This article reviews the parameters that influence free abrasive processing procedures and the models used to predict material removal. The text begins by introducing the common methods and principles of free abrasive processing. It then analyzes the impact of machining parameters on machining efficiency and surface quality. Finally， it summarizes the modeling method for material removal and provides prospects for future research directions.

This article initially provides an overview of the processes and standard regulations involved in the safety verification and validation of autonomous vehicles. Building upon the human-vehicle-road system theory， the article further introduces a novel classification approach， categorizing and summarizing the current technologies and assessment standards for safety verification and validation in autonomous vehicles. It also consolidates and comparatively analyzes three major categories of methods： those based on real-world scenarios， virtual scenarios， and a combination of both. The article conducts a comparative evaluation of the limitations， advantages， and disadvantages of 16 different verification and validation methods across eight characteristic dimensions. Finally， it briefly extrapolates on the challenges and future prospects in the research of safety verification and validation schemes for autonomous vehicles.

The target recognition algorithms and the major problems solved in image classification， detection， and segmentation tasks for side-scan sonar， synthetic aperture sonar， and forward-looking sonar were described. By combining the imaging characteristics and application scenarios of different sonars， the strengths and weaknesses of the target recognition algorithms under the corresponding image processing tasks of the above imaging sonar and the key issues that still need to be addressed were analyzed and summarized， and its future development direction was also looked forward to.

In order to improve the poor adaptability of the traditional equivalent consumption minimization strategy （ECMS）， and to further enhance the fuel economy of hybrid energy storage systems， an adaptive energy management strategy considering the prediction of the state of charge （SoC） of Li-ion battery is proposed. Firstly， based on the domestic tram lines and traveling data， a Markov chain is used to construct the typical driving conditions of streetcars under semi-independent right-of-way. Secondly， the SoC of lithium battery is predicted by adaptive Kalman filtering method， the charging and discharging process of lithium battery is optimized， the reliability of lithium battery is enhanced， and the minimum equivalent energy consumption of hybrid energy storage system is taken as the optimization target， meanwhile， the equivalent factor of traditional ECMS is optimized by combining with particle swarm algorithm， so as to realize the reasonable and effective distribution of load power between fuel cells and lithium batteries. Finally， a comparative analysis is carried out in the typical working conditions of the constructed tram under semi-independent right-of-way. The results show that， compared with the fixed-threshold strategy， the proposed strategy reduces hydrogen consumption by 0.63 kg and fuel cell peak current by 57.2 A. Compared with the state machine strategy， the proposed strategy reduces hydrogen consumption by 1.21 kg and fuel cell peak current by 24.6 A， and the fluctuation ranges of bus voltage and Li-ion battery SoC are both improved.

In order to study the influence factors affecting the preload loss of the cable clamp bolt of the suspension bridge and its sensitivity， a local finite element model of the cable clamp in a suspension bridge in Guangxi province was established， and then was verified by the creep test results. Next， 27 finite element models of bolt clamp were set up by using the orthogonal experiment method to study the influence factors on preload loss which include the stress relaxation rate of the bolt， the creep rate of the cable， the porosity of the cable， the suspender force， the initial preload of the bolt， the cable clamp inclination angle and the temperature difference between cable and clamp. The research results show that： the preload loss of the bolts increases with the stress relaxation rate of the bolts. The cable creep rate has a significant impact on the preload loss of the bolts. The preload loss increases significantly with the cable creep rate. The temperature difference between the clamp and the cable has a significant impact on the preload loss of the bolt， the bolt preload loss decreases as the temperature difference increases. The influence of the cable porosity on the preload loss of the bolts is relatively small. The impact of the suspender force on the preload loss of the bolt is negligible. The preload loss of the bolts initially increase and then decrease with the increasing initial preload of the bolts. The influence of the cable clamp inclination angle on the different bolts in the same clamp is unclear. The research conclusions can provide references for the design and maintenance of the suspension bridge cable.

In this paper， a two-dimensional lattice Boltzmann model based on enthalpy was established to study the temperature distribution， average liquid fraction and melting end time of the oxide layer of micronmeter-sized aluminum particles under different boundary conditions and particle sizes. The results show that the natural convection buoyancy has a significant effect on the melting characteristics. When a single heating wall is heated， due to the influence of natural convection buoyancy， the melting process of the lower wall is the fastest， and the melting speed of the upper wall is the slowest. When the two heating walls act together， the melting speed of the upper and lower heating walls is the fastest， and the melting time is shortened by 18.05% compared to the conditions of the left and right heating walls. As the number of heating walls increases， the improvement effect of melting efficiency weakens. Through calculation， it was found that compared with a single heating wall， the melting efficiency of double heating walls， three heating walls， and four heating walls increased by 21%， 73%， and 75%， respectively. The presence of a cold wall slows down the melting process， but the impact only exists in the latter half of the melting process. In addition， an increase in particle size can accelerate melting， but this characteristic will gradually weaken as the particle size increases.

In order to enhance the applicability of face gear in various compact devices and further expand its transmission advantages， the design of helical face gear with non-orthogonal offset modification integration and its tooth width is studied. The three-dimensional model of non-orthogonal offset modified helical gear is established. According to the restriction conditions of undercut and tip sharpening of non-orthogonal offset modified helical gear， the inner radius and outer radius of gear teeth are designed， and the effective tooth width of face gear is obtained. The influences of axis intersection angle， displacement coefficient， offset distance and helix angle on inner diameter， outer diameter and tooth width are analyzed. The results show that the effective tooth width of face gear first decreases and then increases when the shaft intersection angle increases. The effective tooth width of face gear increases with the increase of modification coefficient； The influence of offset distance on effective tooth width will produce different results because of the size of axis intersection angle； With the increase of helix angle， the effective tooth width will decrease with a small amplitude.

The online optimization method was proposed to solve the problem of hybrid integer nonlinear optimal control in hybrid electric vehicle gearing optimization. Firstly， the real-time control method was constructed based on the rolling optimization idea of model predictive control. The objective function was the total weighted cost associated with vehicle dynamic performance， minimum equivalent fuel consumption， and drivability. Secondly， based on the model fitting of the vehicle power system， the analytical solution of energy distribution was obtained by using the minimum principle， and the gear was optimized by the enumeration method at every moment. The simulation results under standard working conditions show that：①the proposed method can improve the calculation efficiency， and the calculation time is less than 50 ms， which has the potential of online application；②Compared with the dynamic programming method， the proposed method can achieve close to the global optimal fuel economy.

This article focuses on the micro invasive process of variable path wire controlled robots with micrometer precision， and studies the influence of temperature factors on puncture accuracy. By integrating thermal analysis and structural mechanics analysis， the non-uniform temperature field and structural deformation characteristics of key puncture needle components were obtained. By combining the orthogonal experimental analysis method， the influence of transient and steady-state temperature changes on the changes of puncture needle components was analyzed. The results show that the transient and steady-state temperatures from 23 ℃ to 40 ℃ have a significant impact on the precision of micrometer level puncture， and the accuracy impact indicators of the two temperature conditions are consistent： the maximum thermal deformation of the puncture needle with process parameters of 40 ℃-3.1 mm-3 mm/s is 0.012 mm， and the corresponding system impact reaches about 0.1 mm； In addition， based on the characteristics of clinical puncture technology， a 3-factor 5-level analysis of variance and regression analysis were conducted to reveal the non-uniform thermal deformation law caused by thermal factors of titanium nickel alloy puncture needles and the influence of temperature T on puncture accuracy， which is greater than the diameter D of the puncture needle and the injection speed V.

To measure the carbon emissions at each stage of buses throughout life cycle， this paper divided the life cycle of buses into process cycle and energy cycle. Considering the production， assembly， transportation， and recycling stages of each bus system， carried out research to obtain bus production data， combined the data provided by the company and GREET internal data， used Gabi to measure CO₂ emissions of process cycle. Constructed a CO₂ emissions measurement model that included the stages of energy extraction， production processing， transportation， and usage to estimate the energy cycle CO₂ emissions of electric buses and diesel buses. The result shows that the CO₂ emissions of electric buses are 39.2% higher than those of diesel buses during the process cycle. In the energy cycle， the CO₂ emissions of electric buses are 14.2% lower than those of diesel buses. In a comprehensive life-cycle comparison， the CO₂ emissions of electric buses are 9.73% lower than those of diesel buses.

This paper studies the freeway car-following safety in foggy weather. Then a control strategy for freeway car-following safety in foggy weather is proposed based on vehicle-to-vehicle（V2V） communications. Firstly， a foggy car-following model was selected to describe the car-following behavior in foggy weather. Numerical simulation was designed to analyze the influence of different foggy scenes and speed limit conditions on the risk of rear-end collision. Then we conducted sensitivity analyses on the collision time threshold TTC*， the initial speed v of the fleet and the distance L between the lead vehicle and the accident point when the lead vehicle just observed the accident point. Finally， considering the influence of speed difference between the vehicle and preceding vehicle on car-following behavior， a car-following safety control strategy was proposed based on foggy V2V conditions. The results show that the speed limit values of 60 km/h and 100 km/h will lead to the maximum risk of rear-end collision under light fog and heavy fog conditions， respectively. The light fog has the minimum risk of rear-end collision when 40 km/h and 80 km/h are selected as the speed limit value. The heavy fog has the minimum risk of rear-end collision when 60 km/h is selected as the speed limit value. The risk of rear-end collision is positively correlated with the initial speed v of the fleet and collision time threshold TTC*， and negatively correlated with the distance L between the lead vehicle and the accident point when the lead vehicle just observed the accident point. The proposed control strategy can effectively reduce the risk of rear-end collision and improve the car-following safety in foggy weather. Under the confidence level of 95%， the risk of rear-end collision was significantly reduced. The risk of rear-end collision could be reduced by 36.70%~45.14% under different foggy scenes and speed limit conditions.

In order to assess the impact of flooding on urban traffic operation and quantify the degree of failure of defective road sections in the city under the influence of flooding， this paper takes into account the decline in the capacity of road sections under the condition of flooding and the change process of traffic supply and demand in the road network， and quantitatively characterizes the degree of failure of road sections by taking road section smoothness and reliability as an indicator； we construct a Bayesian network model for assessing the degree of smoothness and reliability of the road network， and analyse the influence relationship between road sections and road networks； finally， identify the key road sections by combining the smoothness and reliability of road sections and their influence on road networks. The road network and flooding data of a certain area in Shenzhen are selected for example analysis， and the sensitivity analysis of the model verifies the feasibility of the model. The results of the study provide new ideas for the identification of critical road sections in urban road networks under flooding conditions.

During peak hours， the passenger flow of subway carriages increases sharply. In complex scenarios such as dense crowds and target occlusion， it is difficult to accurately identify each passenger， which can easily lead to missed or false detections. To this end， a subway carriage passenger flow detection method based on the improved YOLO algorithm is proposed. After analyzing the YOLOv8 model structure， the TFE module from ASF-YOLO was added to YOLOv8n. Combined with the spatiotemporal model， the characteristics of high passenger flow at stations and low passenger activity during driving， as well as the different flow characteristics of passengers in train doors and carriages， were considered. The multi frame detection results were fused to achieve accurate detection of passenger flow in subway carriages. Through experimental comparison， the average accuracy of the original YOLOv8n model is 57.0%， the improved model is 69.1%， and after multi frame fusion processing， it is 76.6%. The passenger flow information obtained based on this model supports multiple aspects such as passenger travel guidance， emergency rescue support， and railway operation control.

The vehicle window was introduced to the study to address the problem that the factors considered in the route optimization study of multi-vehicle customized bus are not close to the actual situation. Firstly， the concept of vehicle window was introduced to describe the departure cost， travelling cost， vehicle capacity and travelling speed of different type of customized bus. Secondly， an integer linear programming model with vehicle window was constructed. The model has passenger time-space window and vehicle window as input， and the minimization of enterprise operation cost and passenger travel cost as objective. Finally， a three-segment hybrid encoding genetic algorithm including the customized bus， boarding point and alighting point was designed to solve the model according to the characteristics. It solves the problems of chaotic service order and low optimization efficiency caused by hybrid coding in most stations.An example analysis was carried out in the Sioux Falls network. The results show that： The model with vehicle window is more suitable to the actual scheme， and the greater the difference of vehicle speed ratio， the greater the impact on the scheme.

Aiming at the problem that the traditional queue tail vehicle departure state prediction model is difficult to adapt to the uncertainty of queue dissipation， a queue tail vehicle departure state prediction model driven by trajectory data is proposed. By analyzing the shapes of queue dissipation trajectories and potential influencing factors， the uncertainty of departure state of tail vehicles is uncovered. Starting from the two stages of queue waiting and vehicle start-up， a feature set that influences the tail vehicle departure state is proposed. The extreme gradient boosting algorithm is employed to construct the prediction model， incorporating the SHapley Additive exPlanations（SHAP） interpretable machine learning framework to dissect the contributions of features， and to determine the optimal feature combination and model parameters. The research results indicate that the proposed XGBoost-based departure time prediction model achieves an average mean absolute percentage error（MAPE） of 5.74%， which is improved by 10% approximately compared with the kinematic model. The MAPE for the queue departure speed is 9.98%， improved about 6% over the kinematic model. Furthermore， the performance of the proposed model surpasses three commonly used machine learning methods of random forest， decision trees， and multi-layer perceptron neural networks. The research outcomes provide technical support for adjusting the minimum green light time of intersection signals and eco-driving of connected vehicles in the vehicle-road cooperative environment.

In order to alleviate the problem of imbalance between parking supply and demand by improving the utilization rate of parking resources， and considering the game relationship between user benefits and system benefits in parking allocation， the optimization objective function of maximum parking space utilization and minimum user cost in parking reservation mode is established from the two aspects which are system optimization and user optimization respectively. Then， the optimal parking allocation integer programming model（OPA） considering users' preferences is established. An Augmented Lagrangian-Alternating Direction Method of Multipliers Algorithm is designed to solve the optimal solution of the model. Finally， in order to test the validity of the model， the proposed model and the classical distribution model are compared and analyzed under different supply and demand conditions. The results show that the performance of the proposed model in three performance metrics which are parking utilization， average user cost and request acceptance rate is significantly better than that of the classical allocation model. The research results of this paper can provide theoretical reference for the management of parking reservation platform.

To improve the prediction accuracy and reduce the robustness of the traffic accident prediction model， this paper uses the Stacking integration strategy to construct an integrated traffic accident prediction model. Firstly， single traffic accident prediction models based on eight machine learning models， such as Decision Tree and Extra Tree， were constructed and the MIC test was used to measure the similarity of each traffic prediction model with the graph coloring method， and the models with low similarity and high diversity were selected to participate in the integration. Secondly， Box-Cox transformations were applied to the results of the single accident prediction models and different weights were assigned to each single model separately using feature weighting method. Finally， models such as BP neural network and Logistic regression were selected as meta-learners for Stacking integration. The results of the study show that the prediction accuracy of the integrated model with BP neural network selected for the meta-learner is higher than other integrated models， and the MAE and RMSE of the integrated model have been respectively reduced by 24% and 14% and the R² has been improved by 6% compared to the single accident prediction model with the highest prediction accuracy.

Aiming at the issues of severe missed detections and low detection accuracy for small targets in the perspective of drone aerial photography， an improved YOLOX network is proposed for the detection of drone aerial images. To enhance the feature learning ability of the network， the ASFF module is introduced in the feature fusion part， and the CA mechanism is embedded in the neck of the network. To enhance the network's learning of positive samples， the binary cross-entropy loss function is replaced with the varifocal loss function. Experimental results show that the improved YOLOX network has better detection efficiency， and its mAP_@50 reaches 91.50% and mAP_{@50_95} reached 79.65%. The visualization results in various traffic scenarios show that compared with other algorithms， the optimized network has a lower missed detection rate and higher detection accuracy， which can be competent for the detection task of small target vehicles， and can provide a reference for vehicle multi-target tracking applications from a high-altitude perspective.

This paper presents a bi-level model to address congestion pricing in a multi-modal transportation system. The upper level is an optimization model aiming to minimize the total social cost and determine the optimal congestion price for road segments. Meanwhile， the lower level is a doubly dynamical model including day-to-day traffic dynamics and within-day traffic dynamics. Solving the bi-level model employs a genetic algorithm. Three congestion pricing schemes are proposed and compared， namely no congestion pricing（NCP）， congestion pricing for cars（CPC） and congestion pricing for both cars and car-sharing（CPCS）. The results show that compared to the NCP， the CPC and CPCS result in a 17.44% and 14.89% reduction in private car trips， respectively. Additionally， the average travel times for private cars decrease by 7.54% and 30.18% under the CPC and CPCS， respectively. In the CPC scheme， car-sharing could generate maximum revenue and be the mode accounting for the largest share（39.97%） of the multi-modal transportation system.

A research on urban passenger transportation planning algorithm based on location potential energy and multi-source data was proposed to address the problems of traffic congestion and resource waste caused by unreasonable layout and inappropriate scale of passenger transportation in some cities. Firstly， an appropriate method was selected to integrate urban passenger transportation multi-source data from the fusion of the original data level， feature data level， and transportation theory level. Then， based on the probability model of traffic distribution and the Furness model， a traffic distribution prediction model was constructed， introducing location potential energy to predict the distribution of passenger transportation trips between traffic analysis communities. Finally， a dual level planning model for urban passenger transportation was constructed， and an improved quantum particle swarm optimization algorithm was used to solve it， in order to achieve urban passenger transportation planning. The experimental results show that the proposed method has higher scores than the comparison method in various evaluation indicators， with the highest score reaching 50.492. The highest congestion rate of vehicles is only 29%， which is practical.

To promptly detect， evaluate， and address potential traffic risks in highway tunnels， ensuring the safe and efficient operation of tunnels， a dynamic estimation method for tunnel operational risk states was proposed based on spatial-temporal Transformer network. Tunnel traffic flow holographic detection and key cross-section aggregation information as inputs was utilized， the spatial convolution and temporal LSTM was employed by proposed model for unsupervised extraction of spatiotemporal distribution features of different tunnel traffic operational states. Through extensive sample training of Transformer network layer parameters，it aims to capture the distribution and variances of tunnel traffic states in a high-dimensional risk feature space. This facilitates the estimation of operational risk of tunnel traffic flow. The effectiveness of the proposed method is verified by using real tunnel traffic detection data， and the accuracy of tunnel operation risk estimation is about 96%.

In order to determine the rheological response of ultrasonically disposed petroleum asphalt， a series of rheological tests were used to analyze the changes in the indexes of three typical petroleum asphalt before and after undergoing ultrasonic disposal. It was found that ultrasound reduces the real-time viscosity of asphalt at high temperature by more than 50%， and has an irrecoverable residual effect； the disposal asphalt becomes slightly softer and the thixotropic limit is increased at mid-temperature； and the creep capacity of the disposal asphalt is reduced at low temperature. Chemical and microanalysis showed that the content of heavy components in the asphalt was significantly reduced， and asphaltene aggregates were homogenized and dispersed. Selective inhomogeneous pressurized heating during ultrasound action induced a cracking reaction in the asphalt， which was responsible for the changes in chemical and rheological properties. This significant physicochemical effect of power ultrasound on petroleum asphalt has the potential to be used in road engineering techniques such as warm mixing， recycling， and modified processing.

To investigate the application effect of prestressed steel plate strip reinforcement technology in improving the load-bearing capacity of walls and reducing the degree of deformation， the deformation characteristics of prestressed steel plate strip brick masonry composite walls under applied loads were investigated by considering the failure state of supports.In the experiment， 7 specimens were carefully designed and reinforced with prestressed steel plates， each with different reinforcement situations. In the production of specimens， after pouring the bottom beam， a brick masonry wall is made. After pouring the top beam， the surface is smoothed to form an overall structure with the bottom beam and brick masonry wall. Transport the cured specimens to the laboratory for reinforcement with prestressed steel plates. The cutting length of the longitudinal steel plate strip should be consistent with the height of the wall. If vertical preloading stress is required， the cutting length will correspondingly increase by 2 or 4 mm. The horizontal steel plate strip matches the actual width of the wall. Install testing instruments and strain gauges， start loading until the support function of the composite wall fails， and stop the test loading.The results demonstrate a positive correlation between the reinforcement performance of prestressed steel plate strips on walls and the load-bearing capacity of the walls. Before the wall support fails， the curve shows a decreasing trend， followed by the failure of the wall support. A prestressed steel plate strip with strong reinforcement performance can reduce the strain value of the wall itself， that is， reduce the degree of deformation of the wall itself. The application of prestressed steel plate strips can significantly enhance the overall stiffness and stability of walls.

To address the issue of traditional GAN networks underperforming in single-image rain removal due to imbalanced network capacity， this article introduces a progressive recursive generative adversarial algorithm for this task. This method employs a progressive recursive module generator and a multi-scale feature module discriminator， aiming to enhance the efficiency of the generator and bolster the discriminator's capability. The progressive recursive module， by merging multi-scale features and constructing a progressive recursive structure， not only reduces the burden of network parameters but also elevates the generator's efficiency. Concurrently， the multi-scale feature module aids the discriminator in extracting features at both local and global levels， thereby amplifying its discriminative power. Experimental results indicate that， compared to existing algorithms， our method achieves a peak signal-to-noise ratio （PSNR） and a structural similarity index measure （SSIM） were improved by 1.11% and 1.16% on the Rain100L dataset. On the Rain100H dataset， these metrics were improved by 3.28% and 1.01%， respectively. On real-world datasets， our algorithm excels in rain removal， successfully preserving the majority of detailed features. These experimental outcomes thoroughly verify the effectiveness and robustness of our proposed algorithm.

Unequal allocation of container resources in a cluster environment is currently a pressing issue. In response to container load prediction and resource allocation strategies， this paper proposes an empirical mode decomposition-long short-term memory （EMD-LSTM）-based algorithm for predicting container resource load groups in the preprocessing of astronomical data. An adaptive recommendation value generation algorithm based on load prediction information is introduced， which automatically allocates container computing resources according to the degree of load fluctuation. The accuracy of load prediction was verified using simulated data and real astronomical observation data. Experimental results demonstrate that the proposed algorithm outperforms the triple exponential smoothing method and a single LSTM network model in terms of prediction accuracy. In real-time preprocessing of astronomical data， compared to the default strategy， the recommended value generation algorithm proposed in this paper effectively improves the utilization efficiency of computing resources.

To address the issues of multi-scale object segmentation errors， poor correlation between multi-scale feature maps and feature maps at different stages in the DeepLabv3+ network， the following modules are proposed to incorporate，including a global context attention module， a cascade adaptive Scale awareness module， and an attention optimized fusion module. The global context attention module is embedded in the initial stage of the backbone network for feature extraction， allowing it to capture rich contextual information. The cascade adaptive scale awareness module models the dependencies between multi-scale features， enabling a stronger focus on the features relevant to the target. The attention optimized fusion module merges multiple layers of features through multiple pathways to enhance pixel continuity during decoding. The improved network is validated on the CityScapes dataset and PASCAL VOC2012 augmented dataset， and the experimental results demonstrate its ability to overcome the limitations of DeepLabv3+. Furthermore， the mean intersection over union reaches 76.2% and 78.7% respectively.

In response to the security issues in industrial networks， a new intrusion detection method is proposed. The specific innovations of the method are divided into two aspects. First， in the process of processing， in order to solve the problem of high dimensionality of the original data， a particle swarm optimize genetic algorithm （PSO-GA） hybrid algorithm with dynamically adjusted parameters was proposed for feature extraction. It successfully screened out a subset of features that are meaningful to model training and accelerated training speed. Secondly， when building a machine learning model， theStacking integrated learning framework is used to generalize the output results of multiple models to improve the overall prediction accuracy. The experimental results on both two datasets show that the detection precision on the publicly available intrusion detection dataset CICDS-2017 has reached 95%， and it also has a 93% precision on a real industrial dataset developed by Lan Turnipseed from the gas pipeline control system.

Addressing the issue that the large amounts of data generated by devices during communication transmission are prone to becoming targets for hackers and malicious users， thereby generating abnormal traffic， and that the sparsity of traffic data makes it difficult to capture the associations between global features， which in turn affects the detection effectiveness of abnormal traffic， a spatiotemporal fusion detection method for abnormal traffic in industrial Internet of Things （IoT） based on the improved bidirectional long short-term memory （BiLSTM） neural network algorithm using mean squared error （MSE） is proposed. Firstly，the industrial Internet traffic data is converted into numerical data through the One-Hot coding method， and the SE mechanism in MSE is used to adjust the weight of traffic characteristics to capture the correlation between global characteristics.Secondly，using the forward and backward LSTM of BiLSTM neural network， the spatiotemporal fusion features of network traffic are extracted.Lastly， and the spatio temporal fusion features are input into the softmax classifier to identify traffic and achieve anomaly detection. The experimental results show that when the number of iterations reaches 30， the loss value of the proposed method can reach below 0.4， when the number of iterations reaches 60， both F1 and Matthews correlation coefficients can reach 60， proving that this method has good overall performance.

Distributed fiber optic acoustic sensing signals are affected by environmental noise interference， fiber optic transmission limitations， and multitasking processing requirements in practical applications. To solve this problem， an adaptive task scheduling method based on full waveform inversion is proposed， aiming to improve the signal-to-noise ratio， optimize multi task scheduling processing， and thus enhance the overall performance of the sensing system. Using a full waveform inversion method based on an improved adjoint state source equation， direct full waveform inversion is performed on distributed fiber optic acoustic sensing signals. Using a constructed neural network to learn variable relationships. Based on the variable relationship， provide the remaining number of solutions for each subtask of the selected main task in the current migration stage， update the probability of selecting the migration task， adaptively and dynamically select the migration task， and achieve multi task scheduling. The performance of multi task scheduling was tested on the benchmark test function， and the development trend of fitness and dispersion evaluation results was relatively positive， with fitness exceeding 8 and dispersion below 0.3. It can be seen that this method can effectively handle the problem of low signal-to-noise ratio of sensing signals， improve the quality of multi task scheduling in distributed fiber optic acoustic sensing systems， and have a strong assisting effect on multi task scheduling in systems with low signal-to-noise ratio signals.

In complex outdoor environments， during the landing process of drones， the target may be temporarily obstructed or out of view， leading to tracking failure. To enhance the accuracy and stability of unmanned aerial vehicle attitude control， a quadcopter unmanned aerial vehicle autonomous landing attitude control method based on tracking-learning-detection（TLD） algorithm is proposed. Combining extended Coleman filtering and TLD algorithm to detect specific targets and achieve target tracking through multiple median streams. By accurately capturing target position information， combined with additional inertia term crowd search algorithm and active disturbance rejection control technology， the selection of search step size and directional inertia coefficient was modified to optimize the flight attitude of quadcopter drones， improving the stability and safety of the landing process. The experimental results show that the average center offset of the proposed method is within 1.98 pixels， and the roll angle， pitch angle， and yaw angle the deviation is all within 0.02°， meet the expectations，the operation is smooth， the performance is better， ensuring the safe landing of the quadcopter drone.

In order to solve the problems of high computational complexity and low recognition and classification accuracy in the recognition of various types of laser welding seams， this paper proposes a laser welding seam image recognition and classification algorithm based on the improved Northern Goshawk algorithm（UNGO）， which combines the traditional support vector machine algorithm（SVM） with the improved Northern Goshawk optimization algorithm（UNGO-SVM）， and increases the algorithm search ability through chaos optimization and Levi's greedy learning strategy in flight. At the same time， it helps the algorithm overcome the situation of falling into local optimum， and improves the convergence accuracy and image classification accuracy of the algorithm. The experimental results show that this algorithm （UNGO-SVM） improves the classification accuracy to 99.15% while ensuring the convergence of the algorithm. Finally， compared with SVM， NGO-SVM，DOA-SVM，GOA-SVM improves by 21%，5% ，10% and 11% respectively， proving the feasibility and strong utilization value of this method.

When a traditional single-pixel compressive imaging system obtains measurement values， if the relative placement of the foreground target and imaging system is not static， the reconstructed image is blurred or completely distorted. To solve this problem， a dynamic complementary compressive imaging method based on a complementary mode is proposed. In this method， a single-column digital micro-mirror device is used to modulate the foreground image， two independent single-pixel sensors are used to obtain two optical signals reflected by the digital micro-mirror device， and the compressive measurement values of the foreground target image are obtained column by column， using the recovery mode of the dynamic compressive imaging is obtained；then based on this recovery mode， the traditional algorithm reconstructs the target image. In contrast to the results of traditional reconstruction， the results of each optical channel can be used to reconstruct two target images simultaneously. To improve the quality of the reconstruct image， this paper presents a quality enhancement method based on multi-channel image fusion. The results of simulation experiments show that the proposed dynamic complementary compressive imaging system not only can effectively reconstruct the foreground image， but the quality of the output image is not affected when the moving speed of the system changes within a certain range， demonstrating the good robustness of the system.

Aiming at the problem that the inverse kinematics performance of the mobile loading manipulator needs to be improved， an inverse kinematics solution method based on the multi-population gray wolf algorithm is proposed. Firstly， the inverse kinematics problem of the mobile loading manipulator is transformed into an equivalent optimal problem， and the fitness function is established according to the optimization objective. Secondly， based on the gray wolf algorithm， the grey wolf population is expanded， the particle swarm algorithm and the position updating method of the optimal individual inverse guidance are introduced， and the random reorganization threshold elimination mechanism is set. Then， the algorithm is applied to iteratively invert so that the fitness function approaches 0 to obtain the inverse solution. Finally， the simulation comparison with other solving methods shows that the proposed method has better convergence， solution accuracy and repeatability.

Addressing the challenges inherent in passive unmanned search and rescue missions at sea， including difficulties in target identification， broad search areas， and slow route planning， a strategic process was introduced for maritime search and rescue area planning and a route planning model specifically designed for passive unmanned missions. By thoroughly understanding the emergency response operations at sea and the specific needs for route planning， an optimal routing model have been developed considering factors such as the efficiency of search and rescue area coverage and the cost of rescue routes. The objective function is constructed within these constraints and solved using the whale optimization algorithm. The validity of the model is confirmed through designated scenario experiments， indicating that our proposed model for maritime passive unmanned search and rescue route planning is capable of swiftly identifying the search and rescue area and efficiently discovering a route with reduced costs.

To enhance the real-time performance and accuracy of star point centroid extraction， an improved run-length encoding algorithm based on field programmable gate array（FPGA） was proposed for fast star point extraction. The algorithm combines the characteristics of star point targets with the parallel processing structure of field programmable gate array， addressing the shortcomings of the traditional run-length encoding algorithm that requires setting equivalence tables for label merging and multiple rounds of polling for run-length encoding during extraction. It only requires scanning the image once， and after a delay of several clock cycles， the centroid coordinate of the star point can be extracted. Finally， the algorithm is tested and validated on an field programmable gate array of a star sensor. With a clock frequency of 50 MHz and an input image resolution of 1 280×1 024 pixels， the algorithm takes about 2 us to extract the centroid of the star point， and the position is completely accurate. Compared to the nearly 19 ms consumed by centroid extraction using an Advanced RISC Machine（ARM）， the improved algorithm demonstrates significant advantages and holds notable engineering application value.

Inspired by the microscopic Bouligand structure of mantis shrimp， this paper used four cells （hexagonal， circular， quadrilateral and negative Poisson's ratio） as interlayer basic structures， and established four bionic helical structures and four contrast structures. Using the thermoplastic polyurethane material as the base material， the above structures were fabricated by 3D printing technology. Based on the quasi-static compression tests and numerical simulations， the energy absorption performance of the above eight structures was investigated. The results showed that the energy absorption characteristics of the four bionic helical structures were better， and the specific energy absorption of the bionic hexagon helical structure could reach 531.65 mJ/g. Compared with the corresponding contrast structure， the specific energy absorption of the four bionic helical structures increased by 64.72%， 32.39%， 55.84% and 25.14% respectively， which indicated that the helical stacking distribution between layers effectively increased the energy absorption characteristics of the structures. The deformation and failure of the structures would absorb more energy.