Ant colony algorithm is an intelligent optimization algorithm, which has the advantages of strong robustness, accurate feedback information and strong distributed computing ability.It is widely used in mobile robot path planning.The original algorithm converges slowly and is prone to falling into local optimization.To solve the problems, an improved ant colony path planning algorithm is proposed.Firstly, the solution set is updated and mutated by integrating the idea of brainstorming, so as to speed up the convergence and ensure the diversity of the algorithm.Secondly, the local path attention mechanism is used to extract better path segments, so as to improve the optimization efficiency, and the adaptive t-distribution is added to the pheromone attention mechanism, so as to avoid falling into local optimization.The new pheromone updating method can promote the global search of the algorithm and ensure the convergence rate of the algorithm.Finally, simulation experiments in static environment are carried out in Matlab software, which have verified the effectiveness and feasibility of this algorithm.

In order to solve the problems of complex background of aircraft target in optical remote sensing images, the imbalance between detection accuracy and detection speed, and the likelihood of miss detection, an SPSSD model fusing different network modules is proposed.Firstly, Resnet50 is adopted to replace the feature extraction network in SSD300 algorithm, and a controllable dilated convolution module is added to expand the features receptive field to obtain more feature information favorable to target detection.Secondly, shallow feature information is obtained by adding FPN and SPP network, and the features of the expanded receptive field are fused with deep feature information.Then, the feature information is sent to the ECANet to obtain more complete and more judicious feature information.Finally, NWPU-RESISC45 dataset is adopted to input 3400 high-resolution remote sensing images into SPSSD model for iterative training, and mAP of the improved algorithm model finally reaches 92.68%, which is improved by 5.18 percentage point in comparison with that of the previous algorithm, and the detection speed reaches 25.1 frames per second.The experimental results show that the proposed method can effectively balance the detection accuracy and detection speed of aircraft targets, which reduces the miss detection rate to a certain extent.

In the face of UAV penetration tasks under complex terrain conditions, Particle Swarm Optimization (PSO) is easy to fall into local optimum and suffers from long search time in the process of finding the optimal path.To solve the above problems, the spherical coordinate system is introduced into PSO, and the path obtained is regarded as a vector.The iterative updating of particles is realized through the relationship between the distance, elevation and azimuth of the vector and the speed, pitch and steering angle of the UAV.Finally, the random adaptive inertia weight is introduced to make up for the deficiency of the particles local search ability in the early stage and global search ability in the later stage.The simulation results show that the improved algorithm can effectively avoid the threat region, has faster convergence rate and higher convergence accuracy, and is not easy to fall into local optimum.

The attitude of the quadrotor UAV will be disturbed by atmospheric turbulence when UAV performing tasks.Therefore, a method to optimize the traditional PID controller using variable domain fuzzy theory is proposed.According to the kinematic equation, a nonlinear model of quadrotor UAV with more accurate control accuracy is established.Due to the strong randomness of atmospheric turbulence, the model of components of atmospheric turbulence is obtained by using the Dryden model.Considering the lack of anti-interference ability of traditional PID controller, a variable domain fuzzy adaptive PID controller is designed by using variable domain fuzzy control.In the Simulink environment, the two controllers are compared through simulations after successively adding atmospheric turbulence and step response disturbance.The results show that the optimized controller has better robustness and stability than the traditional controller.

For the exploration of unknown environments, such as search and rescue, chase and escape scenarios, the UAV needs to explore (perceive) the environment while completing current trajectory planning (action selection).Aiming at the above scenarios, in order to achieve efficient environment exploration, an improved Deep Double Q Network (DDQN) exploration trajectory planning method based on Long Short-Term Memory (LSTM) network is proposed.A simulation map is built, the environmental information in the UAVs field of view is taken as input, the LSTM network is introduced, and the choice of action direction is outputted.The priority of exploration experience samples is set to improve training efficiency.Flight dynamics constraints are added, and reasonable state, action space and one-step reward function are designed.Using the proposed algorithm, the UAV can autonomously plan a collision-free track with a wide range of environmental exploration.The simulation results show that the proposed algorithm is better than the traditional DDQN algorithm in the exploration area ratio and the average reward of one-step exploration in unknown environments.

Aiming at the characteristics that the accuracy of object detection and positioning is affected by the loss function of frame regression, a frame regression loss function IAIoU (Included Aspect-ratio IoU) based on IoU (Intersection over Union) is designed.Two optimization terms are designed for this loss.There is the ratio of the difference between the union and intersection area of the prediction box and the labeled box to the smallest enclosed area of the two boxes, as well as the ratio of the difference to the square of the smallest enclosed area of the two boxes.The sum of the two ratios is used as the first optimization term, which avoids the degeneration of loss function when the two boxes are included.The difference between the aspect ratios of the two boxes is adopted as the second optimization term to generate a prediction box that is closer to the labeled box.The designed loss is applied to the single-stage detection algorithm YOLOv3 and verified on the infrared aircraft data set.The mAP reaches 92.17%, which is 1.37% higher than that of the original YOLOv3.

In order to overcome the influence of unknown actuator faults on the flight of quad-rotor UAV formation, an optimal cooperative fault-tolerant control law based on dynamic programming is proposed.The quad-rotor UAV model is established, and the optimal cooperative control law is designed based on dynamic programming.RBF neural network is used to approximate the optimal performance index function, and an adaptive law is designed to estimate the unknown actuator fault.The optimal cooperative fault-tolerant control law can realize the high-precision control of UAV formation flight.The comparative simulation has verified that the designed optimal cooperative fault-tolerant control law has better formation control effect, and the maximum trajectory tracking error of formation flight is only 0.04 m, which shows that the control accuracy is high.The designed adaptive law has better fault estimation effect, and the maximum estimation error is only 0.05 N·m, which realizes the safe and stable control of quad-rotor UAV formation.

Direction-finding cross positioning is simple and requires less measurement information, which is widely used in the field of passive positioning.It is difficult for the traditional direction-finding cross positioning algorithm to solve the positioning problem when the number of targets is unknown.To solve the problem, a direction-finding cross positioning algorithm based on grid density peak is proposed, which introduces meshing and density peak clustering into direction-finding cross positioning, and uses Hough transform to formulate a rule for picking cluster centers.According to the results of simulation experiments and measured data experiments, it is proved that the proposed algorithm can automatically identify the number and position of targets in the presence of noise.It has strong robustness, which can be applied to practical problems.

When the deep learning algorithm is applied to the field of SAR image classification, there are some problems such as long model training time and low accuracy.In order to solve the problems, a convolution neural network model based on hybrid attention mechanism is proposed.The basic module of the model is divided into trunk branch and soft branch.The trunk branch is composed of residual shrinkage network and the improved channel attention mechanism, which is responsible for extracting the main features.The soft branch combines down sampling with up sampling to extract the hybrid attention weight and enhance the mapping ability from input to output.The recognition rate of the model is 99.6% on MSTAR data set, and the training time is short.Noise analysis shows that the model is robust to salt and pepper noise.

Bird strike event is a serious threat to the safety of aviation operation.Monitoring bird targets via quantum lidar is regarded as an effective means to detect potential bird strike events.According to the spatial location information of bird targets with high precision and high resolution obtained by quantum lidar, and based on the operation characteristics and safety standards of aircraft around the airport, a threat level assessment model is proposed, which comprehensively considers the collision probability of bird targets and damage severity.Because of the use of the exponential function for classification, the threat level assessment model can be more accurate in threat level assessment of bird targets under various situations.The simulation results show that the threat level assessment model based on exponential function can evaluate the threat level of bird targets around the airport well, and the results are consistent with the actual movement of bird targets around the airport, which can provide effective warning for preventing the occurrence of bird strikes.

An Improved Spherical-vector-based Particle Swarm Optimization (ISPSO) algorithm is proposed to solve the flight path planning problem of UAVs under multiple threats in complex environment.The ISPSO, combining compression factors with asynchronous learning factors, is used to find the optimal path that minimizes the cost function by efficiently searching the configuration space of the UAV via the correspondence between the particles position and speed and the turn angle and climbing angle of the UAV.To evaluate the performance of ISPSO, two benchmark scenarios are generated from real maps of digital elevation model.The simulation results show that the proposed ISPSO outperforms the spherical vector based particle swarm optimization.

The research on the requirements of anti-submarine decision aid system for shipborne helicopter is conducted from three aspects of military requirements, mission requirements and system function requirements. It is believed that military requirements should satisfy the needs of the crew to master the marine environment and equipment performance and develop programs, and mission requirements should satisfy the needs of multi-type anti-submarine missions, multi-type combat processes and coordinated command.The systems engineering analysis method is applied, and according to the operational process of the shipboard anti-submarine helicopter, the functions of the anti-submarine decision aid system are mainly divided into four aspects, that is, information processing guarantee, anti-submarine data support and program management, task planning and system management.Based on this, the various functions are decomposed in detail.The study has certain reference value for the design of the airborne anti-submarine decision aid system.

In consideration of the special problem of VTOL UAV deck landing, an autonomous ship landing guidance strategy based on on-board vision is proposed.The deficiency of the traditional 2D optical-flow velocity measurement method is overcome, and the 3D velocity vector of the target ship is estimated by using image spherical optical flow.Then, in order to solve the problem of insufficient control accuracy of landing trajectory of the traditional position control scheme, a guidance law based on Tau theory is employed for generating a desired trajectory that meets the constraints of soft landing in real time.In this phase, a trajectory tracking controller is designed taking the 3D velocity vector of UAV flight as the control quantity.Finally, based on Unity3D software, the visual simulation environment is developed, and semi-real, closed-loop simulation for verification is conducted by simulating different sea wave conditions.Compared with traditional trajectory planning schemes, the proposed scheme does not require complex optimization iterations and can meet both the real-time and accuracy requirements of deck landing.

To solve the problems of low similarity of historical trajectories and low prediction accuracy in vessel trajectory prediction, a trajectory prediction algorithm vSeq2Seq based on Sequence-to-Sequence (Seq2Seq) model is proposed.Firstly, the first-order differential method is used to process the AIS data, so as to reduce time dependence, weaken the impact of communication delays, and highlight the rules of vessel movement.Then, the sliding window method is used to process the data, a data set of the model is built, and the trajectory prediction with variable step sizes is conducted through Seq2Seq model.The experimental results show that the vSeq2Seq algorithm can extract the features from varied trajectories of the vessels, changing the prediction step size in different motion states of vessels to predict flexibly.Compared with that of the traditional LSTM and GRU models, the prediction accuracy is significantly improved.

When cooperative multi-agent system lacks individual reward signals, the contribution of different agents cannot be distinguished, which leads to low cooperation efficiency.To solve the problem, the discriminability evaluation index of credit allocation is introduced by using the value decomposition paradigm, and a method based on gradient entropy regularization is proposed to achieve highly discriminable credit allocation.Based on this, an improved QMIX network is proposed by using the multi-agent deep reinforcement learning algorithm.Through SMAC multi-agent learning environment and Starcraft2s built-in map editor, the corresponding simulation environment is established.The results show that the learning efficiency and overall performance of the improved QMIX network are improved compared with that of QMIX network, and it is more suitable for cooperative multi-agent reinforcement learning in partially observable environment.

The airborne optoelectronic platform suffers from interference such as friction, mass imbalance, motion coupling and airborne environment.The traditional PID controller has poor control performance and weak anti-interference ability.To solve this problem, a fractional order controller with good dynamic performance and strong robustness is designed.At the same time, a Particle Swarm Optimization (PSO) algorithm based on the improved velocity and inertia weight is proposed to adjust the five parameters of the fractional order controller.The simulation results show that the improved PSO algorithm has faster convergence rate and more accurate solution.The overshoot of the speed-loop control system of the optoelectronic platform is reduced to 0.13%, the isolation of sinusoidal disturbance is 64.06% higher than that of the traditional PSO algorithm, and the response speed and robustness are greatly improved.

To solve the problems of low precision and large error of traditional navigation modes, a landing sign based on ArUco code is designed in combination with visual navigation.After processing the real-time images collected by the airborne camera, the UAV pose estimation model is established by using the mapping relationship between the world coordinate system and the pixel coordinate system.The estimated value of relative attitude between the UAV and the landing marker is obtained by solving the coordinates of the feature points.The error model is designed and fused to further improve positioning accuracy.Through the outdoor practical application experiment of UAV, it is proved that the designed landing sign and new recognition algorithm greatly improve the accuracy of UAV autonomous landing, which can be better applied to scenarios with high-precision landing requirements such as military UAVs.

The real-time performance and accuracy of the algorithm for UAV ranging are always the focus of research.As for the real-time performance of the algorithm, a bidirectional search strategy and three data streams are proposed to accelerate the GPU, and a butterfly sorting algorithm is introduced to improve the computing efficiency.As for the accuracy of the algorithm, the improved Absolute Difference (AD) algorithm is combined with the improved Census algorithm, which effectively solves the problem that the algorithm relies too much on the center pixel.In order to reduce the complexity of the algorithm, the moving average algorithm is introduced into target detection based on the background subtraction method.As for the limited storage space of UAV platform, an optimization scheme of data storage is proposed to reduce the space occupied by data.A comparison with Boyer-Moore (BM) algorithm and Semi-Global Block Matching (SGBM) algorithm is conducted on Middlebury data set, and then the outdoor scenes are collected for the comparison with BM algorithm on ranging performance.Finally, the UAV ranging is implemented outdoors.A large number of experimental data has verified that the algorithm meets the high precision requirements of UAV real-time ranging.

According to the characteristics of IEEE1394B bus, a high-speed serial bus, the application of SAE AS5643 protocol in the key field of aviation safety is analyzed.The technical characteristics of the protocol is studied, and the deterministic aspects are expounded, such as time synchronization, data communication, root node, topology and bandwidth.Based on the advantages of time trigger, the AS5643 protocol is improved, the functions such as time trigger scheduling and sending and receiving time are improved, and a time trigger scheduling design architecture based on AS5643 protocol is put forward, so as to further improve the robustness and certainty required for the application of aviation safety critical systems in future high-speed (1.6 Gibit/s, 3.2 Gibit/s, etc.) communication applications.RTL-level implementation and simulation analysis of the design scheme are conducted, which verify the correctness of the communication.

With good concealment, the locating and tracking system based on azimuth measurement is an effective means of counter reconnaissance, anti-jamming and surprise attack.In view of the locating demand to ground moving target by ground observation stations, the target models of uniform and uniform-acceleration rectilinear motion are established.In order to reduce the influence of angle measurement error, the target locating algorithm is derived according to the least square estimation criterion.The main factors affecting the locating accuracy are analyzed, and the simulation experiments and numerical analysis are carried out from aspects of the number of observation stations, target distance and angle measurement error.The mean and standard deviation of locating error are analyzed by using Monte Carlo method.The calculation results show that under the same conditions, better locating accuracy can be obtained by using three to four ground observation stations.The locating error and error dispersion degree increase with the increase of target distance and angle measurement error.The experimental results have certain reference significance for optimizing the configuration of ground observation stations, improving the accuracy of target locating and evaluating the expected results of equipment quantitatively.

With the extensive application of wireless equipment, electronic warfare is gradually becoming the key to winning modern warfare.Improving the anti-reconnaissance performance of radar has become an important means to enhance radar electronic countermeasures.The design of anti-sorting signals can enhance the anti-reconnaissance capability of radar fundamentally.Aiming at the algorithm of sequential difference histogram, which is widely used in the field of signal reconnaissance, the anti-sorting signal is designed from the perspective of destroying the estimation of signal Pulse Repetition Interval (PRI) by the sorting algorithm.Firstly, the operating steps of the sorting algorithm of sequential difference histogram are summarized, and the key steps of signal sorting by the algorithm are analyzed.Then, the signal design scheme is proposed.Finally, simulation verification is carried out.The simulation results show that the designed signal can effectively counter the sorting algorithm.

Vibration performance is an important index to evaluate the environmental reliability and performance stability of high-precision helicopter-borne fiber-optic gyroscopes.As the source component of helicopter-borne fiber-optic inertial navigation, the fiber source needs to ensure its stable output under vibration conditions.Aiming at the problem of the mean wavelength of the fiber source changing with vibration, based on the basic mechanism, the Jones matrix model of the optical path is established to deeply study the relationship between the polarization state of the transmitted light in the fiber and the polarization-dependent loss.Without increasing the complexity of the fiber source system, a vibration-resistant fiber source for helicopter-borne fiber-optic inertial navigation based on Faraday rotating mirror is proposed, and its performance on vibration resistance is verified through experiments.The experimental results show that the proposed scheme can improve the stability of the fiber source and fiber-optic inertial navigation under vibration conditions to a great extent.Especially under the conditions of high-frequency vibration (1000～2000 Hz), the mean wavelength stability of the fiber source is improved by 6.8×10-6.