This PDF file contains the front matter associated with SPIE Proceedings Volume 12759, including the Title Page, Copyright information, Table of Contents, and Conference Committee information.

A theory, namely the train derailment energy analysis theory, was proposed after the study. The theoretical knowledge of the train's self-excitation characteristics and feedback theory demonstrates that lateral instability can cause train derailment to occur. Vehicle derailment has been a difficult problem to solve so far, and in the derailment tests that have been done, sometimes derailments have not occurred even though the operating vehicles have greatly exceeded the existing evaluation criteria. In recent years, domestic and foreign railroad scholars have conducted a lot of research on derailment, but until now the derailment mechanism is still very vague understanding. For more than a century, the study of derailment problems has been limited to the static aspects of single wheel pairs and less to the whole vehicle. Train derailment should not only consider the static mechanics, but also involve dynamic derailment problems, especially at home and abroad after the train speed increase, the train on the track at high speed, dynamic and non-linear effects are becoming more and more serious, making derailment problems become more complex.

The current fuzzy control nodes for mobile obstacle avoidance of patrol robots are mostly set to the single-layer form, and the control speed is slow, leading to the increase of fuzzy control time consumption. Therefore, the design and verification analysis of the fuzzy control method for mobile obstacle avoidance of patrol robots in substations based on laser radar are proposed. According to the actual control requirements and standard changes, it is necessary to determine the basic control indicators of the patrol robot first, increase the control speed of the robot in a multi-level manner, deploy multi-level 3D fuzzy control nodes, establish a grid multi-level control matrix, and build a fuzzy control model for the moving obstacle avoidance of the laser radar patrol robot on this basis. Fuzzy control is realized by improved artificial potential field processing. The final test results show that: through the verification of four test areas in the substation, the final fuzzy control time of the patrol robot is well controlled below 1.5s, indicating that the practical application effect of this robot control method is better, the speed and range of control are further expanded, and the handling of robot obstacle avoidance and displacement tasks is more diversified and flexible. It has stronger pertinence and practical application value.

It is advantageous to use a robot in automatic winding, for solving the problems of low efficiency and low degree of automation of manual winding. In this paper, the trajectory planning algorithm of the robot is studied, and the trajectory planning of the robot is carried out by using the quintic polynomial interpolation. The simulation verification of Matlab is used to obtain the displacement, velocity, and acceleration curves of each joint of the robot, The observation results show that the curve fitted by quintic polynomial trajectory planning is smooth, and its speed and acceleration have no sudden changes, which ensures the stability of the moving process of the automatic winding robot.

In order to explore the feasibility of brain-computer interface in the weapon system application, an experience of automatic thrower model controlled by brain-computer interface is designed to compete the firing task. According to the lens of automatic thrower model fed back from the screen, the testee carries out the practical or imaginary actions, and then through the nonintrusive BCI equipment, load 6 kinds of EGG signal which are collected from the testee to the trained convolutional neural network. 9 kinds of signal controlling the actions of automatic thrower model would be output. Control the switching functions of automatic thrower model through setting to realize 9 actions including Forward, backward, left, right, turret left, right, muzzle up, down, and launch, so as to complete the firing task. The comprehensive accuracy rate of implementation is 78.83%. The result of this experience shows that the brain-computer interface has the feasibility in the military field, which is worthy of researchers to carry out deeper studies.

At present, the researches for bus priority are mostly limited to the background of fixed signal timing. And there are few studies on bus priority under the background of inductive control. In view of this, an inductive control method for bus priority based on dynamic green time constraint is proposed. Firstly, a basic background scheme is generated according to the minimum green time of phases. Then, the green extension time required for bus priority is calculated. Finally, a calculation model for green time constraint is established to dynamically allocate the demand green extension time and optimize the green time constraint interval of the phases under inductive control in real time. A simulation scenario is established by SUMO software. The simulation results show that the proposed method can improve the efficiency of bus travel while taking into account the traffic demand of private vehicles.

The behavior of spacecraft in tracking game is studied by using the theory of differential game. This study is based on the premise that both sides of the game have continuous small thrusts, and the terminal distance is set as the payoff function. Since the altitude of the two sides of the orbit is time-varying, the proposed model has time-varying characteristics. Finally, the problem is transformed into a two-point boundary value problem with high dimensional timevarying nonlinear. In order to solve the two-point boundary value problem, the multi-shooting method is usually adopted, but this method is very sensitive to the initial value problem, and the model has large error. To solve this problem, a hybrid algorithm combining multi-shooting method and particle swarm algorithm is proposed and numerical solutions are given. In the method, the initial value of the co-state variable is estimated by the particle swarm algorithm, and then it is brought into the multiple shooting method to obtain the solution of the problem. The simulation example shows that the hybrid algorithm has high calculation accuracy for the tracking game problem, and the calculation example finally gives the trajectories of both sides of the game.

In order to reduce the production efficiency of welding machine faults in the process of shipbuilding, a fault diagnosis model of welding machine based on hidden Markov chain is established in this paper. Firstly, based on the historical fault data of the welder, the fault diagnosis model can be used to select the corresponding data segments of fault states such as internal line break, power phase loss and sensor damage. Secondly, the effective data is used to train the hidden Markov model, and the parameters of each fault model are optimized and iterated. Finally, the experimental platform of welding machine fault diagnosis is built, and the welding machine fault diagnosis experiment is carried out. The experimental results show that the model can accurately diagnose welder faults, reduce equipment downtime and improve production efficiency by calculating the matching degree between welder real-time working data and each fault model.

Vacuum switch based on permanent magnet operating mechanism is widely used in power system. In order to keep the opening and closing time of permanent magnet vacuum circuit breaker consistent, the method of controlling coil current can ensure the stability of opening and closing time of permanent magnet vacuum circuit breaker. Firstly, the paper analyzes and calculates the dynamic characteristics of the opening and closing of the 40.5kV vacuum switch operating mechanism, and establishes a simulation test model. The improved hysteresis PID algorithm controller is used to detect and control the coil current and displacement characteristics of the mechanism in real time. The simulation test proves the feasibility of the controller in improving the mechanism action stability. To test the control algorithm, an intelligent controller is designed. The opening and closing experiments of the circuit breaker are carried out in different voltage ranges by selecting the energy storage capacitors with different capacities. The results show that the controller designed based on the above control principle can effectively control the travel track of the moving contact of the circuit breaker, so that the movement track of the moving contact is consistent with the movement track of the reference current curve; Under the selected coil reference current curve, the error of actual circuit breaker opening and closing action time is ≤ 0.35ms, and the dispersion can still be maintained at about ± 0.35ms under different ambient temperatures, which proves the effectiveness of adding this control system to improve the stability of action time.

The test equipment is used to measure the deformation of the shell of the phase change energy storage device during the melting and solidification of the phase change material and to monitor whether bulging occurs to ensure its structural safety. When the test device is used to simulate different orbit operating environments, due to its structure and process characteristics, this temperature control process has time lag problems, non-linear problems and various unknown interferences using conventional control methods. It is difficult to obtain satisfactory static and dynamic performance. To this end, a modified Smith prediction compensation control method is designed to improve the control speed and robustness by introducing a Smith predictor in the fuzzy PID control loop to compensate the time lag of the control in advance. Finally, MATLAB is used to program and simulate the system with the modified Smith prediction compensation control method, and the results show that the modified Smith prediction compensation control method is more superior compared with the traditional PID control method and fuzzy control method.

With the development of the times, the concept of modern warfare is constantly changing and precision strikes have become an important mode and means of combat on the modern battlefield. It is of great strategic value to research and realise hard target penetration munitions that can strike accurately and effectively destroy high-value targets. The key to the efficient destruction of hard targets by penetration munitions is the penetration fuze, the core of which is the initiation control strategy in high overload conditions. By investigating the current state of research on intrusion fuze initiation strategies, the future trends of layer counting identification and cavity counting identification are envisaged.

The ball mill grinding system in the mine has strong non-linearity, large time lag and coupling, and it is difficult to control, and the application of traditional control methods cannot timely and accurately control the mill grinding and grading process in practical applications. To solve this problem, a generalized predictive control algorithm with proportional integration with improved gray wolf optimization is proposed to control the grinding process of ball mill. Considering the feedback structure of proportional integral type generalized predictive control and the implicit decoupling and prediction function of generalized predictive control algorithm, the optimal control of the feed amount and graded overflow concentration of the ball mill is realized, and the parameters are tuned by improving the gray wolf optimization, so as to improve the performance of the control system. The simulation results show that the control method has good tracking and anti-disturbance ability, and can realize efficient control of the grinding and grading process.

In order to solve the problems of nonlinear time-varying and large instantaneous impact disturbance of hydraulic valve servo system, the application of fuzzy control theory in hydraulic valve control system is studied. According to the working principle of hydraulic valve servo system and the basic principle of fuzzy PID control, the MATLAB model of hydraulic valve servo system is established, and the application of fuzzy PID control technology in hydraulic valve system is researched. The simulation results show that fuzzy control theory can effectively reduce the overshoot of the system and increase the ability of the system to anti disturbance. Compared with the classic PID control algorithm, it is more in line with the practical application needs of high-precision positioning of hydraulic valves.

This paper mainly focuses on the coupling effect and decoupling strategy of the active magnetic bearing (AMB). First, the AMB’s operating mechanism is examined, and a linearized model of the electromagnet-rotor is built. Second, the state equation for the five-degree-of-freedom (5-DOF) electromagnet-rotor is obtained using this linearized model. Combined with PID control, a radial magnetic bearing control model was built using MATLABSimulink to observe its control effect. Finally, to solve the coupling problem, a feedforward-feedback decoupling approach is suggested. Via simulation, it has been demonstrated that this approach can lessen the coupling effect and increase the static stability of the system.

In this paper, an integrated control strategy integrating trajectory tracking and lateral stability is proposed for a distributed drive autonomous vehicle (DDAV). First, a model prediction controller is established based on a 3 degree-of-freedom (DOF) vehicle dynamics model and a path following model, and the optimal front wheel steering angle and external yaw moment are simultaneously obtained. Then the torque distribution controller optimises the distribution of longitudinal forces at each wheel with the tyre load rate minimisation as the objective function, and applies the interior point method to solve the optimal torque distribution problem. Finally, simulation experiments are carried out based on the Carsim-Simulink joint simulation platform. The results show that the proposed integrated control strategy can effectively control the lateral error and heading error in vehicle trajectory tracking, and ensure reliable lateral stability of the vehicle at the same time.

Aiming at the problem of dynamic performance degradation caused by disturbance of internal and external factors of the carrier when the stability and tracking system of the photoelectric stable viewing platform moves, this paper proposes an active disturbance rejection and integral backstepping control strategy, adopts the model-assisted reduced order LESO controller as the outer loop control of the system, adopts the design concept of integral backstepping algorithm, and introduces the virtual control quantity of the system. The basic control law composed of inner and outer rings can be suitable for the use of motor vector control system, so as to achieve accurate control of motor speed.

This paper proposes an anti-saturation dynamic surface control strategy based on position guidance to achieve safe and accurate positioning of engineering vessels in the context of restricted inputs in the anchor mooring system. By introducing a position guidance function to limit the rate of change of anchor cable tension and using a dynamic compensation system to compensate for input saturation, the engineering vessel can be quickly and accurately positioned. Simulation results show that the proposed anti-saturation dynamic surface controller effectively addresses the problem of input saturation in the mooring system, thereby improving system safety and effectiveness.

During the process of laying gravel, the inclination angle of the leveling barge needs to meet high-precision process requirements. However, the ballast anti-heeling system is susceptible to disturbance, which results in poor control performance. The use of adaptive feedforward PID controllers can directly compensate for load disturbances and eliminate the effects of interference. However, the valve pump switch is fixed at the beginning of the control cycle, and the fluctuation of the heeling angle is significant due to the load speed changes. To address the above issues, this paper proposes a control method that increases the degree of freedom of the valve pump switch time. The PSO algorithm is used to optimize the best switch time, and the switch time is shifted to the inside of the control cycle in the adaptive feedforward PID control strategy. Simulation results show that compared with the traditional fixed switch point strategy, the proposed strategy can maintain smooth fluctuations of heeling angle around the expected value when the load speed is constant or changing, and the RMSE is reduced without increasing the switch frequency.

In recent years, STATCOM/BESS (Static Synchronous Compensator/Battery Energy Storage System) has received widespread attention due to its superior power four-quadrant output characteristics. To address the problem of delayed reactive power withdrawal and voltage fluctuations in compensating traditional DC transmission under large disturbances, a fast and effective outer-loop control switching strategy has been proposed. Under small disturbances, a constant bus voltage control method based on PI control is adopted to provide real-time and complete compensation of the receiving end voltage, maintaining the receiving-end bus voltage at the rated value. Under large disturbances, the control mode is switched to open-loop control with constant reactive power, without the delay of the PI loop, enabling timely withdrawal of reactive power upon fault clearance and ensuring stable receiving-end voltage. The proposed control strategy was verified by simulation on the PSCAD/EMTDC platform, demonstrating its feasibility and effectiveness.

Aiming at solving the problem of long-time consumption in the process of testing the gun vertical stabilizer system, a mathematical model was established, and the fuzzy PID control algorithm was designed, and its simulation model was established based on the AMESim/Simulink joint simulation platform, and the control performance of the PID control algorithm and the fuzzy PID control algorithm for the artillery vertical stabilizer system and the loading system at a given sinusoidal signal was analyzed. The results show that compared with the traditional PID control algorithm, the maximum loading error of the system is reduced by 5000N, and the fuzzy PID control algorithm has a better control performance, which provides a reference for the research of load simulation system.

In this paper, a novel self-organizing radial basis function neural network (RBFNN)-based nonlinear model predictive control (RBF-NMPC) scheme is proposed. First, the RBFNN is initialized on the training data using clustering and extreme learning machine (ELM) algorithms, and it serves as a dynamic predictor of an unknown plant. In addition, an adaptive growing and merging strategy is utilized in the neural network so that the RBFNN can automatically adjust its structure. Second, an improved Levenberg-Marquardt (LM) algorithm with a fixed time window is used to increase convergence speed while tuning network parameters. Then, the optimal control signal is calculated by gradient method. Finally, the validity of the developed method is demonstrated by a simulation of continuous stirred tank reactor system.

Most of the current automatic planning and execution codes for the obstacle avoidance path of the patrol robot in the substation are single structure, and the planning scope is limited, which leads to the increase of the optimal time consumption of the automatic planning. Therefore, the research on the automatic planning of the obstacle avoidance path of the patrol robot in the intelligent substation is proposed. According to the actual planning requirements and standards, the basic data collection of the robot is carried out, and the obstacle avoidance operator is set. In the form of multi-objective, expand the actual planning scope, design the robot multi-objective obstacle avoidance path code, and based on this, calculate the fitness function, design an adaptive three-dimensional obstacle avoidance path automatic planning model, and use depth grid modification to achieve automatic path planning processing. The final test results show that after three stages and five independent unit areas of test comparison, the TR20 robot selected in the G intelligent substation has relatively high completion rate for the test task, and better control the optimal time consumption of automatic obstacle avoidance path planning to less than 0.5s, indicating that this kind of planning scheme will be more targeted and stable. The patrol robot can also avoid the area where obstacles appear faster and more timely, and design the shortest path, which has practical application value.

The interconnected operation of multiple microgrids can effectively deal with the power fluctuation caused by largescale distributed power supply access, and enhance the anti-interference ability of the system. A distributed autonomous coordinated control strategy is proposed for AC-DC hybrid multi-microgrid system. This strategy can reduce the dependence on communication system, realize power exchange among subgrids, and reduce the voltage or frequency offset of each subgrid effectively. Firstly, the reference voltage or frequency of each subgrid is established by the energy storage unit. Then, the bidirectional DC-DC converter (BDDC) or bidirectional AC-DC converter (BADC) of each subgrid adjusts the power output to ensure the reasonable power distribution of the system. Finally, the AC-DC hybrid multi-microgrid system simulation model was established in PSCAD/EMTDC, and different working conditions were simulated and verified. The simulation results show that the control strategy can realize the coordinated operation of multiple microgrids.

To study the chaotic phenomenon of fractional-order permanent magnet synchronous motor (FOPMSM) and its control problem, a strategy of decoupled adaptive backstepping chaotic control is proposed based on feedback decoupling control for fractional-order permanent magnet synchronous motor (FOPMSM) system to reduce the order, and the adaptive backstepping method is used to design the chaotic controller according to the decoupled system. The fractional-order permanent magnet synchronous motor (FOPMSM) system's stability is analyzed using the Mittag-Leffler stability theorem. The simulation results show that the decoupled adaptive backstepping control can bring the permanent magnet synchronous motor (FOPMSM) out of the chaotic state quickly and with a fast response.

A large amount of grid connected wind power will have a certain impact on the stability of the power grid system, so transmission system operators have proposed power regulation regulations for grid connected wind farms. In order to satisfy with this requirement, firstly, an proportional integral controller is constructed based on the deviation of grid grid commands and the power measure value of the wind farm. Secondly, based on the different operating states and power generation capacity of the wind turbines, the active power distribution scheme is designed. At last, the wind farm controller are modeled and simulated. The simulation results show that the wind farm power control scheme realizes the active power distribution of the wind farm according to the grid dispatching instructions, meeting grid demand.

The main work of this paper is to study exponential stability under the premise of T-S fuzzy model. First, by constructing a new suitable Lyapunov Krasovskii function and Jensen inequality, the uncertain disturbance problem with time-varying delay is solved. Secondly, based on linear matrix inequality (LMI), sufficient conditions for exponential stability of systems and controller design are established. Finally, an industrial example is provide to show the effectiveness and feasibility of the work.

This paper presents a new control method to synchronize multiple stepper motors with acceleration and deceleration. The S-shape velocity profile is chosen to generate a frequency look-up table. Adjust the look-up table by multiplying a factor by the frequency for each stepper motor in real time so that the tables have equal time length. In the meanwhile, the velocity still shows the S shape. Since this method only saves one look-up table in memory and needs simple calculation in real time, it can be used in low-end microcontrollers. The method was tested using three stepper motors on the Arduino platform and the result shows good time accuracy and good performance.

Aiming at the problems of large energy consumption, weak penetration and small coverage in the application of wide-area Internet of Things in home control system, a smart home control system based on ZigBee+NB-IoT heterogeneous networking technology is designed. With STM32 as the main control chip, the system uses sensors to collect temperature and humidity, light intensity, smoke concentration and other parameters, and uses ZigBee+NB-IoT technology to transmit the data to the client, so as to carry out intelligent control and management of home gateway. The test results show that the application of ZigBee+NB-IoT technology in smart home control system has the advantages of convenient networking, low power consumption, stable system connection and reliable operation, and has a good application prospect.

Conventional electrical equipment installation quality control methods mainly use EPLAN electrical drawing software for quality assessment, which is vulnerable to the impact of dynamic installation environment changes, resulting in low quality factor evaluation indicators. Therefore, a new electrical equipment installation quality control method needs to be designed based on machine learning. That is, according to the key points of the installation process, the factors affecting the installation quality of electrical equipment are analyzed, the installation status of electrical equipment is diagnosed using machine learning technology, and the hierarchical control points for the installation quality of electrical equipment are constructed, thus completing the installation quality control of electrical equipment. The case analysis results show that the quality factor evaluation index of the designed electrical equipment installation quality control method based on machine learning is high, which proves that its control effect is good, reliable, and has certain application value, and has made certain contributions to the subsequent electrical equipment installation optimization.

The adaptive tracking control issue for the unmanned surface vehicle (USV) with actuator failures and disturbances is investigated. More specifically, a fixed-time high-order sliding mode observer is designed to estimate the actuator fault. Furthermore, a non-singular sliding mode controller (NSMC) is designed to solve the singular problem that may occur in the sliding mode function under certain circumstances. And then, by using the Lyapunov stability theory, we prove that the proposed control method can ensure all signals in the closed-loop system are bounded and the USV can track the desired trajectory in a fixed time. At last, the simulation example of the USV operating in the fixed-time tracking strategy demonstrates the effectiveness of the control method.

The predictive control of Wiener systems involving modeling nonlinearities is of great importance for analyzing industrial processes. In this paper, a data-driven predictive control method for Wiener systems is proposed. The methods used in this paper are based on some common simple linear algebra tools such as least squares and QR decomposition, where no complex solution process is involved, resulting in high estimation accuracy and numerical robustness. the data-driven predictive controller is designed according to the subspace predictor. The simulation results reveal that the proposed method are more accuracy and stable in Wiener model predictive control.

This paper introduces the development of a certain type of law enforcement equipment and analyzes the optimization direction of the control system. Aiming at the random disturbance problem of the law enforcement equipment under special working conditions, an improved approach law is designed to solve the contradiction between the approach time of the sliding surface of the traditional approach law and the system jitter, so as to improve the rapid response of the system. Based on the improved reaching law, the speed sliding mode controller suitable for permanent magnet synchronous motor is designed, and the simulation results verify the effectiveness and feasibility of the proposed control algorithm, which has engineering guiding significance for the algorithm optimization of this type of law enforcement equipment.

With the appearance of high-power focal plane module, in order to meet the refrigeration demand of infrared optical remote sensing load under complex working conditions, it is necessary for space refrigeration controller to have high-precision temperature control effect. For the sake of improving the temperature control accuracy of traditional space refrigeration controller, a high-precision temperature control strategy is proposed. By designing a temperature measurement circuit with secondary amplification and low-pass filtering mechanism, high-precision temperature measurement data can be obtained as the temperature feedback input source of the system. Through experimental test, the static temperature measurement accuracy can be improved from 20mk to 5.4 mk by the secondary amplification and low-pass filtering module. By designing the incremental PID with limiting amplitude and speed, the algorithm realizes the fast convergence of temperature curve and small overshoot. The experimental results show that the temperature control accuracy of the high-precision refrigeration controller is ±0.025k, and it stabilizes within 7 minutes after being disturbed, thus realizing high-precision and high-stability refrigeration control.

A run-to-run optimal control strategy for batch processes is presented in the paper. Generally it is very difficult to build an accurate mechanistic model for a batch process. Because principal component analysis is powerful for the problems characterized by multivariable and strong correlation, principal component regression (PCR) models are employed for end-point optimal control of batch process from batch to batch. That is, based on the model, the proposed method exploits the repetitive nature of batch processes to reach the desired end-point properties by determining the optimal operating policy. The optimization problem in the method is solved by quadratic programming (QP). Simulation examples show that the run-to-run control method can improve the batch process performance when batch run repeated. It is also demonstrated that the run-to-run control performance is convergent even when process constraints exist.

Hydraulic technology has been widely used in machinery manufacturing, and synchronization control is an important issue in the field of hydraulics. With the continuous development of the aerospace industry, the application of double-cylinder hydraulic technology in modern machinery is becoming more and more widespread, metallurgical equipment, metallurgical equipment, machine tools and other high-precision actuators, as well as the application of double-cylinder electro-hydraulic servo systems, are subject to inconsistent movements. As a result of non-uniform load, leakage, friction, resistance, manufacturing precision, hydraulic coil difference, structural elastic deformation, and so on. Therefore, the study of hydraulic synchronization control technology is especially important. This paper presents an immune neural network based PID controller to improve the synchronization precision of a two-cylinder hydraulic system. On the basis of the above, a new design method for the valve control cylinder of a double-cylinder forging hydraulic press is presented. In order to overcome the drawback that the conventional PID controller cannot be adaptively adjusted, the immune algorithm is adopted to monitor the operation condition of the system, and the PID parameters are dynamically adjusted on this basis, so as to improve the timeliness and effectiveness of the PID control; in the immune algorithm, the accuracy of the adjustment function construction will have a great impact on the accuracy of its construction. The validity and validity of the proposed method have been verified by simulation experiments.

As power system see increasing application of big data and artificial intelligence, the new generation of power quality data requires cross-platform interaction, larger database support volume, and higher data processing speed, which makes traditional power quality data collection and processing methods difficult to adapt to the power grid development. This paper proposes a power quality Thompson tau algorithm data cleaning strategy based on LES feature extraction, first extract LES feature value of power quality data and cluster partition of the eigenvalues, and then, according to the power quality characteristics of the data partition, divide the sub-interval step flexibly for Thompson tau data cleaning and multipoint interpolation data filling. Finally, power quality data from a practical power grid is taken as an example and the simulation results verify the feasibility and accuracy of the proposed method.

This paper takes Nanjing online car-hailing as the research object, aiming to analyze and model the spatial autocorrelation of the passenger flow of online car-hailing. Python and ArcGIS software are used to obtain the temporal and spatial distribution of residents' travel and find out the hot spots. Using the spatial analysis tool of GeoDa software, the spatial correlation of online car-hailing passenger flow is obtained; Using the regression analysis function of GeoDa software, the passenger flow forecast model is obtained. Our conclusion is that the online car-hailing passenger flow has spatial autocorrelation and positive spatial correlation, and the increase of road length and the number of buildings, schools, parking lots and catering points in the region can promote the increase of online car-hailing passenger flow.

In order to improve the SLAM accuracy of robots in dynamic scenes and realize the tracking of dynamic objects, this paper proposed a dynamic objects recognition method based on reprojection error and depth estimation. Firstly, recognize objects by semantic segmentation and calculate reprojection error of objects between adjacent frames to filter out dynamic objects. Then, estimate the depth of objects to improve the recognition accuracy. Finally, complete SLAM and real-time estimation of dynamic objects pose according to static background and objects. Results of experiments on KITTI datasets show that the algorithm can effectively improve the recognition accuracy of dynamic objects and ameliorate the SLAM accuracy in dynamic scenes.

In the current climate of traffic congestion and frequent traffic accidents due to increasing pressure on transport, the birth and development of driverless cars is particularly important. However, as the technologies of unmanned vehicles are not yet perfect, it becomes more practical to carry goods at low speed - unmanned delivery - than to carry people at high speed. In order to achieve the goal of making full use of the existing road resources and improving the efficiency of distribution in logistics, this paper uses an intelligent algorithm - the Pelican algorithm - to optimise the cost of the unmanned vehicle and solve for its minimum cost.

For the sake of satisfying the needs of material handling in intelligent logistics warehouses and industrial buildings better and improving the deficiencies of existing material handling forms, in this paper, we design an intelligent material handling robot servo system that integrates DTW (Dynamic time warping) and LSTM (Long Short Term Memory) algorithms. The operator can command the robot to move in multiple directions according to the motion intention of the human body through the human manipulative mode based on the hand movement or the rod traction mode with weak traction. This project collects various motion pattern data through experiments. After preprocessing the data, features are extracted and repeated motion pattern recognition training and template matching are carried out to deeply learn and perceive human intention. We input the data collected in real time into the generated LSTM network model, identify the motion pattern in real time, and introduce the DTW algorithm for template matching verification. Through testing, the accuracy of motion pattern recognition can exceed 95% under the effective combination of the two algorithms, which is higher than that of the DTW and LSTM algorithms.

Aiming at the shortcomings of sparrow search algorithm, such as dependence on initial solution, sudden drop of population diversity in the late iteration period, and easy to fall into local optimal value, a multi-strategy algorithm was proposed to improve sparrow search algorithm. In the initialization stage of the algorithm, the standard Brownian motion is introduced to enhance the global search ability of the algorithm and improve the population diversity. In the later stage of the algorithm iteration, the improved artificial bee colony search strategy is introduced to accelerate the convergence of the algorithm, and the influence of local extreme points on SSA can be reduced to improve the ability of the algorithm to jump out of the local optimal solution. The experimental results show that the improved sparrow search algorithm has faster global convergence speed, better optimization accuracy and stability in function optimization.

Distributed power supply system has the advantages of high redundancy, flexibility and reliability. This paper starts from the distributed power system structure, using modular design means, based on solar panel array module and bidirectional battery charge and discharge module of distributed power system working mode analysis and control strategy design, finally through the Matlab simulation experiment to prove the effectiveness of the control strategy, for the distributed power system design and development, provides a useful reference.

Aiming at the problems of dwarf mongoose optimization algorithm with weak optimization ability and slow convergence speed, an improved dwarf mongoose optimization algorithm is proposed. Firstly, the dwarf mongoose optimization algorithm performs secondary reverse learning solution on elite individuals, and selects the best individuals to form a new population to improve the initial population diversity. Then, combined with the sparrow position update strategy in the Sparrow Search Algorithm, the dwarf mongoose location update method is improved to improve the local development ability of the algorithm. Then, the Gaussian mutation strategy is used to improve the algorithm's optimization ability and convergence speed. Finally, eight benchmark functions are compared with four different types of algorithms and three DMO algorithms with individual strategies, and the results show that the improved dwarf mongoose optimization algorithm has higher optimization accuracy, convergence speed and stability.

Air pollution is a major environmental issue that affects human health and the environment. In recent years, deep learning has been applied to the prediction of air pollution expansion with promising results. This paper provides a comprehensive review of the recent literature on the application of deep learning related algorithms to the prediction of pollution expansion. The paper focuses on the use of deep learning models such as Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM), and Hybrid models for air pollution forecasting. The literature review covers studies published between 2018 and 2023, and includes articles from various journals with high impact factors. The results of the reviewed studies show that deep learning models have outperformed traditional statistical models in terms of accuracy and robustness for air pollution forecasting. The paper concludes by highlighting the challenges and opportunities for further research in this area.

With the continuous development of deep learning, image classification and object detection technology have become one of the hottest research directions. With the help of deep learning, image classification and object detection have made breakthroughs, and have a wide range of applications in many fields, such as face recognition, autonomous driving, environmental art design and other fields. Supervised deep learning models require training with large amounts of labeled data and multiple rounds of iteration. Due to the limitations and particularities in some practical applications, it is difficult to obtain sufficient image samples. Therefore, this paper focuses on the problem of small-sample image classification and object detection. This paper deeply studies the training mechanism of meta-learning, improves and optimizes the EGNN algorithm based on graph neural network, designs a task-adapted feature extraction model for feature extraction of query images, and adjusts the contribution of different features of query images by using image feature learning to predict a set of parameters acting on the query image feature map, so as to make it more suitable for the classification task of small samples. At the same time, prototype node representation and feature fusion measurement modules are introduced on the original network, which are used to reduce the scale of graph nodes and improve the classification accuracy, respectively.

This thesis introduces a Python-based data processing system that utilizes wearable fitness devices to collect and analyze data on college students’ physical fitness training, progress, and overall health. The system stores the data in a MySQL database and analyzes it using statistical and machine-learning techniques. Its goal is to enhance the effectiveness of physical fitness programs in colleges and universities by providing trainers, coaches, and administrators with informed decision-making tools. The system’s performance will be evaluated using metrics such as accuracy, speed, and usability, with the ultimate aim of promoting a healthier lifestyle among college students.

Aiming at the problem of low tracking accuracy caused by the sparse deployment of UWSN nodes in the process of target tracking based on underwater wireless sensor network (UWSN), a target tracking algorithm based on the regulate of node height and the consideration of node mobile energy consumption is proposed. First of all, the optimal height solution methods are given when the target height is known and unknown. Secondly, a dynamic coefficient is introduced to adjust the weight between tracking accuracy and energy consumption, and an objective function is established to comprehensively consider the target tracking accuracy and node moving energy consumption, so as to realize the dynamic regulate of node height. The simulation results show that the algorithm can achieve good and stable tracking effect and effectively reduce the energy consumption caused by node movement.

The path-tracking method of the unmanned surface vehicle (USV) consists of guidance law and heading control. For solving the problem that the LOS guidance law leads to low path-tracking accuracy, this paper proposes an integraldifferential LOS (IDLOS). The proposed algorithm innovatively introduces the adaptive integral term and the differential term on the LOS. The integration term can improve the convergence speed of the USV. The differential term can reduce the overshoot. Therefore, the USV can quickly approach the desired path and improve the accuracy of path-tracking. To verify the accuracy of the IDLOS algorithm, simulation experiments are conducted in MATLAB in this paper. The results are as follows. The convergence time of the IDLOS algorithm is 1.4s faster than that of LOS. The overshoot of IDLOS is about 13 times lower than that of ILOS (integral LOS). IAE (integral absolute error) of the cross-track error of IDLOS decreased by about 0.89 compared with LOS.

With the popularity of heat treatment technology, the accuracy of temperature control and the correctness of control method selection are crucial. More and more researchers combine genetic algorithms, model predictive control algorithms, and particle swarm algorithms for the parameter identification of resistance furnace models have disadvantages such as local optimum and slow convergence speed. Therefore, the paper establishes an approach to the parameter identification of resistance furnace model based on the Improved Moth-flame optimization algorithm (IMFO) algorithm, selects the Particle Swarm Optimization (PSO) and Moth-flame optimization algorithm (MFO) as comparison algorithms, and obtains the parameter identification results of the above three intelligent optimization algorithms through MATLAB simulation tests and compares them with the actual resistance furnace output. It is found that the IMFO can identify the resistance furnace temperature sintering process more accurately, and the parameter identification results of the resistance furnace model using the IMFO are better fitted with the actual resistance furnace parameter variation curve. Compared with Moth-flame optimization algorithm (MFO) and Particle Swarm Optimization (PSO), the Improved Moth-flame optimization algorithm (IMFO) has the advantages of fast convergence, solving the local optimum problem, and improving the speed of finding the optimum, etc. It is very effective in the field of parameter identification of furnace temperature sintering process and better meets the needs of general industrial control.

Deep Convolutional Neural Networks (DCNNs) have been used in food image classification, segmentation, gradient identification and many other applications. There are many small segments in food industry, and it is very expensive and time-consuming to build a large training dataset for each of them. Transfer learning can play a role in this aspect by reusing knowledge learned from large datasets, like ImageNet, to target tasks where only small datasets are available. In this work, we conduct an empirical study on the transfer learning capability of DCNNs pre-trained with ImageNet on four small food datasets. We find that models perform better on ImageNet dataset also have better transfer performance for small target datasets; In two transfer learning approaches, fine-turning performs better than feature extraction; Data augmentation helps in transfer learning.

For the problems of long planning paths and slow convergence speed of the ant colony optimization (ACO) in mobile robot path planning, this paper improves the ant colony optimization and introduces the artificial potential field method (APF). Firstly, the initial pheromone concentration is differentiated to make the ants more directional at the early stage of the algorithm. Second, the heuristic factor in the heuristic function is dynamically adjusted to optimize the pheromone update strategy so that the algorithm can effectively distinguish the high-quality paths from the common ones. Finally, the distance parameter is added to the repulsion function of APF, the improved ant colony algorithm is fused with APF, and a fusion algorithm based on the improved ant colony algorithm-artificial potential field method (IACO-APF) is proposed. In the simulation experiments, the trajectories are optimized by the B-spline curve method, and the experiments show that the IACO-APF algorithm can obtain better solutions in a shorter time compared with the traditional ant colony algorithm and other algorithms mentioned in the literature.

This paper proposes a novel, improved discrete particle swarm optimization algorithm (IDPSO) to solve the multiobjective flexible job shop scheduling problem (MOFJSP) under dynamic perturbations. Firstly, the idea of variable neighborhood search (VNS) is fused to design three neighborhood structures to enhance the ability of the discrete particle swarm algorithm to jump out of local extremes. Secondly, a non-linear descent inertia weight is designed to help the DPSO algorithm better balance the global and local search capabilities. Furthermore, this paper redesigns the individual optimal as well as global optimal particle update methods of the algorithm for the characteristics of the MOFJSP problem. Then, an external archival solution set of elite strategies is designed to store the non-dominated solutions in the algorithm's solution process. Finally, the effectiveness of this algorithm is verified on the Kacem dataset; compared with other algorithms, the results show that this algorithm has a strong performance in finding the best.

Choosing the optimum route is the most critical planning content in economic tourism planning. This paper takes the coordinates of tourism attractions in Enshi as the research background, considers the problem of travel route planning as a tourism salesman problem, and solves the shortest circuit path for traveling salesman as the research object. In our research, we chose the genetic algorithm, the Ant Colony Algorithm, and the combination of the two algorithms and compared and analyzed them with the best parameters. The experiment shows that combining the ant colony and genetic algorithms can more effectively find the shortest circuit path in developing tourism planning in the Enshi Scenic Area. At the same time, the effect is obviously better than the results obtained by any of the algorithms alone.

Aiming at the problem of artifacts in the direct back-projection reconstruction algorithm of industrial CT, this article first uses Radon transform to obtain projection data, and then obtains the original function of the industrial CT image function through inverse Fourier transform, and performs convolution operation on the original function. Use the selection property of Dirac function to get the reconstructed image of industrial CT. The method in this paper can reduce a Fourier transform operation. The simulation results show that the method in this paper can effectively improve the blooming phenomenon of the tomogram reconstructed by direct back projection, and the reconstructed tomographic image is better.

This paper studies the problem of production scheduling in a discrete production workshop, which involves the selection of equipment and the sequence of processes. Firstly, two vector coding structures are used to represent the solution to the problem. Secondly, to solve the problem of difficult search in the case of many devices, this paper improves the music chair algorithm, which has a fast convergence speed. Then, the algorithm is combined with Pareto optimality and an improved fuzzy decision-maker to solve the multi-objective production scheduling problem. Finally, it is compared with the genetic algorithm, the particle swarm optimization algorithm, and the simulated annealing algorithm to solve the single-objective scheduling problem on the actual production scheduling data. The results show that the algorithm can quickly converge on the optimal solution. Compared with the non-dominated genetic algorithm II and the multi-objective evolutionary algorithm based on decomposition to solve the multi-objective scheduling problem, the results show that this method can solve the multi-objective scheduling problem better.

To address the issue of low accuracy in simulating the motion trajectory of water-floating garbage due to multiple factors, a method for simulating the drift trajectory of water-floating garbage based on Sa-LSTM was proposed. The method taking the drift trajectory of water-floating garbage in Lanzhou Section of the Yellow River as the research object, integrated multiple influencing factors through feature derivation and enhanced the memory and generalization ability of LSTM model by using spatial attention module, which further improved the accuracy of water-floating garbage simulation data. The experimental results show that the proposed method can effectively reduce the interference of multiple influencing factors on the simulation of water-floating garbage drifting trajectory, improve the accuracy of drifting trajectory simulation, and provide a method and location information support for the accurate management and management of water-floating garbage.

This paper researches a grid-optimized DBSCAN algorithm called Grid-DBSCAN. Different from the traditional DBSCAN algorithm, we introduce a square grid array with side length ε. The whole study area is divided twice into several grids to study the computation. In this algorithm, any data point only needs to make a distance operation with the data points in the range of four different 3*3 grids that it may be in. This significantly reduces the computational complexity and required runtime of the algorithm. In addition, the two different center positions of the grid partition ensure the transitivity of the clustering algorithm and reduce the possibility of data points being misclassified. The simulation results indicate that the Grid-DBSCAN algorithm can effectively handle various clustering conditions. Both the computational time and clustering performance are superior to those of traditional DBSCAN and K-means algorithms.

We propose a new obstacle avoidance method for mobile robots in unknown and dangerous environments. The method involves scanning the environment with a laser radar to obtain map data and building a grid map environment model, followed by initial path planning using an improved A* algorithm. To address the shortcomings of the A* algorithm, such as paths that are too close to obstacles and large turning angles, we propose combining the A* algorithm with an improved artificial potential field method. This method establishes a repulsive field between the robot and obstacles, ensuring that the robot's motion path stays away from obstacles. Simulation experiments have demonstrated that the obstacle avoidance mobile robot has excellent motion control capabilities.

In the electronic warfare reconnaissance system, interferometer direction finding has been widely used with its advantages of simple structure, high accuracy and high sensitivity. However, it also suffers from the phase difference ambiguity problem, so the accuracy of interferometer direction finding largely depends on the correct rate of phase ambiguity resolution. This paper proposed an adaptive ambiguity resolution algorithm for phase error equivalence space, which can adaptively accomplish the resolution of phase ambiguity with arbitrary baseline number and arbitrary baseline ratio, and the effectiveness of this algorithm was validated through engineering practice and simulation examples.

Wheat powdery mildew spore detection is one of the most common scenarios in the field of agricultural pest and disease detection. Since wheat powdery mildew spore images possess the characteristics of small spore targets, easy adhesion, shallow feature dimension, and many background clutter, it leads to the increased difficulty of wheat powdery mildew spore detection. By analyzing the above detection difficulties, a YOLOv5 wheat powdery mildew spore detection algorithm based on attentional feature fusion is proposed in this paper. Improvements are made on the basis of YOLOv5s. Firstly, an attentional mechanism-based feature fusion module AM-FF (Attentional Mechanisms-Feature Fusion) is proposed to strengthen the fusion of features at different scales, which enhances the learning ability of network contextual association by autonomously learning and selecting the optimal features through the attentional mechanism. Secondly, in order to enhance the shallow features and small target detection accuracy, we modify the original network structure to enhance the acquisition of shallow features and add fused reinforced shallow features in the neck network FPN (Feature Pyramid Networks). Finally, the SIoU (SCYLLA-IoU) loss function is introduced in order to increase the consideration of vector angle in the boundary box regression calculation, which effectively improves the detection accuracy. Experiments were conducted on the expert-labeled wheat powdery mildew spore dataset, and the final result of the algorithm experiment AP (Average Precision) was 94.23%, which was a large improvement compared with the original YOLOv5 detection algorithm vertically and also compared with the algorithm made by previous authors horizontally, and the experimental results proved the superiority of the proposed method in this paper.

The current conventional distribution network operation risk metric model mainly realizes risk metrics by calculating risk index values, which leads to poor metrics due to the lack of reasonable calculation of sub-project risk values. In this regard, a low-carbon distribution network operation risk metric model based on information entropy and threshold comparison algorithm is proposed. The risk information entropy indices of different subprojects are calculated, and the subproject risk values are summed up. Combined with the threshold comparison algorithm, the metrics are selected in terms of severity as well as probability to construct the risk metric system and finally the risk metric model. In the experiment, the proposed method is validated for metric accuracy. The experimental results show that when the proposed model is used for operational risk measurement, the risk value calculated by the model has a low error and has a more desirable risk measurement effect.

In this paper, for a battlefield material point siting problem, we do not adopt the traditional ideas, but treat it as a classification and solution module model, and choose to bring in a nested algorithm model to fuzzy cluster the points on the battlefield in order to divide them into multiple independent siting problems, and change the aggregation path of each node according to a heuristic algorithm to traverse all reach possibilities as much as possible. Dividing the battlefield area into n regions for clustering optimization: the fuzzy C-mean clustering algorithm using genetic simulated annealing algorithm for fuzzy clustering within the battlefield area, and then the immune algorithm to make a selection of all possibilities, can guarantee the accuracy rate while greatly reducing the model running time and achieving higher efficiency. (1) regional selection stage, based on the difficulty of attacking each point of the battlefield to establish an evaluation index system, the use of cluster analysis methods[1] to divide the battlefield area into n categories to achieve for different looking for the corresponding command center, material center target group; (2) Immune algorithm optimization, based on the immune algorithm to construct the corresponding model and design a modified genetic algorithm to optimize the optimal addressing[2]. After using the genetic algorithm with adaptive parameters, a more stable centroid is obtained in 500 iterations. Based on the above two-stage planning method, the empirical study results show that it can provide scientific personalized optimal addressing points for addressing class problems.

Make the existing product data as a new production factor to enable intelligent design. Based on the historical data of various shower ontology and molds in a sanitary ware company, this paper presents a method to calculate the similarity of case data for shower handle products, and uses similar products for mold reuse design. By combining the structural characteristics of shower ontology and molds, the shower design data is transformed into a characteristic information model, and Hierarchical clustering combining K-means algorithm is then studied to improve the accuracy of clustering results. Euclidean similarity algorithm is used to calculate the similarity between shower bodies. The experimental verification shows that the case data is divided into three categories and the similarity threshold of 0.7 can retrieve the product design with high similarity. The correlation algorithm provides the judgment basis for realizing the reuse design of product mold.

This paper studies the scheduling optimization problem of process shop. The gray wolf algorithm is used to solve the problem. The algorithm has fast convergence speed and simple structure, but it also has the disadvantages of easy to fall into local extremum, slow convergence speed in the later stage and poor population diversity. In order to overcome the above shortcomings, the gray wolf algorithm is improved to match the characteristics of the scheduling optimization problem. The improved method is to add Tent chaotic map, nonlinear convergence factor and Levy flight algorithm to improve the shortcomings of gray wolf algorithm.

For analyzing risk behavior based on multi-source time series data, a fuzzy Mann-Kendall method for autonomous or ADAS system equipped vehicles is proposed. With the combination of Mann-Kendall and fuzzy rules, more accurate detection of the data breakpoints, outliers, and other features are achievable, and the whole statistical patterns are fully considerable. It can eliminate ambiguity and uncertainty in driving data with assessment efficacy, can provide interoperability and compatibility, can support several data distribution with trend modes, and can modify inference rules to adapt to different data sets and application scenarios. For validating the suggested approach purpose, 200 sets of cut-in data are abstracted. By comparing the result with the actual triggering sign, the result shows the method's viability.

A speed profile optimization method based on improved particle swarm algorithm (IPSO) is proposed to address the problems of on-time performance, comfort and energy consumption in the process of automatic train operation (ATO) of high-speed trains. Since the traditional particle swarm algorithm (PSO) has many shortcomings, this paper adopts the nonlinear inertia weight decrement of cosine instead of the linear inertia weight decrement of the traditional particle swarm algorithm, as well as the adaptive adjustment strategy for the acceleration factor, which expands the global search range and avoids the local optimum, while accelerates the convergence speed. Taking the safety of train operation and the train dynamics model as constraints, a multi-objective operation optimization model for high-speed trains is established. The multi-objective operation optimization model of high-speed trains is solved by combining the improved particle swarm algorithm with the maneuvering conditions. Finally, the effectiveness of the improved particle swarm algorithm is verified through the simulation of the selected lines, which makes each performance index relatively optimal.

The map-based global localization technique leverages global descriptors to articulate overall features of point clouds, and facilitates vehicle localization through comparison with the map keyframe. However, this technique is capable of facing challenges when there is an offset between the vehicle position and map keyframes, making it difficult to achieve a match and thereby adversely affecting the accuracy and robustness of the localization system. To mitigate this challenge, this paper presents a novel approach for global localization that designs ring-wise encoding descriptors. This approach involves the integration of ring-wise elevation difference encoding and the shifting of the point cloud coordinate system to construct global descriptors that possesses rotational and translational invariance, thereby enabling rough localization. Additionally, the search range is dynamically adjusted established upon the similarity score of descriptors, and fine localization is achieved employing the branch and bound scan matching algorithm. The presented approach is validated on the KITTI dataset and JLU Campus dataset. Results demonstrate a high success rate in global localization, as well as excellent generalization and practicality.

The traditional processing of distribution substation heavy overload data uses manual method or association rule method, but its processing efficiency is slow and processing accuracy is low. In this regard, a BP neural network method for processing distribution substation heavy overload data is proposed. The BP neural network is constructed to build a model, extract data features, and train the model for data processing using large-scale data. In the experiments, the processing accuracy of the proposed model is verified. The analysis of the experimental results shows that the model processing efficiency is high and the data processing accuracy is high when the proposed model is used for data accuracy detection.

In the process of civil aircraft test flight, different test flight subjects need to be carried out in different natural environments, so it is necessary to conduct frequent transfer flights; In addition, there are functional and reliability test flights, route demonstration flights and other route flight tasks during the process of civil aircraft test flight. As the test flight aircraft has not obtained the type certificate, there are many special restrictions on airworthiness conditions. So it is necessary to make special route for test flight aircraft. In this paper, we present a fast planning algorithm for civil aircraft test flight routes, which takes into account the special requirements of test aircraft for routes, incorporates Dijkstra's algorithm, and can quickly construct routes suitable for test aircraft by combining existing route resources, and gives an implementation example of the algorithm. The results show that the route planned by this method is effective and reliable.

With the advancement of industry and the advent of Industry 4.0, the original traditional factories also need information and intelligent changes to adapt to the times. Our new system demonstrates a support system for mid/low-volume production, designed to help employees assemble different products with visual or tactile feedback in a traditional factory. The system records the entire workflow using multiple sensors (mainly image, motion, and electrical contact sensors) and learns to analyze each production stage through deep learning networks to set optimal values. The system compares the current state of the job with the learned target state and provides information to the employee when deviations occur so that corrections can be made on-site. After testing, the system has improved the product quality and production efficiency of the factory, meeting the standards of modern factories.

In view of the gradual diversification of users' needs for motors in practical engineering, it is required that the motor turntable can be stopped at any specified position according to users' needs, and in addition to that, the turntable can be controlled by users to achieve automatic and accurate positioning. Therefore, this paper designs a set of servo drive system, which is one kind of automatic control system, and can be used to control the angle of rotation (or displacement) of the controlled object, so that it can run automatically, continuously and precisely according to the change of the input command, while this set of servo drive has a closed-loop control system with negative feedback, which can realize the control of the corresponding function more accurately and quickly.

This paper proposed an adaptive road segmentation method based on deep learning for the typical features of variable video scenes and unfixed monitoring perspectives of general national and provincial trunk highways, and constructed an adaptive road segmentation algorithm based on Unet-VGG16 by comparing backbone feature extraction networks to achieve intelligent extraction of variable scenes and perspectives of highway pavement area. The detection accuracy of the algorithm was 95.77% and the recall rate was 96.32%, which achieved good detection performance.

Short-term traffic flow prediction is of great significance for road construction and operation. Aiming at the possible problems of missing data or insufficient sample data, the short-term traffic flow prediction of parallel road sections is studied using the traffic flow data collected in Shandong Province and the transfer learning idea. Firstly, the spatiotemporal correlation of the traffic flow of the parallel road section is analyzed, and the correlation coefficient is calculated; Secondly, the long-term and short-term memory network model is used to predict the section flow of expressway; Finally, the model based transfer learning is used to predict the cross section flow of parallel roads. The prediction results show that the average absolute error, the mean square error, the root mean square error and the average absolute percentage error of the transfer learning based prediction algorithm decrease by 2.56%, 3.59%, 1.81% and 4.41% respectively compared with the direct prediction of the short-term memory network model. In addition, the prediction model based on transfer learning can significantly improve the prediction efficiency, and can provide feasible methods for traffic flow prediction in areas lacking equipment collection or data.

Aiming at the problem of how to safely and efficiently pass the manned robot in the dense crowd, the path extension algorithm based on the mixed social force model are studied. Firstly, starting from the traditional social force model, a mixed social force model is established in the scene of the integration of pedestrians and robots. Secondly, to address the problem of blocking the passage of robots in dense crowd, the interaction behavior is designed to trigger the movement of individual pedestrians, so that the distribution situation of the crowd that hinders the passage of robot is changed. Finally, guide the robot through the crowd according to the expansion of path until it reaches the destination. Simulation experiments show that the proposed algorithm can improve the passing efficiency of robot in indoor dynamic and static dense crowd environment.

As a non-renewable fossil energy source, oil is closely related to the development of the national economy. The output of oil is equivalent to the blood of the national economy, so reasonable prediction and planning of the output of fossil energy are particularly important. Based on the 10-year daily oil production data of 62 oil wells, this paper comprehensively analyzed some main factors affecting oil production. Through data preprocessing and eigenvalue correlation research and analysis, 13 different characteristic attributes of oil wells were selected. The data were grouped and ungrouped by blocks, respectively. The polynomial regression algorithm model and the random forest model were developed and trained for the two kinds of data. The prediction results of different models were compared, and the models with the smallest error were selected as the quartic polynomial regression model and the random forest model. The prediction of the daily production of the oil well satisfied the requirement of accuracy.

Aspect-level text sentiment analysis is a more complete and specific task of text sentiment analysis, which determines the sentiment information expressed by a text with respect to each aspect. In recent years, the use of syntactic dependency trees as well as sentence structure information for aspect-level sentiment analysis has become more common, but they often ignore the role of sentence semantics involved in sentiment analysis and fail to combine them. In this paper, we improve and propose an interactive graph convolutional network (SInterGCN) model incorporating syntactic semantic structure for text-level sentiment analysis tasks. Specifically, a dependency graph is first constructed for each sentence in the syntactic relational dependency tree. This is followed by combining the graph of structural relations with the associated semantic graph, which is used to capture key aspect-oriented and contextual aspects. Furthermore, in order to interactively extract dependencies between aspect words and other aspects, an aspect-oriented GCN is used to model the representation learned by the aspect-oriented GCN based on the inter-aspect graph. Thus, the model can be aware of important contexts and aspect words when interactively learning aspect-specific sentiment features. Experimental results on four relevant benchmark datasets demonstrate that our proposed model outperforms the original interactive approach, achieving superior performance and validating the effectiveness of our model.

Convolutional neural networks perform extremely well in current image classification tasks based on deep learning. In our research experiment, we found a flaw and made a simple but extremely effective improvement. In this article, we used the traditional ResNet and VGG16 as the benchmark networks. By parallelizing multiple networks into a multi-branch structure model, we designed a multi-branch structure classification header, arranged in parallel order into a whole model, and sent the entire image into the detection header of the multi-branch structure. Through different convolution kernels, we obtained the category scores of multiple categories, and output the highest score. Adding a multi-branch architecture as an external module to the target detection network has been experimentally found to improve performance. Our experiments on cifar-10, cifar-100, and ImageNet 2012 datasets have shown that our method can achieve extremely high classification network prediction accuracy under extreme conditions. Under conventional conditions, the accuracy rate also reaches 97%. Applying related modules to target detection networks also has relatively improved performance.

Moving object tracking plays a very important role in computer vision. With the advancement of science and technology, effective tracking of targets in various scenarios is a hot research direction in many fields. The emergence of fisheye cameras provides a new research direction for target tracking of video streams. With a fisheye lens, a fisheye camera can capture a wider range of images at once and obtain more information. In this paper, the inter frame difference method (IFD-method) and optical-flow method (OF-method) commonly used in the target tracking algorithm are mainly studied to realize the tracking of moving targets in fisheye video streams. Experiments show that the IFD-method, especially the three-frame IFD-method, can effectively extract the target area, while the OF-method can realize the real-time tracking of the feature points on the moving target in the array-to-array video stream.

It is mainly used in the tennis field to accurately identify the current position of the tennis ball through the Opencv cascade algorithm. Use the A * algorithm to move the robot to the location of the tennis ball and plan a reasonable route. When reaching the tennis position, use a gripping mechanism to pick up the tennis ball and throw the picked up tennis ball at a reasonable position for the trainer to achieve the purpose of man-machine training. In addition, the car is also equipped with a Bluetooth control method, which allows you to remotely control the robot using APP, and observe the car route and constructed map on the PC. Experiments have proved that the vision system can effectively track tennis balls, and the robot has many functions and stable performance.

Small size, high efficiency, low loss, these are the advantages of permanent magnet synchronous motor. Based on the disadvantages of the traditional high-frequency square wave injection method, such as strong sensitivity to measurement error, poor robustness and sampling delay, an immune algorithm of optimized filter is proposed and a proportional resonance controller is adopted. Compared with the traditional PI controller, proportional resonance control improves the control precision and robustness of the system. Finally, the simulation is realized in Matlab.

To address the issue of large torque ripple in traditional model predictive torque control for permanent magnet synchronous motors, we propose a torque-flux hysteresis-based model predictive voltage control approach. Firstly, a mathematical model is formulated in a two-phase rotating reference frame, and the computation process for the suggested model predictive voltage regulation technique grounded on the torque-flux hysteresis is delineated. To reduce the switching frequency, a spatial vector modulation technique and a hysteresis-based predictive torque-flux control are combined. To minimize the impact of parameter variations on model predictions, the least squares method is used for motor parameter identification. The simulation results indicate that the proposed method is capable of efficiently decreasing both switching losses and torque ripple, while simultaneously ensuring satisfactory system performance.

With the continuous development of our economy and the rapid development of national infrastructure and computer industry, the development of discharge control system for concrete mixing station plays an important role in improving production efficiency. As the core module of concrete mixing station, unloading system of concrete mixing station has shortcomings of high labor intensity, slow speed and easy to occur accidents in manual operation. Therefore, the research and development of unloading system of concrete mixing station based on machine vision has important application value. As an important component of the concrete mixing station system, this paper analyzes the problems of traditional mixing station unloading systems based on the mixing station structure and unloading methods, and uses visual algorithms to perform mixing station unloading operations and compare them with manual operations. Improve the efficiency of system unloading, increase the intelligent diagnosis function of the system, and reduce enterprise operating costs.

The target detection by fusion of LIDAR and camera can effectively improve the accuracy of target detection and recognition. The current fusion algorithm mainly stays in feature-level fusion, and the real-time performance is poor to be deployed in mobile, this paper proposes a target-level fusion scheme to fuse the target frames detected by two sensors based on the overlap, and the experiments show that the method can effectively improve the detection rate and classification correct rate.

Campus access management is an essential component of school security. As the notion of intelligent campuses progressively evolves, there is an increasing appreciation of the significance of the campus access control system. Currently, there are two face recognition-based access control schemes available. On the one hand, the cloud server employs online authorization delegation to grant access, while on the other hand, the edge device utilizes local authorization to regulate access control. However, due to many school staff and access control terminals, the current method still needs to improve authorization delays and parallel server pressure. Therefore, this paper proposes an online verification priority strategy based on cloud-edge collaboration, which fully utilizes the characteristics of cloud servers and edge devices to solve the above-mentioned problems.

To realize the robot's visual autonomous localization in crowded scenes, the methods of camera observation position optimization and environmental visual invariant information extraction are studied. Firstly, a visual observation position model is developed to achieve efficient extraction of visual invariant information in pedestrian interference, and the visual localization camera observation position optimization method is designed. Secondly, fusion of point-line feature and localization algorithms based on line feature constraints in visual invariant information are investigated. Finally, experiments are conducted in public data sets and airport terminal environments to verify the effectiveness and practicality of the method.

In recent years, Visual-Inertial Simultaneous Localization and Mapping (VI-SLAM) has become a prevalent technology for 6-DOF state estimation. Current VI-SLAM approaches achieve advanced localization accuracy in the air while becoming brittle in the underwater environment due to broad untextured areas, inadequate illumination, and limited visibility. Additionally, existing underwater visual-inertial datasets were usually collected in open seas, making it challenging to obtain ground truth for the tracking of the robot, which limits the evaluation and development of VI-SLAM algorithms for underwater environments. To address these challenges, we proposed TagVins, a VIO system that integrates artificial features with natural features to construct reliable frame associations. Moreover, an efficient visual-inertial fusion algorithm based on the extended Kalman filter is presented. And a new underwater visual-inertial dataset, with groundtruth trajectories obtained via rigid structure and reliable localizations in the air, is presented to facilitate the further development of VI-SLAM algorithms in the underwater environment. Our evaluation results on the proposed dataset demonstrate that TagVins outperforms existing state-of-the-art VI-SLAM algorithms.

In the field of unmanned ships, which has attracted widly attention in recent years, the small terminals distributed on the huge ship need to independently judge the collision risks. However, the performance of terminal devices is significantly limited to ambient noise which affect the docking guidance. In this paper, Adaptive Intelligent Particle Filter (AIPF), an assisted tracking method based on Intelligent Particle Filter (IPF) is proposed to smooth and continuous state observation, thus effectively eliminating abnormal and leaping data. Specifically, this method regards the coordinates of the target in a period of time as a process with stationary independent increment, and the possible motion model is explored from an engineering perspective, so as to optimize low weight particles and adjust the number of particles dynamically. The experimental results on the ARMv7 platform demonstrate that even under the harsh conditions of computational power, the suggested method can still reduce particle degeneracy, improve the performance of particle swarm to better achieve auxiliary tracking.

Aiming at the complex mechanism of traditional mining accidents, various influencing factors, and difficult to predict accident risks, this paper uses the text description of the accident occurrence process to extract the unsafe actions of personnel and equipment that lead to the occurrence of accidents, and establishes a prediction model for the level of hidden dangers in coal mine accidents through in-depth learning methods. Select 426 coal mine accident processes as a dataset to train and validate the prediction model, with a ratio of 7:3 for the number of training and validation accidents. The research results show that the accuracy rate of the model in predicting the level of coal mine accidents reaches 83.53%, which can provide a theoretical basis for mining production management and safety decision-making to a certain extent.

With the rapid development of internet technology, cloud computing provides a new path for the increasing demand for information. To improve data storage capacity, balancing the load of data read and write nodes is of utmost importance. Research on improving node information transmission and balancing load storage by utilizing write load and bandwidth. The experimental results show that the overall standard deviation of file distribution in the system has decreased by more than 80%, and the coefficient of variation of file distribution in the system has decreased by more than 85%. The efficiency of disk space utilization is sufficient, and the utilization variance of the optimization algorithm is below 4. The improved algorithm reduces the time required to complete tasks to less than 1800 seconds, and as the data increases, the increase in task completion time is slow. The improved model can effectively balance the load of distributed hash tables.

There is a great correlation between data and machine learning is essential to explore methods for analyzing data. As an important component of machine learning prediction and classification, logistic regression algorithms are widely applied in various disciplines. This study delves into the theoretical connotation and algorithm steps of logistic regression algorithm, and selects enterprise samples as calculation examples to construct the model and analyze the algorithm prediction, in order to gain a deeper understanding of logistic regression algorithm. The results show that the accuracy of the constructed model reaches 80%, and the area below the ROC is 0.82. The prediction model is better, providing a new method for predicting the development of enterprises.

The wireless rechargeable sensor network mainly consists of a data processing unit, a sensor (data acquisition unit), and a mobile charger (energy supply unit). For the wireless rechargeable sensor network to work properly, the mobile charger needs to charge the sensors periodically through the route given by the data processing unit, but the mobile charger incurs unnecessary energy loss on the charging journey. This paper discusses the problem of route planning and the minimum capacity of sensor batteries for single versus multiple mobile chargers. First, the TSP and MTSP problems of minimum energy consumption of mobile chargers over the distance are solved by modeling them with Monte Carlo optimization simulated annealing algorithm, and improved genetic algorithm, respectively. Then, the problem is further transformed into a linear programming and multi-objective optimization model to solve for the minimum capacity of the battery. In addition, computational experiments using the LKH algorithm and CA learning methods are conducted in this paper to propose optimization scenarios for the route planning model.

In order to reduce the cost of intelligent power logistics system and improve the reliability under complex working conditions, a model reference adaptive speed sensorless intelligent power storage control strategy based on super spiral sliding mode is proposed. Design the warehousing and transportation process of electric power materials, without personnel participation, the system can complete the extraction and storage of goods by itself; Taking the voltage and current of permanent magnet synchronous linear motor as input, the superspiral sliding mode was used to replace the PI ring in the traditional adaptive law, and the model adaptive velocity observer was designed to estimate the rotor velocity and displacement; Simulation results show that the proposed strategy has fast response speed and small observation error, and can realize accurate identification of system motion information.

The insulator images taken by intelligent unmanned inspection equipment often have many small insulators in large-resolution scenes, and sparsely distributed self-destruct defects with complex backgrounds. These phenomena have brought great challenges to the efficient and accurate insulator recognition and defect detection based on insulator images. Therefore, this paper builds a large insulator image database, and proposes an insulator detection method through improving the YOLOv5 model by adding the smaller detection layer and optimizing the loss function to accurately classify and identify small insulators and detect their self-explosion defects. The proposed method can achieve a recognition accuracy rate of 97.8%, a recall rate of 96.2%, and a self-explosion defect detection accuracy rate of 95.1% in the constructed insulator database. The experimental results demonstrate that the proposed model has outstanding performance in terms of insulator recognition and the self-explosion defect detection, which can improve the efficiency of intelligent inspection of power grid system in practical application.

At present, ADCP is the most effective method for measuring multi-layered streamflow profiles internationally. Therefore, this paper is dedicated to presenting the research progress of ADCP measurement methods in a more systematic and complete way, focusing on the application of bio-visual mechanisms in ADCP measurements. Based on the comparison of traditional river flow measurement methods and ADCP river flow measurement methods, the current status of ADCP applications at home and abroad is analyzed. When ADCP is used to measure river flow, for a certain measurement section cannot accurately reflect the shortcomings of river flow data, based on the use of bio-vision mechanism, the use of visual neural computation is considered to process the river flow and the resulting section data, etc., while using The visual perceptual field adjustment mechanism for global or local features of the image makes the acquired data more accurate to estimate the data of the whole river flow.

At present, photovoltaic system transfers solar energy to electricity and has become a fundamental component in electronic power supply module. However, the occurrence of system faults extremely influence the stabilization and effectiveness of photovoltaic system. Additionally, existing researches on photovoltaic system faults detection are primarily concentrated on the voltages signals or mathematical analysis models, which leaks the detection accuracy and costs extraordinary computation times. In this paper, we utilize deep learning method to obtain the features from input system data, which contains system fault information. Indeed, the trained model can distinguish the faults nodes from these features and response faults information. Subsequently, we devise loss function to improve the model learning rate and enhance the detection accuracy. From our extensive simulation results, we can conclude that our devised method can successfully identify the faults nodes in photovoltaic system with reasonable computation costs and precise detection results through comparing with traditional detection methods.

An intelligent identification method of 10kV cable insulation fault under damped oscillation wave voltage is designed. Control the amplitude of damping wave, infer the equivalent circuit, confirm the reflection law of electrical signal, measure the distance of fault points through classification method, distinguish the fault points, and realize the intelligent identification of 10kV cable insulation fault. The experimental results show that the proposed identification method has high accuracy in the analysis of the current flowing through the main insulation of the cable, and the waveform at the breakdown voltage of 20kV and 10kV is basically consistent with the actual waveform.

In this paper, a video gesture recognition intelligent home control system based on FPGA is proposed. The system can realize the control of intelligent home devices by recognizing user gestures. The system adopts the latest FPGA technology and deep learning algorithm, which can realize the real-time analysis and processing of gestures, and convert them into the command control of smart home devices. The program is developed by using Pango Design Suite software. The simulation results verify the correctness of the system. The system not only has a broad application prospect, but also provides a new technical route and solution for the development of smart home.

The massive multiple-input and multiple-output (MIMO) system based on channel state information (CSI) is the core technology of next-generation communication. As the complexity of the CSI matrix gradually increases, CSI feedback becomes more challenging. CSI feedback technology based on deep learning (DL) has been successful in frequency-division duplex (FDD) MIMO systems. In this paper, we propose a complex-valued lightweight neural network MADNet for CSI feedback. The network is based on an encoder-decoder structure, which enriches the information extraction of the CSI matrix by adopting channel information extraction modules and coordinates information extraction modules, which reduces the computational complexity of the network by using lightweight convolution. Experimental results show that our network outperforms most CNN-based network architectures at multiple compression rates and performs significantly better in both indoor and outdoor scenarios.

Vector hydrophone is a new type of underwater acoustic sensor, which is one of the hot spots in the field of underwater acoustic technology in the 21st century. In order to explore the feasibility of vector hydrophone circumferential array in engineering applications, a 16-yuan composite vector hydrophone circumferential array is designed and manufactured in this paper. The array diameter is 1.92m and the working frequency band is 20~3000Hz. The experimental results show that the array has better target azimuth estimation ability than scalar circular array.

The conventional target feature matching method of remote sensing image of power tower mainly uses deformable pooling framework for sampling processing, which is vulnerable to the dynamic deformation of geographical space, resulting in low registration rate. Therefore, a new target feature matching method of remote sensing image of power tower needs to be designed with the support of grid coding, that is, combining the principle of thermal infrared image. The target data of remote sensing image of power tower is processed, and the target feature matching algorithm of remote sensing image of power tower is designed on the basis of grid coding support, so that the target feature matching of remote sensing image of power tower is realized. The experimental results show that the designed remote sensing image target feature matching method for power tower has a high registration rate, which proves that the designed remote sensing image target feature matching method has good matching effect, accuracy, and certain application value, and has made certain contributions to improving the quality of remote sensing images in various fields.

Traction systems provide the traction power of high-speed trains. Because the complex operation mechanism of train under actual working conditions and the measured data are nonlinear and non-Gaussian, and the sampling frequency of the sensor is high in actual working conditions. Directly using a neural network or the multivariate statistical method is challenging to obtain the ideal fault detection (FD) result. Therefore, this paper proposes a data-driven method based on broad learning system (BLS) and convolutional neural network (CNN) assisted principal component analysis (PCA). Two neural networks are used to enhance the robustness of the algorithm, so that the proposed method has better fault detection ability in nonlinear and non-Gaussian systems. The advantage of this method is that it does not require the establishment of a complex high-speed train data model. Instead, by processing the collected data, the proposed algorithm can ensure good fault detection capabilities. Finally, the effectiveness and feasibility of the proposed method are verified on the simulation platform of traction drive control system (TDCS).

The proportional-integral-derivative (PID) controller is the most common controller in industrial production. However, the PID controller cannot change own parameters once it is designed, so that the conventional PID control method is difficult to cope with complex nonlinear systems. In order to overcome the shortcomings of classical PID controller, an optimized control method based on the back propagation (BP) neural network has been proposed. At the same time, the particle swarm optimization (PSO) is combined with the BP neural network to avoid BP neural network falling into local optimization. The validity of combined PSO-BP PID optimization algorithm is verified by a direct current motor servo control system. Results show that comparing with the PID, the overshoot of PSO-BP-PID to the response of step signal is reduced by 88.84%, the setting time and rise time is reduced by 61.88% and 34.47% respectively. The PSO-BP PID controller for direct current motor servo control system is more reasonable and effective than PID controller.

To address the problems of low classification accuracy in bird sound recognition algorithms, a bird sound recognition method based on FBank feature extraction is proposed. FBank features are extracted from bird audio data, trained with Transformer model, and confusion matrix of the model is obtained by adjusting learning rate and optimizing loss function. Tests have shown that the proposed method has good accuracy in identifying 10 types of bird sounds.

The main purpose of this paper is to improve the accuracy of the prediction of aircraft delay time. Using K-Medoids clustering algorithm to cluster the time intervals between arrival and departure time based on historical data, and then use the principal component analysis to reduce the dimensionality of the data. Finally, The Bayesian Classifier is used for classification processing to predict the category of the time interval to which the data belongs. The improved Bayesian Classifier has greatly improved accuracy and reduced errors, then we can use this data to predict the actual departure time of the aircraft. It is convenient for us to predict the delay time of the aircraft, and the prediction result is more accurate.

During the actual construction of a cutter suction dredger, the control of mud density and flow rate in the mud pipeline are two important parameters that affect the dredging output and the safe transportation of mud. Mud density is affected by many factors such as the speed of the cutter, the speed of the cutter and the speed of the submerged pump. Real-time prediction and control of mud density are important means to achieve efficient and safe dredging. Using actual construction data, an optimization method based on Sparrow Search Algorithm (SSA) is proposed to predict mud density. LSTM and ELMAN models are established and compared with the optimized SSA-LSTM and SSA-ELMAN models. The results show that the SSA optimization method can predict the density of cutter suction dredger with high prediction accuracy and stability.

Accurate traffic flow prediction results can help to reduce traffic congestion, optimize travel routes and improve road capacity, etc. Therefore, a short-term traffic flow prediction model based on GWO-Attention-LSTM is proposed, which can deeply explore the single-factor information features of traffic flow. The model is based on Long Short-Term Memory (LSTM), and Attention mechanism is added to enhance the focus on important information. The initial weight parameter values of Attention mechanism are optimized by Grey Wolf Optimizer (GWO), so as to mine and predict the time series of traffic flow. The traffic flow data of the actual road is simulated and analyzed. The results show that the RMSE values of GWO-Attention-LSTM model decrease by 1% and 2% compared with Attention-LSTM model and LSTM model, respectively. It proves that GWO-Attention-LSTM model has lower prediction error and better model performance by adding Attention mechanism and GWO algorithm.

With the expansion of urban rail transit network, the passenger demand becomes more uneven and complex. An accurate passenger flow forecasting of urban rail transit is important for both operation planning and service quality improvement. This paper proposes a passenger flow forecasting method based on prophet-GRU combined model. Firstly, Seasonal- Trend-Loss decomposition is used to decompose the urban rail transit passenger flow time series into trend, seasonal and residual terms. Then, on the basis of STL decomposition, prophet model is established for the trend and seasonal terms, GRU model is established for the residual terms as well. Taking the Nanjing Metro Line Network as the example, all the prediction results are obtained by linear integration. The results show that the weighted accuracy of passenger flow prediction reaches 96.13% for 159 stations. The root mean square error (RMSE) of the combined Prophet-GRU model is at least 49.69% lower and the mean absolute error is at least 19.00% lower compared to the ARIMA, GRU, and Prophet models.

Accurate and real-time bicycle demand forecasting helps shared-bicycle operators to allocate shared-bicycle more reasonably and efficiently. Thus making full use of resources and providing convenience for people to the greatest extent. We use convolutional LSTM to form Encoding-Forecasting Structure, and use open-source mobike data set of Beijing to train our model. By evaluating the model through these metrics RMSE,MAE,R2, we found that the Encoding-Forecasting Structure has better fitting ability and accuracy.

Excessive delivery of shared bicycles will lead to a waste of resources. In order to improve the utilization of resources, it is necessary to forecast the demand for shared bicycles. Most of the previous papers only focus on demand data's temporal or spatial correlation. Our paper uses the T-GCN model, which takes into account not only temporal but also spatial correlations. Our research achieves better results on the problem of demand forecasting.

With the rapid economic development and the advancement of urbanization, the structure of urban road networks is becoming more and more complex. It is particularly important to provide real-time and accurate traffic flow forecasting for traffic management departments. Aiming at the importance of traffic flow parameter prediction in intelligent transportation systems, this paper proposes a traffic flow prediction method based on the Residual Network (ResNet) model. Through the processing of taxi GPS trajectory data, effective and complete traffic data is obtained. Considering characteristic factors such as weather and holidays, the spatial characteristics of traffic flow are modeled by residual convolution units for different time granularities. The different branches and regions are weighted to predict the final traffic flow for each region. The experimental results show that the design model considers the complex dynamic space-time characteristics of the region, and the predictability is improved compared with other models, which is an effective traffic flow forecasting method.

In order to study the problem of short term prediction of parking demand in the region, long short-term memory (LSTM) neural network model has been used to predict the corresponding parking demand at subsequent time points based on the historical parking demand changes. Using the historical order data of on-street parking in Guilin, the data were organized into time series of 15-minute periods, and processed by noise reduction using wavelet threshold denoising method to train and test the model. The experimental results show that the prediction accuracy of the LSTM model is higher (MSE=11.588, RMSE=3.404, MAE=2.079, R2=0.945) compared with the traditional back propagation (BP) and wavelet neural network (WNN) neural network algorithm, and the prediction results are more similar to the real results. It can be seen that the use of LSTM recurrent neural network is effective and feasible for short time forecasting of berth demand in the region.

With the development of flight control technology, near-surface vehicle will play an important role in the future naval battle. In order to solve the complex problem of path planning for near-surface vehicle in island and reef environment, a method of dividing island and reef areas by using improved clustering algorithm and generating feasible path planning by using binary tree algorithm is proposed. Simulation results show that the proposed method can solve the problem of path planning in dense island and reef areas, and has high feasibility. It has certain practical value and reference significance for the study of path planning of near-surface vehicle.

In order to solve the problem that the current short-term traffic flow prediction algorithm ignores the multi-dimension of traffic flow data, which leads to low prediction accuracy, this paper introduced the tensor decomposition theory and established a dynamic tensor model based on the multi-time dimensions of week, day and period. This paper designed a short-term traffic flow prediction algorithm based on Tucker Decomposition on Gradient (TDG), and took the traffic flow velocity data of Dujiakan section of Beijing-Hong Kong-Macau Expressway as an example for empirical verification. The result shows that the algorithm can quickly achieve good prediction results based on less historical data, and can effectively achieve traffic flow prediction for weekdays and weekends, and the average mean absolute error (MAE) is about 4.4%. Compared with the classical prediction algorithm, it can show better prediction accuracy.

The magnetic levitation track uses the contact rail and the collector shoe to form a loop to provide power to the car. When there are hard points on the contact rail, they will affect the power supply process. In order to effectively supervise the height of the contact rail hard points and carry out railway maintenance in a timely manner, the effective prediction of the contact rail hard points height are achieved by constructing a BP neural network dual-channel regression model. According to the signal characteristics, determine the parameters design the parameters of the BP neural network model. Reasonably divide it into training, verification and testing data sets. Finally complete the prediction of the hard points height based on the trained model. The results show that the proposed method for predicting the height of the hard points of the contact rail based on BP neural network can realize the quantitative detection of the height of the hard points under different working conditions.

In this paper, temperature-time models with different characteristics, melting and crystallization versus temperature and time as well as melting rate versus temperature and crystallization rate versus temperature are developed around the design problem of mold flux. Finally, the dynamic changes between successive photographs of the melting and crystallization processes of the nearby die flow are successfully analyzed and quantified by means of RGB analysis, ORB algorithm and Fourier series fitting and finally, the temperature and time variations, the melting and crystallization processes of the die melt as a function of temperature and time, and the melting and crystallization kinetics of the die melt, i.e., the relationship between temperature, melting rate and crystallization rate are discussed. The above research is beneficial in guiding the design of mold fluxes to meet the solidification requirements of steel grades. Because this process wastes manpower and hinders the development of experimental process information, there is an urgent need to develop automatic feature extraction and mathematical modeling techniques for automatic feature sequence images.

Aiming at the difficulty and inaccuracy of the bearing preload test of the spindle system, the pressure sensor is used to monitor the force of the bearing cover to realize the real-time monitoring of the axial preload of the bearing outer ring of the spindle system. The mechanical finite element analysis model of the spindle system is established, which applies preload to the outer ring of the bearing point by point and monitors the force of the outer ring of the bearing point by point. The force of each bearing outer ring in the spindle system under different axial preload is analyzed by simulation and test. The results show that the experimental data of the bearing outer ring force changing with the axial preload are consistent with the simulation force prediction structure, which verifies the accuracy and reliability of the proposed real-time monitoring method for the bearing outer ring force.

The dual-vehicle system uses TI's 32-bit high performance, low-cost, low-power MSP432P401R as a control kernel to design a car-following system for a given multitasking situation. The vehicles have a controllable speed in this system, follow the track, and are capable of performing multiple tasks in conjunction with each other. The main body of the vehicle includes an encoder to obtain the current wheel speed, a greyscale detector, an OpenMV to obtain the track information, and an ultrasound distance measurement module to control the vehicle distance. In addition, the dual vehicles communicate with each other through Bluetooth modules to ensure that the vehicle tracking system can start and stop nearly simultaneously. During the driving process, both cars can smoothly follow the black track and perform the corresponding tasks of overtaking and parking. This system meets the requirements of design and the operation process shows excellent performance.

Target segmentation is an important aspect in autonomous driving systems. In this paper, we propose an improved algorithm based on the Euclidean clustering algorithm for the segmentation of LiDAR pedestrian point clouds, which often results in under-segmentation due to the close distance. After the point cloud pre-processing, the target segmentation is performed based on the point cloud density and spacing as the clustering basis. The experimental results show that the improved algorithm improves the target segmentation accuracy by about 4% and the pedestrian target segmentation accuracy by about 6.4% for the traditional Euclidean clustering algorithm.

In recent years, the devices of the Internet of Things have grown exponentially. As an emerging wireless access mode in the Internet of Things, NB-IoT is gradually replacing the traditional GSM/GPRS communication mode. Aiming at the requirements of environment monitoring and device control application scenarios of smart home, a smart home control and management system based on NB-IoT is designed. The hardware node used STM32F429 as the main controller, equipped with 4.3-inch touch screen, connected to NB-IoT network with M5311 module, and used BME280 and BH1750 for indoor data and information acquisition. An AT framework is designed based on FreeRTOS, a touch screen program is designed based on the MVP architecture of TouchGFX, Ali Cloud is used as MQTT server, and the Web interface is built using the NodeRED platform. The test results show that the system can realize the device login to the cloud server, complete the home environment real-time monitoring and intelligent device control functions.

As a complex large-scale automatic spraying equipment, the spraying operation object of the ship segment intelligent spraying robot system is large ship segment, which has the characteristics of large structure size and non-standardization, and these characteristics make its spraying process and quality control more complicated. In order to explore the automatic spraying process of ship segments and improve the spraying quality of the intelligent spraying system of ship segments, the spraying process of the intelligent spraying robot system of ship segments is studied by orthogonal experiments, and the spraying shift speed, spraying height and spraying pressure are selected as the main factors for analysis, and the results show that the best process combination of the intelligent spraying system of ship segments is the spraying height of 300mm, the spraying speed of 0.3m /s, and the spraying pressure is 0.6Mpa.

We build the Utility Management System software architecture evaluation model for reducing the risk of software system. we got t the Utility Management System software architecture needs, Combined with the Quality Attribute Workshop method and the PALM method, from the main function, external cross-linking relationship and development trend, so as to build five evaluation dimensions of software architecture, including modularity, technology maturity, competition processing, independence, reliability. Then grading each evaluation dimension to get the Utility Management System software architecture evaluation model. The model can realize the risk identification of early development of the Utility Management System software system through the comprehensive evaluation of five dimensions of software architecture, so as to help improve the quality of software system and reduce the cost of the whole life cycle of the system.

In the hot strip rolling process, anomalies can potentially have a considerable impact on the overall production process owing to the interdependent nature of the production module and variables. To address this concern, this paper proposes a data-driven framework that leverages a spatiotemporal feature extraction method based on Symbolic Dynamic Filtering to identify the causal relationships among the variables. These extracted features are then used to learn more representational features via a Restricted Boltzmann Machine, and the monitoring statistics are built on its concept of energy to form an anomaly detection method. The experimental findings demonstrate the efficacy and feasibility of the proposed framework in detecting faults, exhibiting notable accuracy and promptness in real-time scenarios.

Engineering problems typically involve complex structures and physical processes, which often make them difficult to solve directly. To address these challenges, surrogate models have become a common approach for approximating and optimizing engineering systems. Although high-fidelity (HF) models offer higher accuracy, collecting HF data points and performing calculations can be very challenging due to their high cost and computational complexity. In contrast, low-fidelity (LF) data points are easier to obtain, but their computational accuracy is relatively low. Therefore, developing multi-fidelity surrogate (MFS) models has become a necessary research direction. In this paper, we propose a multi-fidelity surrogate model by optimal model selection (MFS-OMS). The optimal model (OM) that reflects the difference between HF and LF is selected by combining cross-validation (CV) and mean squared error (MSE), and the proportion factor between high-fidelity and low-fidelity models is calculated using genetic algorithm. MFS-OMS is compared with Co_Kriging and LR_MFS using numerical examples and a press machine optimization problem. The experimental results show that MFS_OMS can provide more accurate prediction accuracy than Co_Kriging and LR_MFS models. The research in this paper can promote the development and improvement of models and algorithms, and provide more possibilities and feasibility for the application of surrogate models in a wider range of fields.

This paper proposes a visual inspection algorithm based on OpenCV to address the issue of bubbles in polarizers during attachment, which has the potential to reduce yield and reliability due to hardware, attachment parameters, and material characteristics. An industrial camera is used to capture real-time images of flexible OLED panels that already have attached polarizers. The captured images are subjected to grayscale conversion and median filtering to remove noise. Edge detection is then performed using the Canny operator and binary processing to mark the bubble edges. The bubble edge area is filled, and the area of the bubble is calculated, with larger bubbles filtered out. Comparative experiments demonstrate that this visual inspection algorithm based on OpenCV can achieve reliable detection results even under uneven illumination and severe reflection conditions. The proposed algorithm has significant value in addressing the issue of bubbles in polarizer attachment and can be applied in industrial production.

In order to reduce the failure of software system caused by demand defects and demand changes, the evolution process of requirements should be completed in a stable state. A nonlinear equation is established to describe this process. The parameters of the equation are mainly determined by the demand change rate, which can be divided into stable region and unstable region. If A is in a stable region, the demand can be completely evolved in a stable way, and the change of demand will not affect the final evolution result. In addition, if A is in an unstable region, the evolution process is unstable or chaotic, and the evolution result is abnormal. In this case, it is impossible to determine all subsystems. Because a small change in initial demand in evolution will cause great changes. We use the above theory to analyze the home telephone service management system and get a stable evolution process. The results show that the actual evolution trajectory is consistent with the theoretical trajectory. Therefore, we should adopt an appropriate demand change rate to get a stable demand evolution process and make the demand change.

To achieve precise seeding operations, an electric control precise seeding system was designed in this paper, in which a brushless DC motor was used to provide power for the seeding device. A double-closed-loop control model of the brushless DC motor was established and simulated, and a seeding device test bench experiment was conducted. The simulation results showed that the system had fast response, no overshoot, and good dynamic and static performance. The seeding device test bench experiment showed that the qualified seeding index was greater than 96.64%, the missing seeding index was less than 1.34%, the redundant seeding index was less than 2.01%, and the variation coefficient was less than 17.61%. The seeding uniformity was good, and the difference in seeding quality index was within 5%, which met the requirements of the JB/T 10293-2013 "Technical Conditions of Single-Grain (Precision) Seeder". The seeding stability was also good.

The versatility and agility of the assembly system is the key to affect the cost and efficiency. This paper focuses on the application scene of wing-body attitude adjustment, and puts forward a modular reconfigurable layout method and a spatial motion data calculation method based on measured position for tooling equipment; for the control system, a general communication method library based on ModbusTCP protocol is constructed; for process monitoring, a decisionmaking and monitoring model driven by all elements of PPR is established. The data show that the method can reduce the infrastructure construction cost and improve the assembly efficiency.

At present, issues such as high fuel cost, heavy environmental contamination, and high energy loss are found in the hydraulic system which uses fuel engines to drive the hydraulic pump in the stacking and picking forklift trucks, especially in high-level stacking and picking forklift trucks. In order to cope with the above problems, this paper achieves the goal of energy-conservative design by implementing an energy-efficient design for the hydraulic system of forklift trucks, i.e., adopting a servo drive system in place of a fuel engine to drive the hydraulic pump, and designing a reasonable accumulator to recover energy when the forklift truck unloads the goods.

The multi-joint mechanical arm has become the necessary production equipment. This paper aims to research space trajectory planning of multi-joint bionic scorpion manipulator to realize the polarizer attachment motion. By modeling of the structural connection relationships of bionic scorpion manipulator, the quintic polynomial method can be used for modeling of joint space trajectory planning algorithm. Working space of bionic scorpion manipulator for attachment motions can be determined by means of simulations. The underactuated and fully actuated motions are compared to analyze the planned space trajectory by motion simulations of different driving modes. This method can provide the high practical value and application prospects for the space trajectory planning during the dynamic simulation.

In order to solve the problem of speed matching between image data output and acquisition, it is convenient to provide simple and flexible transmission for high-speed digital cameras and image acquisition cards. This paper proposes a video processing module based on CameraLink interface technology. By making full use of the internal resources of FPGA, the algorithm is executed efficiently. The system not only supports the clock 85MHz, but also the acquisition cache is not less than 64MB. Real-time acquisition and processing of CameraLink with image characteristics of 200Hz/50Hz(16bit/8bit); Moreover, it supports two channels of 50Hz(8bit/8bit×3) CameraLink image output, and realizes the video conversion from CameraLink signal to SDI signal. The functional correctness and performance stability of the video processing module based on CameraLink proposed in this paper are verified through function and pressure tests.

For the sake of improving the dynamic response, and the anti-disturbance ability of IPMSM speed regulation, a novel nonsingular terminal sliding mode control (NNSTSMC)is designed in this paper. Firstly, to accelerate the convergence time of IPMSM speed tracking error, the manifold of novel nonsingular terminal sliding mode control (NNSTSMC) is constructed based on tanh function. Secondly, using equivalent control to design the novel nonsingular terminal sliding mode control (NNSTSMC), and introducing a softsign function in the switch robust control to attenuate system chattering. Finally, compared with PI, traditional sliding mode control (SMC), simulation results indicate that the nonsingular terminal sliding mode controller exhibits strong dynamic response performance and robustness.

On account of the special properties, throwing robot can be used as a military robot, which is suitable for detection, anti-terrorism, harsh environment detection and other scenarios, and has a good application prospect. In this paper, a buffering pleated mechanism based on dumbbell type throwing robot are proposed. The buffering energy absorption characteristics of pleated structure are compared and simulated by FEM based on Abaqus software. Designed for water and land patrol, the throwing robot contains two symmetrical trailing wheels coupled with the pleated structure to meet the requirement of balance. The results show that the buffer mechanism is compact, economical and possesses a high space utilization rate, reflecting its practical value.

In order to improve the cruising speed and range of the biomimetic fish underwater robot, reduce energy consumption, a microstructure is added to the lower surface near the robot head to study the influence of microstructure on the resistance characteristics of the robot. Firstly, transverse and longitudinal triangular grooves were added to the lower surface of the robot to analyze the resistance characteristics. Compared with smooth surfaces, the total drag reduction rate of longitudinal microstructures is 12.4%, and that of transverse grooves is 11.2%. It was found that the drag reduction effect of longitudinal grooves is better than that of transverse grooves. Then, numerical simulations were conducted on three types of microstructures: V-shaped, triangular, and trapezoid shaped. When the dimensionless size of the microstructure are h = 13 and s = 17, the triangular microstructure has the best drag reduction effect, with a total drag reduction rate of 12.4%. The V-shaped drag reduction rate is 11.4, and the trapezoidal drag reduction rate is 6.3%. Finally, the drag reduction mechanism of the microstructure is given by analyzing the velocity distribution and wall shear stress of the flow field around the microstructure.

We present a tension optimization method for the control force allocation problem of barges. Firstly, the mathematical model of tension distribution is established with the rate of change of anchor cable tension as the optimization objective and the control force and moment and the size of anchor cable tension as the constraints, and the tension distribution problem is transformed into a convex quadratic programming problem. Then, according to the idea of mathematical programming, the original pairwise interior point method is used to solve the convex quadratic programming problem. Finally, the effectiveness of the primal-dual interior point method is verified through simulation experiments; it has a high convergence speed and accuracy.

The construction industry, a major contributor to domestic carbon emissions, has great potential for energy preservation and emission decrease, as the country has put forward a goal of "carbon peak and carbon neutrality." Therefore, it is necessary to establish a structure carbon emission monitoring system based on CPS technology to achieve small carbon administration and provide scientific administration strategies. A genetic algorithm is used to create a multi-objective optimization arithmetical model of "duration-cost-carbon emission-construction quality" and the optimal construction administration strategy is then attained. This system will be composed of a CPS sensing covering, a network covering, and a control covering to gather, transmit, analyze, and give feedback on the carbon emission data of the construction area, thus permit for true-time monitoring of carbon emissions. The goal of verdant structure scheme administration can be effectively accomplished through visualizing carbon emissions. The CPS technology-based carbon emission monitoring system will enable true-time monitoring and early warning of carbon emissions during construction, there fulfilling the national demand for "carbon peak and carbon neutrality".

Camouflaged Object Detection (COD) is a challenging vision task that aims at locating objects having high similarity in appearance with its background. Existing COD methods are highly dependent on the high-dimensional feature of camouflaged object while neglecting details within the low-dimensional feature. In this paper, we highlight lowdimensional information to enlarge the discrepancies between objects and their surroundings. We propose a CNN-based network ShallowNet with enhanced low-dimensional feature to retain more differences in the process of encoding. The proposed approach shows enhanced effectiveness both qualitatively and quantitatively compared with existing COD methods regarding various datasets.

Advancements in artificial intelligence technology are leading to significant transformations in the generation of news. This article presents the News Video Style Feature Automatic Extraction System (NVFAE), which leverages state-of-the- art open-source technologies such as Automatic Speech Recognition, Optical Character Recognition, and Natural Language Processing. The NVFAE extracts news video style features from multiple modalities: textual features, visual features, and delivery features. The system uses automated annotation to extract five distinct style features from news videos: high-frequency words, overall tense, video color tone, main character, and video length. 38 BBC videos about the topic of "Turkey earthquake" from the YouTube platform are chosen for testing NVFAE, whose five features can be all successfully extracted. Experimental results show that the system proposed in this paper can provide a digital and structured way to explore the narrative characteristics of news videos, only with a mouse click and style features of news videos can be obtained in batches. NVFAE is aimed to overcome the difficulties of manual annotation for video news, as well as supply news workers with valuable references to study the news styles and content planning.

The position adjustment of unmanned aerial vehicles (UAVs) in formation flight is of great significance in various fields such as military defense and maritime surveillance. Although the existing active positioning technology for UAVs has been developed relatively well, it is highly susceptible to exposing position information. To solve this problem, the paper presents a position adjustment model for bearing-only passive positioning UAVs based on the greedy criterion, which adjusts the position by the received angle information and the geometric properties of the formation. For circular formations that can perceive height information: Firstly, determine the center and radius of the circle. Secondly, UAVs are arranged in a circle line through equal circumferential angles of the same chord and received angular information. Finally, distribute them evenly. For conical formations that cannot perceive height information: Firstly, convert three dimensions into two dimensions. Secondly, adjust the position of UAVs on the peripheral equilateral triangle. Finally, repeat the previous steps to adjust the position of the tapered interior UAVs. It needs 26 pieces of angle information after at most 15 adjustments to complete the circular formation; the conical formation can be adjusted up to 39 times, requiring 63 pieces of angle information. The fewer the required adjustments, the higher the accuracy.