2 as the backbone network.In combination with atrous convolution and spatial attention mechanism,a single-stage receptive field enhancement module is designed to expand the receptive field range of the single-stage feature map and enhance the correlation between adjacent pixels.In order to improve the back propagation efficiency of the model and balance the positive and negative samples,the adaptive training sample selection method is adopted.Finally,a lightweight detection algorithm with the model size of only 1.6×32 Mibit and the floating-point operation of only 5.63GFLOPS is obtained.On the constructed MTS-UAV data set,the mAP reaches 89.6%,and the FPS reaches 105 frame/s on RTX2080Ti.

To solve the problem that model bias and external disturbances increase the difficulty of position tracking and anti-swing control of the slung-load flight system of quadrotor UAVs,a nonlinear control method based on compensation function observer is proposed.Firstly,the dynamic model of the slung-load flight system of a quadrotor UAV is built and a compensation function observer is designed to observe the unmodelled dynamics and unknown disturbances.Then,in view of the tracking error constraints of system reference inputs,a barrier Lyapunov function is built to design a backstepping controller,and the total disturbance is compensated for by introducing the estimation value of the compensation function observer in the controller,thus ensuring the UAV position errors and the slung-load swing angle to change within the constraint ranges and making the system achieve the expected control effect.Lyapunov stability analysis proves the stability of the closed-loop system,the convergence of position errors and the suppression of slung-load swing.The simulation results show that the proposed controller can effectively suppress the swing of the slung-load while realizing the accurate control of UAV positions.

In order to solve the problem that the adaptive monopulse phase discriminating curve is distorted and the performance of angle estimation is seriously degraded under the condition of main-lobe interference, a new method to improve the performance of angle measurement is proposed by introducing mirror virtual interference to offset the error generated by the interference signal.The distortion of phase discriminating curve and residual interference signal are analyzed mathematically to be the main factors of generating angle measurement errors,and the main angle of interference signal has a corresponding relationship with the deviation.Therefore,the sampling covariance matrix is reconstructed by mirror virtual interference to estimate the signal adaptively.The proposed algorithm does not require additional restrictions on the sum and difference beams,does not consume the degree-of-freedom of the antenna,and has higher stability under low SNR conditions.The simulation results demonstrate the effectiveness of the proposed method.

In order to obtain a systematic and standardized comprehensive evaluation method of laser angle deception jamming equipment,firstly,after comparative analysis,the Analytic Hierarchy Process (AHP) is used to build a three-level evaluation index system and a comparison matrix of each level index.Secondly,the weight values of indexes at all levels are obtained by using sum product and power multiplication,as well as the least square method,and the consistency test results are less than 0.1.Finally,according to the two cases of scheme selection and physical comparison test,and in combination with the existing calculation methods for evaluating the operational effectiveness of laser angle deception jamming and the fitting results of a large number of outfield test data,a more reasonable scoring method and model are established for each index,supplemented by application examples.The results show that this method can comprehensively evaluate the equipment more effectively and rigorously,and solve the problem that the previous evaluation process is difficult to refine and quantify.

As an interdisciplinary domain,the microwave photonic technology is one of the important development directions of new airborne radar systems and can realize the functions of RF signal processing and high-speed transmission by optical methods that are difficult or even impossible to realize by electronic technology.However,almost all the domestic microwave photonic radar systems developed up to now are only tested for function demonstration,and the system-level validation approach oriented to the practical operational requirements is urgently needed to promote the engineering development of the airborne microwave photonic radar.After the illustration of necessity,an integrated validation system of airborne microwave photonic radar oriented to system-level performance evaluation is proposed,and the key components of the integrated validation system and the performance indices mapping and evaluation method are expounded.The superiority of this integrated validation system is summarized.

GMSK is the main digital frequency modulation mode used by airborne sonar buoy at present.To solve the problems of low accuracy and waste of resources when Look-Up Table (LUT) technology is used to replace the complex filtering,integration and sine cosine function solution in GMSK digital modulator,based on the research and analysis of GMSK modulation signal,the mathematical model of Gaussian pre-modulation signal is improved,the mapping relationship between the associated symbol and GMSK signals at all levels is derived,and the optimized and simplified LUT is selected to design the GMSK orthogonal modulation scheme,which not only ensures the algorithm accuracy,but also minimizes resource utilization.

This paper studies the modeling and solving method of echelon usage problem of equipment measured by dual-life indexes and one of them cannot be controlled in the case of uncertain task quantity and consumption.The decision-making model of echelon usage problem of dual-life equipment under uncertainty is constructed.The maximization of the echelon uniformity index,life matching index and life utilization index is taken as the objective.The I-NSGA-Ⅲ is used to solve the problem.The improved segmented coding method and operators are adopted,and the repeated individual control mechanism is introduced to improve the diversity of the population.When the scale of the problem is large,the search of a larger range in a shorter time can be realized,which verifies the feasibility of this method.

At present,most of the research on test optimization selection is limited to the qualitative research on whether the fault can be diagnosed,without further considering the difficulty degree of fault diagnosis.To solve the above problem,a test optimization selection method based on quantitative evaluation of fault diagnosability is proposed.Firstly,the fault diagnosability is quantitatively evaluated based on Earth Movers Distance (EMD).Then,based on the evaluation results and comprehensively considering the number, reliability and cost of tests,a multi-objective test optimization selection problem is constructed,and a binary tuna swarm optimization algorithm is proposed to solve the optimal test set meeting the testability index and fault diagnosability level.Finally,a switching power supply of a certain type of equipment is taken as an experimental case to verify the effectiveness of the proposed method.The simulation results show that the proposed method can realize the test optimization selection aiming at improving the system fault diagnosability, and can fundamentally improve the system fault diagnosis ability.