The traditional Particle Swarm Optimization (PSO) algorithm is prone to fall into local optimum during the running of the algorithm, which may lead to non-optimal result of final path planning.To solve the problem, the pheromone in the ant colony algorithm is introduced for each particle in the process of iterative optimization of the traditional PSO, giving the particle more information connotation to speed up the convergence.Then, the 3D path planning problem of Unmanned Aerial Vehicle (UAV) based on fuzzy-logic PSO is studied in consideration of fuzzy logic.The input of the path planning is controlled by fuzzy processing to prevent the system from falling into local optimum.The proposed algorithm combines the advantages of the PSO, ant colony algorithm and fuzzy logic, and its effectiveness is verified by simulation.

In order to solve the problem of trajectory jamming in multi-target tracking, a Distance-Cheap Joint Probability Data Association (D-CJPDA) algorithm based on distance for re-weighting is proposed on the basis of CJPDA.The probability of interconnection between the candidate measurements and the tracks is calculated by using CJPDA algorithm, and the probability of interconnection is re-weighted according to the distance relationship between each candidate measurement and track prediction point.The state and covariance of the target are updated according to the equivalent value obtained from the new probability distribution, thus to maintain and update the target track.Simulation results show that, compared with CJPDA algorithm, the proposed algorithm can further improve the tracking accuracy and reduce the possibility of track merging and false tracking.

In order to improve the allocation efficiency of multi-UAV dynamic reconnaissance, and to better meet the real-time requirements of modern warfare for multi-UAV dynamic reconnaissance, a model for multi-UAV dynamic reconnaissance resource allocation of multiple UAVs is established.The model transforms the dynamic task allocation problem into a multi-stage static allocation problem, and updates the initial state of the static allocation problem of each stage by using the status updating method, so as to improve the overall allocation efficiency.The improved artificial bee colony algorithm is used to solve the model, and the bidirectional evolution is used to increase the diversity of the population at the stage of selecting the honey source, which improves the optimization ability of the algorithm.The simulation results show that the dynamic reconnaissance resource allocation model has the advantages of high reconnaissance efficiency and strong endurance compared with the traditional static allocation model, and is more suitable for large-scale UAV cluster operations.

In order to solve the problems of premature convergence and local optimization in Particle Swarm Optimization (PSO), a strategy for non-linear improvement of inertia weight is proposed, an inertia weight based on exponential function is constructed, and a random adjustment strategy is added to realize the dynamic adjustment of inertia weight.In addition, mutation and crossover operations in differential evolution algorithm are introduced to update the position of particles in order to increase the diversity of particle population.To verify the optimization performance of the proposed algorithm, four typical test functions are selected to compare the improved PSO algorithm with other algorithms.The experimental data shows that the improved algorithm proposed in this paper has higher search accuracy for complex problems and faster convergence speed for simple problems.

Aiming at the control of multi-USV formation, a distributed coordination strategy was proposed based on Multi-Agent System (MAS) with the implementation algorithm.The distributed formation coordination was realized by using the interaction between the Agents carried by the USVs.At the same time, a controller for “meta-formation” was designed based on the leader-follower method.After that, the correctness and stability of the strategy and the algorithm were verified by simulation experiments and analysis.Finally, the USVs of “Jellyfish” and “Skate” were used to test the distributed coordination control method, and the test results verified the effectiveness of the method in practical engineering applications.

The error registration problem is the key problem to be solved in radar networking to accurately estimate and eliminate system errors.Dimensional expansion filtering can estimate the system error in real time, but the calculation amount increases with the increase of the dimension.The two-stage Kalman filter can greatly reduce the calculation amount, but it is prone to calculation error divergence.Based on the research of ASEKF and TSEKF algorithm, this paper combines the square root filtering with high numerical stability with the TSEKF algorithm, introduces the sequential idea, and proposes an error registration method based on SSR-TSEKF.The simulation shows that the new algorithm effectively improves the filtering stability on the basis of reducing the amount of calculation.

Aiming at the problem that the suppression of noise Amplitude-Modulation (AM) jamming has poor real-time performance, a step-by-step algorithm is proposed based on FFT frequency offset fast correction for the noise AM jamming suppression.Firstly, the FFT-based frequency offset fast correction algorithm is used to accurately estimate the center frequency of noise AM jamming, and the calculation amount is effectively reduced under the premise of ensuring the estimation accuracy.Then, the estimated interference center frequency is used to demodulate the signal and the interference by a demodulation formula, and the bilateral spectral distribution of the signal and the interference is obtained.Finally, cancellation is carried out twice in steps to the demodulated signal in the frequency domain, and the signal is recovered in the time domain, thereby suppressing of the noise AM jamming is implemented.The simulation analysis shows that the algorithm can effectively suppress noise AM jamming under the premise of significantly reducing the computational complexity.

To solve the trajectory tracking problem of a quadrotor UAV, an optimal iterative learning controller based on the disturbance observer is designed.A discrete kinematics model of the quadrotor UAV is established in consideration of external disturbances.A discrete disturbance observer is designed to estimate the unknown disturbance, and the convergence of disturbance estimation errors is strictly proved by using the Lyapunov method.On the basis of compensating for the disturbance, an optimal iterative learning controller is designed to ensure the asymptotic convergence of the tracking error of the quadrotor UAV to the desired trajectory.Finally, the effectiveness of the disturbance observer and the iterative learning controller is verified by simulation.

In specific application, the Gaussian Process Regression (GPR) algorithm depends too much on the selection of the kernel function and has limited space for improving the accuracy.In order to obtain higher prediction accuracy, this paper proposes a Gaussian Process Regression (GPR) prediction algorithm combined with AdaBoost.RT algorithm.The algorithm introduces the concept of preset threshold from the perspective of statistics, and divides the GPR prediction results of different kernel functions into two parts, i.e.correct and error.After two layers of training, the credibility of the GPR algorithm with different kernel functions is reflected by the final weight.Fninally, combination is made to each kernel function, and the output with high-quality prediction value is obtained.In the failure prediction simulation experiment of lithium battery, the prediction error of the algorithm is reduced by 82.14% compared with the traditional GPR algorithm, which proves the rationality and practicability of the method.

Aiming at the problems of high mismatching rate and poor robustness of the ORB matching algorithm, an improved ORB-LBP feature matching algorithm based on fusion descriptors is proposed.First of all, the algorithm constructs the pyramid scale space for the input image, detects the oFAST key points on each layer to improve the scale invariance of the algorithm.Then, the image block is used instead of the pixel to improve the anti-noise performance of the LBP algorithm, and the minimum value selecting and sorting are used to make its rotation invariant.Finally, in the process of generating rBRIEF-LBP descriptor, the 128-bit modified LBP descriptor is used instead of the last 128-bit of the rBRIEF descriptor with low variance, to make full use of the image information and improve the matching accuracy and robustness.The experimental results show that, the proposed algorithm has much better matching accuracy and robustness in the case of scale changing, rotating and brightness changing than the traditional ORB does, which can better satisfy the requirements of fast and accurate matching of complex images.

In the field of track initiation, considering the problem of false tracks caused by a large number of residual clutters in complex electromagnetic environment, a new clutter suppression method is proposed.This method is based on Adaptive Boosting (AdaBoost) algorithm, which transforms the clutter suppression problem into that of AdaBoost decision tree classification.First, the base learner is trained to get a rough classification result.On this basis, the results of multiple base learners are combined to classify more comprehensively, and the problem that the weak learner performs poorly on clutter points with insignificant features is solved.The experimental results show that, compared with SVM and KNN classifiers,the proposed method can suppress the residual clutter and reduce its influence on the track initiation more effectively.

For the requirement of high-precision rapid alignment of Strapdown Inertial Navigation System (SINS), a fast initial alignment method with the constraint of constant attitude angle is proposed on the basis of traditional position-observation filtering method.Based on the fact that the equivalent rotation vector of the initial body coordinate system relative to the body coordinate system remain constant or only changes a little under static conditions, the status and measurement models are established, and a closed-loop Kalman filter is applied to implement the fine alignment.Simulation and experiment results show that, the new method can accomplish the alignment with higher precision compared with the traditional methods, and the alignment accuracy of SINS is improved by 43%.In addition, the new method can be easily applied in engineering with fine effects.

The pseudo-spectral method can transform the continuous Optimal Control Problem (OCP) into a discrete Nonlinear Programming(NLP)Problem, and the solution of the original OCP is obtained by interpolation on discrete solution of NLP, which is a continuous function of the state and control.Firstly, the basic theory and general process of solving the OCP with the pseudo-spectral method are analysed.The consistency principle of the pseudo-spectral method is concluded, including the existence, feasibility and optimality.Then, methods are derived to ensure the consistency of the solution, and the applicability of the consistency principle and the method is verified through three case simulations.Finally, the characteristics of pseudo-spectral method are summarized, and general solution strategies of different scenarios and new research challenges are proposed.

Due to the underactuation and strong coupling of quadrotor UAVs, the visual servo effect of the camera fixed on the UAV airframe is not good.To solve the problem, an image moment based visual servo method is proposed.Firstly, the virtual image plane is built to select image moments as features, and then image dynamics is decoupled.Furthermore, the servo relationship between image errors and velocity is established by using the quadrotor dynamics model.Secondly, according to the dynamic decoupling of the quadrotor, the system is decoupled into position controller and attitude controller.The velocity control law is obtained by the position controller according to the servo relationship, and the desired attitude is obtained directly by means of dynamic expansion.Finally, a sliding mode controller is designed for attitude tracking control aiming at the attitude disturbance caused by the nonlinearity of the system.Simulation results verify the superiority and robustness of the proposed method.

The locating of remote (over 500 m away from the monitoring station) laser spot center in the actual environment with weak laser intensity and laser path interference was studied.Firstly, image preprocessing was used to suppress camera noise.Secondly, the laser spot intensity distribution model was constructed to segment the laser spot.Then, morphological corrosion operation was used to suppress the false segmentation, and the processed segmentation image was meshed.By calculating the credibility of the grid and the coordinate of the center of the grid, the combination of the coordinate of the center of the grid was completed to locate the center of the first-order spot.Finally, in combination with prior knowledge and the score of center possibility, the center point search was adjusted to complete the secondary positioning, and the laser spot center was obtained.In verification under actual scenario, the proposed algorithm can perfectly locate the center of the remote laser spot in an environment with weak laser intensity and obvious optical path interference.

Aiming at the shortage of on-site data samples in equipment maintenance evaluation, a multi-source data fusion method based on D-S evidence theory is established to make full use of the maintenance test data of the early stages or from other sources.Firstly, the traditional Bayes Bootstrap method is improved.The discretized maintenance test data from each source is more accurately fitted to its distribution parameters, then the information such as sample size and distribution characteristics is mined from the data source to construct an evidence.The evidence synthesized by D-S evidence theory is taken as a weight for constructing a multi-source data fusion model of maintenance.Finally, the maintenance test data of a certain type of tank is used to analyze the example, and the effectiveness of the fusion method is verified.

Aiming at the problem that the reconnaissance robot has poor immediacy in the real-time positioning and global map construction process because of the slow selection of key frames in the RGB-D SLAM algorithm,this paper proposes a SLAM key frame adaptive selection algorithm suitable for RGB-D cameras.The inter-frame similarity of the continuous images acquired by the camera may be changed due to the different motion states of the robot, which will affect the selection of key frames.In this paper, the neural network is used to construct the relationship model between the motion state of the robot and the step length of the key frame.The key frame adaptive selection algorithm based on the motion state of the robot is proposed to realize the adaptive selection of key frames.The experimental results show that, the algorithm reduces the amount of calculation of the system and improves the overall operating efficiency of the system.Therefore,the algorithm can effectively improve the running speed of the system while ensuring the accuracy and has strong practicability.

Aiming at the uncertainty and external disturbance of the longitudinal model of hypersonic vehicles, a nonlinear dynamic inverse controller with sliding mode disturbance observer is designed based on the terminal sliding mode control theory.Firstly, feedback linearization is made to the longitudinal motion model of the aircraft to realize the decoupling of the height and velocity channels.The nonlinear dynamic inverse controller is designed by using the pole configuration method.The model parameter uncertainty and external disturbance are regarded as “equivalent disturbance”.The sliding mode disturbance observer is designed to estimate the equivalent disturbance of the system, and it is fed back to the dynamic inverse controller for compensation.The simulation results show that the controller can track the command signal quickly and stably, has strong robustness in the presence of interference, and can meet the requirements of hypersonic flight control.

In order to eliminate the influence of the non-linear phase resulted from the non-symmetry of orthogonal wavelet on decomposition result, on the basis of zero phase filtering principle and the Undecimated Discrete Wavelet Transform (UDWT), we proposed a UDWT decomposition algorithm with zero phase characteristics, called Zero Phase UDWT (ZPUDWT).The new method can not only overcome the defects of Mallat algorithm, such as overlap of frequency and shift variance, which are caused by down sampling, but also eliminate the displacement and distortion phenomenon in decomposition results of conventional wavelet transform.Therefore, ZPUDWT is very suitable for weak feature extraction.Analysis of an example for weak feature extraction of bearings proved the effectiveness and advantages of the new method.

The phenomenon of light leakage may appear when a rugged LCD is operated in vibration environment.An effective method is proposed for analysis of the phenomenon.At first, a simplified rugged LCD model is used for making analysis to the finite element, and the optical property simulation software is used for calculating and comparing the transmittance of light in stationary state and vibrating state.Experiment is made to verify the effect, which explains the mechanism of white spot generation from the perspective of the transmittance changing of liquid crystal screen.In addition, the effects of different LCD modes and different thicknesses of rugged glass on the change of light transmittance in random vibration environment are also studied.The proposed method is helpful for improving the yield rate of rugged LCDs and reducing the production cost.

Airborne passive radar is faced with interferences of Doppler broadening direct waves and strong clutters, but the traditional Normalized Least Mean Square (NLMS) algorithm is difficult to eliminate such interferences.In order to solve this problem, a Variable-Step-Size Range-Doppler NLMS (VSS-RDNLMS) algorithm is proposed.The algorithm is improved on the basis of RDNLMS algorithm to eliminate clutters on the distance and Doppler synchronously.At the same time, the output error signal and system noise are used to control the step size.Therefore, the steady-state error of the algorithm is reduced on the premise of ensuring the convergence speed.The simulation results show that, under the same simulation conditions, the VSS-RDNLMS algorithm can obviously improve the clutter cancellation performance for airborne passive radar, effectively reduce the cancellation residual, and improve the SNR of target.

Aiming at the fault estimation problem of a class of nonlinear systems with disturbance, a new method combining H∞ robust observer with fault estimation filter is designed.Firstly, this method inhibits the influence of unknown disturbance and system noise by introducing H∞ performance index, and inhibits the influence of the rate of fault change on fault estimation error, thus to increase the rapidity and robustness of fault estimation.Secondly, the estimation fault is adjusted by the residual signal after suppression of the H∞ robust observer, and then the threshold value of the filter is designed, and the threshold is compared with the residual norm to determine whether the filter is started or not.The design of the filter further enhances the robustness of the system.Finally, the simulation results verify the validity and correctness of the proposed method.