The traditional Mayfly algorithm for UAV path planning has the problems of poor stability, low accuracy, slow convergence speed and local optimization.To solve the problems, a multi-strategy improved Mayfly algorithm applied to UAV path planning is proposed.Firstly, the cost function model and the environment model are established, and the UAV path planning problem is transformed into an optimization problem that meets the feasible route requirements and the route safety constraints.Secondly, Particle Swarm Optimization (PSO) is initialized based on Lévy flight principle, and the traditional Mayfly algorithm is improved by using adaptive t-distribution and Pareto-based elite retention strategy.Finally, the validity of the proposed algorithm is verified by simulation experiment.The simulation results show that the performance of the improved Mayfly algorithm is better than that of the traditional Mayfly algorithm and the particle swarm optimization, and the quality of the planned trajectory is higher.

Object detection is decoupled into two subtasks, namely, classification and localization, in mainstream detectors. Each task possesses a separate detection subnetwork and is trained with an independent loss function. In this way, the correlation between classification and localization is disregarded, and thus the classification score predicted by the model is not capable of representing the localization quality of the prediction box. Consequently, predictions of high localization quality may be suppressed by their poorly localized counterparts in the procedure of Non-Maximum Suppression (NMS), inducing precision degradation. To tackle this problem, a consistency loss is proposed to constrain the rank similarity between the classification score predicted by the model and the localization quality in training process to reinforce about their consistency. Based on FCOS-ResNet50 model and PASCAL VOC dataset, the proposed loss function brings about 1.3 percentage points of mAP0.5, 4.3 percentage points of mAP75, and 5.4 percentage points of mAP90 improvements.

In practical applications of image recognition, the training data tend to follow a long-tailed class distribution without regard to manual balancing.The recognition effect of long-tailed image recognition algorithms based on deep learning is not good, and the recognition accuracy of middle and tail categories is unsatisfactory.In order to address the problems, a Personalized Multi-expert Recognition Algorithm (PMRA) is proposed.Firstly, a multi-expert network is constructed by integrating multiple branches based on the residual network.Then, a personalized learning module is built by assigning personalized training data to different experts to improve the recognition accuracy of middle and tail categories, and a personalized information enhancement module is built by fusing the experts information to deal with the lack of information of middle and tail categories.In the multi-expert network fusing multiple modules, the overall recognition accuracy of long-tailed images is improved by two stages of learning.Finally, the experimental results on benchmark datasets of CIFAR-10-LT, CIFAR-100-LT, ImageNet-LT and iNaturalist2018 show that the recognition accuracy of the proposed algorithm on multiple datasets is better than that of other algorithms.

As for the image rotation compensation servo mechanism in the aerial imaging system, the disturbances will influence its tracking accuracy.To solve the problem, an improved composite control algorithm is proposed.Firstly, as for the equivalent disturbance, a finite-time hyperbolic sinusoidal tracking differentiator (FSTD) is proposed to construct an improved Nonlinear Disturbance Observer (NDO), which makes the system insensitive to parameter perturbation and more accurate in disturbance estimation.Secondly, in order to improve the tracking accuracy of the system, a New High-order Sliding Mode Controller (N-HSMC) is designed by using the idea of the variable-speed reaching law, which ensures the rapidity of the system and also effectively reduces chattering.Finally, through a comparative simulation, the effectiveness of the proposed method is verified.

To solve the problems of input delay and saturation during helicopter maneuvering flight with uncertain model, a model-free incremental adaptive optimal control scheme is proposed.Firstly, the approximate time-varying model of the system is obtained by using incremental nonlinear technique, and the parameters of the relevant matrix are identified by Recursive Least Squares (RLS) estimation.Secondly, the functional performance indexes are used to deal with input delay and saturation, and Incremental Adaptive Dynamic Programming (IADP) is used to design the approximate optimal tracking control law.The Time Difference Error (TDE) function based on real-time state and delayed input is approximated by neural network, and the weight updating rate of the evaluation network is obtained by its instantaneous integral.Finally, the stability of the closed-loop system is proved by the Lyapunov function analysis.Simulations of helicopter maneuvering flight velocity tracking control are given to verify the effectiveness of the proposed control scheme.

According to the technical characteristics of new naval gun weapons and advanced guided ammunition, the idea of intercepting aerial targets by using guided projectile is proposed, which combines the advantages of ordinary ammunition and air defense missile, and the mathematical model of killing airspace of air defense guided projectile is constructed.According to the characteristics of the exterior ballistic curve of air defense guided projectile, the exterior ballistic curve is modeled.Based on the exterior ballistic model, the factors that influence the killing zone and the launch zone are analyzed, and the mathematical models of the killing zone and the launch zone are established.The numerical model of vertical killing zone under zero route shortcut is constructed, and an algorithm that can quickly generate the numerical model of vertical killing zone under different lengths of route shortcuts is designed.The numerical simulation analysis of a specific example is conducted.The simulation results show that with the increase of the length of the target route shortcut, the range of the vertical killing zone of the projectile decreases gradually.The constructed killing airspace model of air defense guided projectile is reasonable and effective, which has great value for the operational application of air defense guided projectile.

Multi-core processing is widely used due to its high performance.From the perspective of multi-core system architecture, the processor architecture and the software architecture are analyzed respectively.According to the characteristics of airborne applications, homogeneous multi-core processing and software architectures such as AMP, SMP and BMP are discussed.Based on the status quo of researches and to meet the requirements of inheritance and incremental certification for airborne avionics systems, the commercial P5020 multi-core processor is chosen, which is configured with Tianmai operating system based on AMP architecture, and a performance test case is transplanted.The test results show that the multi-core processing system can achieve double performance improvement to a certain extent, which can meet the requirements of computing-intensive applications of avionics systems.

To identify the interference of polarization-modulated false targets generated by the forwarding of Digital Radio Frequency Memory (DRFM), a method of identifying polarization-modulated false targets is proposed.Firstly, a polarization measurement model of the circularly polarized antenna array based on time-space joint domain coding is proposed, and it is proved that this model can accurately obtain the polarization information of echoes after pulse compression.Secondly, according to the principle of the tensor product signal reception model of the polarization radar, the distribution difference between the target echo and the polarization-modulated false target echo in the four-dimensional complex space is obtained.Finally, based on the difference and in combination with the maximum and minimum distance algorithm and the density method, the K-means algorithm is improved, and a classifier taking the optimal clustering number K as the polarization discriminant quantity is designed.The simulation experiment takes 64 transmitted pulses as examples, and verifies that the proposed method has good discriminating effect under different Signal-to-Noise Ratio (SNR) and Jamming-to-Noise Ratio (JNR), and can effectively discriminate between targets and polarization-modulated false targets.

In order to realize the self-alignment function of the low-cost Micro Inertial Measurement Unit (MIMU) and improve the self-alignment accuracy, this paper proposes a method of moving MIMU self-alignment assisted by GPS single antenna based on Kalman Cascade Filtering (KCF).Firstly, the Modified Ensemble Empirical Mode Decomposition (MEEMD) is performed on the random error of the velocity vector measured by the GPS single antenna, and after signal reconstruction, Kalman filtering is used for denoising calculation to obtain the heading angle measurement.Secondly, the state equation of the system is established by taking the attitude angles of the base and gyroscope constant drift as the state variables, and the observation equation of the system is established by fusing the accelerometer and GPS single antenna measurement information.Then, the Adaptive Unscented Kalman Filter (AUKF) is used for information fusion, and the optimal estimation of the attitude angles of the base is realized.After simulation verification and comparison, it is proved that the proposed method effectively improves the self-alignment accuracy.The simulation results validate the superiority of the proposed algorithm in MIMU self-alignment assisted by GPS single antenna.

Image compressive sensing is a technology that reconstructs the original image as far as possible under the condition of under-sampling.Most of the image compressive sensing methods based on the framework of Convolutional Neural Network (CNN) are prone to be limited by the receptive field of convolution and pay less attention to global information.To solve the problem, an image reconstruction network using compressive sensing based on Swin Transformer is proposed.The network uses the convolutional layer for image sampling, and then uses the structure of Residual Swin Transformer Group (RSTG), which combines the self-attention mechanism with the residual structure, to focus on the details of the image.The experimental results show that the image reconstruction network using compressive sensing based on Swin Transformer can make full use of the prior information of the image, further improve the image reconstruction accuracy of compressive sensing, and obtain better reconstruction performance and visual effects than that of other compressive sensing methods.

During the pursuit and evasion process on the battlefield, the fighter aircraft needs to complete super-maneuver actions to quickly enter the dominant attack area.However, in the range of large angle-of-attack flight, the flight process is strongly coupled.Traditional dynamic inversion methods have good fast decoupling capability, but the robustness is poor.An L1 adaptive flight control method based on nonlinear dynamic inversion is proposed, which improves the dynamic performance and robustness of the system by introducing PI-type dynamic inversion control and L1 adaptive structure.Finally, a certain software is used to simulate the cobra maneuver of the fighter aircraft, and the results show that besides improving the dynamic performance of the system, the method can effectively compensate for such perturbations as parameter uncertainty and improve the robustness, which provides a technical reference for air combat of the fighter aircraft.

A fractional-order, Super-Twisting sliding mode controller is proposed for the trajectory tracking of quadrotor UAVs under external disturbance.Firstly, based on the double closed-loop control strategy, the quadrotor UAV system is decoupled as a position subsystem and an attitude subsystem.Secondly, the sliding mode controller is designed by using the Super-Twisting algorithm, which can realize global finite-time convergence and eliminate system chattering.In order to further improve the control precision and anti-interference performance of the system, a fractional-order differential and integral operator is introduced.Finally, a simulation is conducted in comparison with the Super-Twisting sliding mode and the integral terminal sliding mode, and the superiority of the proposed algorithm is verified.

There are weak or no GPS signals indoors, so the accuracy of indoor UAV positioning is low.To solve the problem, the fusion of visual and inertial data is introduced to realize indoor UAV positioning.At the front end, the feature matching algorithm is improved.As for the rotation motion of the image, the Principal Component Analysis (PCA) method is used to calculate the rotation angle, the grid of the feature point and its 8 neighborhood grids are changed, and the Gaussian threshold is set according to the Euclidean distance between the neighborhood feature point and the matching feature point.A new score statistics model is proposed to increase the number of correct matching pairs, so as to improve the rapidity of feature matching and the accuracy of indoor visual positioning.To solve the problem of mismatching caused by local similarity of images, a data set is determined by using the geometric relationship between feature points.The similarity of data is analyzed by using Pearson correlation coefficient, and the threshold is set to eliminate the feature matching pairs with low confidence, so as to optimize the visual estimation of UAV pose information.At the back end, the visual inertia is used to optimize the pose information based on the tight coupling of the sliding window.The UAV hovering experiments under normal indoor illumination and dim indoor illumination are designed, and the flight log is analyzed.It can be seen that the feature matching speed of the improved Grid-based Motion Statistics (GMS) algorithm is 3 times higher than that of the original algorithm, and the mismatching in similar local areas is eliminated.The feature matching accuracy can reach 94%, and the accuracy of indoor UAV positioning can reach 0.02 m.The algorithm can be better applied in complex military environments.

The problem of hybrid event-triggered consistency of uncertain multi-agent systems under network attacks is studied.The traditional time-triggered mechanism and event-triggered mechanism cannot guarantee the best performance of the system.Therefore, a more efficient hybrid trigger mechanism is adopted.The switching between event-triggered mechanism and time-triggered mechanism conforms to random Bernoulli distribution, and the interval between any two adjacent triggers of event-triggered mechanism is at least one sampling period to effectively save network resources.The effects of time-varying delay control input, model errors, external disturbances and network attacks are considered.The mathematical model of the closed-loop multi-agent system is established.According to Lyapunov asymptotic stability theory and LMI theory, the consistency theorem of multi-agent system under hybrid trigger mechanism is given.Finally, the validity and feasibility of the theory are verified by simulation experiment .

To solve the problem of collision avoidance control in UAV formation cooperative operations, a distributed collision avoidance control method of UAV formation based on dipolar navigation function and Model Predictive Control (MPC) is designed.Firstly, the navigation function and control law of each UAV are established by using current position and target position of other UAVs in the neighbor set.Then, a distributed high-level formation control law based on MPC is designed and a distributed MPC algorithm is given.Finally, five UAVs flight to the target points designated to each UAV at the stage of terminal attack is taken as an example, and the collision avoidance control method of UAV formation is simulated.The simulation results show that the collision avoidance control method of UAV formation cooperative operations based on navigation function and MPC is effective.

To solve the problems of color distortion and loss of details in underwater images, an underwater image enhancement algorithm based on Multi-Scale Triple Attention (MSTA) is proposed.The algorithm uses the Generative Adversarial Network (GAN) as the basic architecture, and the generative network adopts the encoding and decoding structure.An MSTA module is designed.The combination of the multi-scale structure and the Triple Attention (TA) mechanism can realize the cross-dimensional interaction of information at different levels, making the network better learn the features of underwater images and suppress the features of noise.The discriminant network adopts a structure similar to Markov discriminator.Multiple loss functions are constructed to make the generated image consistent with the reference image in terms of structure, content and color.The experimental results show that the proposed algorithm is superior to the comparison algorithms in terms of subjective visual effects and objective evaluation indicators.The proposed algorithm can effectively improve the feature extraction ability of the network, restore the color of underwater images in different scenes, and enhance the contrast and clarity of the images.

Remote sensing ground object images have the characteristics of complex background and numerous varieties, and traditional segmentation algorithms will lead to edge blur, information loss and low segmentation accuracy.To solve the problems, a semantic segmentation algorithm based on the improved DeepLabV3+ network is proposed.Firstly, the improved feature extraction network CHRNet is introduced into the backbone network.Secondly, the Non-Subsampled Contourlet Transform (NSCT) algorithm is used to reconstruct the global pooling operation in the Atrous Spatial Pyramid Pooling (ASPP) module.Finally, the parameter-free attention mechanism SimAM is added in model encoding and decoding stages to enhance feature transfer among modules and improve feature utilization ratio.The experimental results show that Mean Intersection over Union (MIoU) of the improved algorithm is 81.56% on PASCAL VOC2012 data set and 64.2% on WHDLD data set, which are about 4.61 percentage points and 2.8 percentage points higher than those of the original algorithm.The improved algorithm can enhance segmentation accuracy while ensuring segmentation speed.

To solve the problem of UAV rendezvous formation control for the formation consisting of a tanker and multiple receivers in the process of autonomous aerial refueling, a formation control approach is designed based on Lyapunov Guidance Vector Field (LGVF) and the information consensus theory.Firstly, the UAV formation models are established, including the communication topology model and the formation geometry model.Based on this, the cooperative variables of the multi-UAV system are defined.Then, the expected refueling trajectory is designed, and the flight control law of the tanker (the formations reference point) is designed based on the LGVF method.Then, according to the position and velocity information of the tanker, the UAV rendezvous formation control method is designed based on the consensus theory, and the stability of this method is analyzed.Finally, a numerical simulation is carried out to illustrate the effectiveness of the proposed method.

Performing missions in complex environments has become an inevitable trend in the development of rotorcraft UAVs.In order to enable the rotorcraft UAVs to efficiently and safely execute missions in complex environments, scholars have conducted extensive research on binocular-vision obstacle avoidance technology for rotorcraft UAVs.In order to understand current situation and development trends of binocular-vision autonomous obstacle avoidance technology for rotorcraft UAVs, the composition of the binocular-vision obstacle avoidance system is introduced, and the role each technology plays in the binocular-vision obstacle avoidance system is analyzed at first.Then, the camera calibration methods, stereo matching algorithms and obstacle avoidance algorithms commonly used in binocular-vision obstacle avoidance technology are introduced, the advantages and disadvantages of each are summarized, and the latest research progress is expounded.Finally, the development directions of binocular-vision obstacle avoidance technology are predicted.