This PDF file contains the front matter associated with SPIE Proceedings Volume 13224, including the Title Page, Copyright information, Table of Contents, and Conference Committee information.

With the rapid urbanization in China, the construction of high-rise residential buildings has become increasingly vital. However, the increasing scale and depth of substructures in tall buildings have posed challenges in the excavation of deep foundation pits. The excavation of such deep foundation pits demands comprehensive monitoring and management to ensure structural stability and safety. This study aims to address the issue of foundation pit deformation prediction using machine learning methods. Traditional mechanics models and numerical analysis methods have demonstrated substantial errors in this context. Therefore, this paper reviews the applications of Gaussian processes, genetic algorithm-optimized BP neural networks, and Long Short-Term Memory (LSTM) networks in foundation pit deformation prediction. However, these methods exhibit limitations, such as the stationary assumption in Gaussian processes, high computational costs associated with genetic algorithms, and LSTM's limited performance with small datasets. To mitigate the challenges posed by limited monitoring data, this study introduces a transfer learning strategy based on the CNN-LSTM-Attention model. Validation results from practical case studies suggest that this strategy effectively enhances the accuracy of foundation pit deformation prediction, especially when dealing with limited data.

This paper constructs a comprehensive value evaluation model of Energy Internet based on an integrated approach AHPentropy method and cloud model theory. Annual data from 2010 to 2019 is selected to assess the value creation of digital technology-orientated energy internet in China. By separating it with economic, energy, environment and social dimensions based on the process of energy internet value creation, the comprehensive value of digital technology-orientated energy internet is comprehensively assessed in a macro perspective. Guided by the policy agenda, different scenarios of digital technology inputs are set to simulate the change of Energy Internet value creation. The results show that digital technology significantly enhances the value creation for Energy Internet in all aspects. Compared with the BAU scenario, the comprehensive value score increases by 5.18% under the High-input scenario and by 1.75% under the Low-input scenario, while the excessive investment in digital technology does little to improve the value creation from Energy Internet. The approach proposed is versatile and can be adapted to perform a comprehensive assessment of any other system.

This paper investigates the use of a Back-Propagation Neural Network (BPNN) to identify the correlation between relevant parameters and ship collisions. Subsequently, a model is constructed to forecast the number of collisions over the upcoming three years. This model is expected to provide insight into the prevalence of maritime incidents and aid in the formulation of preventative measures. In this paper, BP neural network is used to study the water incidents provided by TSB (The Transportation Safety Board of Canada). The purpose of this paper is to predict the number of collisions in the future from the macro level. To achieve this goal, this paper establishes a data model based on Python. The results of this study can help researchers and channel management agencies better understand collision risk.

Electricity theft detection is a major concern in modern power systems, and has drawn the attention of both academia and industry. However, this problem has not been fully solved because of its complex intrinsic patterns, which are beyond the capacity of human decisions. Machine learning methods have been proven to be effective in handling such complex problems. Based on an open electricity theft detection dataset, a deep learning-based solution, the Temporal Convolutional Network (TCN), was proposed in this study and compared with existing machine learning schemes to demonstrate its effectiveness. The influence of the model parameters was also investigated in numerical experiments. The proposed TCN model is effective in detecting electricity theft behaviour in real-world electricity markets.

According to the national fire situation statistics in 2022, electrical fires caused by residual current failures account for about 14% of the total. At present, the user-side load is increasingly complex, and the residual current signal is difficult to effectively identify, which seriously hinders the safe and healthy operation of smart cities, and it is urgent need for the residual current monitoring technology. Aiming at the hidden dangers of electricity consumption caused by residual current, this paper combined with the monitoring principle of residual current, proposed a cluster analysis of residual current hidden danger diagnosis technology based on particle size calculation, obtained the nonlinear relationship between characteristic data and other variables, output the residual current diagnosis results, and verified them in the laboratory and on-site conditions, realizing the effective monitoring of residual current in smart cities, which may improve the safety and reliability of city electricity.

Aiming at the demand for low-power monitoring in complex environments to ensure gas pipeline safety, this design focuses on a monitoring system for gas manhole covers. The system enables real-time monitoring of combustible gas concentration, temperature, and humidity emanating from the manholes. It uses a Narrowband IoT (NB-IoT) module to transmit data to an IoT platform via a base station, enabling data visualization on the application side. To counteract the effects of temperature and humidity changes on gas detection accuracy, a method using a Long Short-Term Memory (LSTM) network for temperature and humidity compensation is proposed. This network is trained to align measured values with actual values under varying conditions, leading to the construction of an effective error compensation model. Test results demonstrate the system's excellent low-power performance and measurement accuracy, and it has been successfully scaled up for wider applications.

At present, the physical quality and health status of college students are declining year by year. Carrying out physical health test in colleges and universities can not only fully understand the physical condition of students, but also provide useful reference for physical education in schools. Facing the huge student population, the physical fitness test mode and means in colleges and universities still rely on a large number of manual operations, which have the disadvantages of complicated processes, recording errors and low efficiency, thus affecting the actual effect of physical fitness test. In this regard, based on the actual needs of colleges and universities, this paper will integrate technologies such as Internet of Things, data communication and cloud computing, and put forward a set of construction scheme of college students' physical fitness test intelligent system, making full use of intelligent devices to improve the efficiency and accuracy of college students' physical fitness test, and promote the college physical fitness test model to be more intelligent and scientific. Practice has proved that the RFID reader arranged in the playground can quickly read the personal information in students' IC cards, complete the perception and judgment of students' test scores, and realize the automatic collection and processing of test data. At the same time, the RFID reader will also upload data in the cloud through the intelligent gateway, which realizes the remote monitoring and calling of test data and has certain promotion value.

Trajectory prediction plays a crucial role in achieving autonomous driving, as it significantly reduces driving risks by predicting the movement trajectory of other vehicles. The key challenge lies in effectively encoding scene information and generating accurate multimodal results for each agent. To address this challenge, we propose a graph neural network framework that enables multi-centric modeling of relationships between heterogeneous inputs. This framework establishes various spatiotemporal adjacency relationships among scene nodes, leveraging graph attention mechanisms to allow each scene node to learn neighborhood features effectively and generate scene context enriched with valuable information. To tackle the problem of declining prediction accuracy as the prediction time increases, we propose a variable length window structure. The structure consists of a long window prediction module for multi-agent multimodal prediction, followed by a short window optimization module for refining the predictions. By utilizing this structure, we successfully strike a balance between model size and prediction accuracy. To validate our proposed model, we conducted experiments on the Argverse 1 motion forecasting dataset, and the results showcased excellent predictive performance.

In this paper, we propose a low complexity method to eliminate random object motion in real time. This method is based on the radar system composed of AD8302 and phase shifter. RNN is used to predict the output voltage signal of AD8302, and the phase difference caused by motion is predicted and compensated in real time according to the relationship of phase voltage, so as to obtain vital signs. At the same time, we also modify the phase voltage curve of AD8302 to increase the accuracy of eliminating random body motion. The experiment shows that the method can eliminate the linear motion signal and retain the sinusoidal signal of the loudspeaker in the case of the combination of linear and sinusoidal motion, which greatly validates the feasibility of the method to eliminate human motion and extract vital signs.

To fulfill the requirements for unmanned aerial vehicle (UAV) equipment testing, usage, maintenance, and repair, and to achieve the goals of forward support, rapid support, and comprehensive support, a comprehensive calibration and testing system for UAVs has been designed and implemented. This system focuses on calibrating and testing multiple subsystems including the wireless data link, flight control system, onboard sensors, power This system focuses on calibrating and testing multiple subsystems including the wireless data link, flight control system, onboard sensors, power system, and overall wiring of the UAV. It incorporates various subsystem calibration and testing equipment to ensure accurate and reliable results. It incorporates various subsystem calibration and testing equipment to ensure accurate and reliable results. The system design is rational, utilizing scientific methods and employing advanced technology, resulting in stable and consistent performance. The system design is rational, utilizing scientific methods and employing advanced technology, resulting in stable and consistent performance. The experimental outcomes affirm the effectiveness of system in meeting the calibration and testing requirements of UAVs.

The multi-agent cooperative search has garnered increasing attention in recent decades. Within the decentralized framework of multi-UAV cooperation, precise task allocation and assignment are paramount to balancing the workload of each UAV. To achieve reasonable task allocation, we employ spatial segmentation within the search domain, dividing it into distinct partitions. The aim is to allocate equivalent search workloads to individual UAVs across these partitions. To enhance algorithm efficiency, we utilize the Voronoi Diagram as the spatial segmentation generator and integrate deep reinforcement learning to refine the topological structure of these partitions. Finally, to assess the robustness of our proposed algorithm, we conducted experiments under various search scenarios. The results demonstrate significant improvements in the overall search efficiency of the swarm.

Point cloud completion involves repairing incomplete and fragmented point clouds into complete ones to facilitate downstream tasks more effectively. In current learning-based methods, if the network fails to adequately consider the structural relationships between point cloud features and the prior knowledge embedded in incomplete point clouds during learning, it may lead to completed point clouds lacking in local details, thereby affecting the completion accuracy. To address this issue, this paper proposes an improved network, namely the Attention-based Structural Analysis Point Cloud Completion Network. The entire network continues the structure of the pf-net network, employing a cascaded approach for feature extraction to extract and merge high-dimensional features. It introduces a self-attention mechanism to analyze the structural relationships inherent in high-dimensional features, and a cross-attention mechanism to assist the decoder in utilizing the prior features from incomplete point clouds to recover complete point clouds. Experimental results on the Shapenet dataset show a 4.818% improvement in completion accuracy compared to the baseline network. Additionally, there are better visual results, and the network demonstrates advantages in CD compared to other networks.

For an especial multi-agent system (MAS) characterized by non-repetitive parameter perturbations, this brief proposes a composite iterative learning control (ILC) algorithm to achieve complete trajectory tracking of the system. Firstly, employing 2-D system theory, the ILC system model consisting of an uncertain multi-agent system and ILC law is transformed into an typical uncertain 2-D Roesser system. Secondly, a controller with a real-time state information feedback term was designed to this system, featuring characteristics of feedback on the time axis and feedforward on the iteration axis, thus constituting a composite ILC scheme with a control effect in two directions for the ILC system. Subsequently, utilizing Lyapunov stability principles and LMI methods, some sufficient conditions are obtained for robust stability of the system under the action of iterative learning control, and feasible solutions for the learning gain matrix are derived.

It is challenging for robots to complete tasks in dangerous environments. Traditional approaches require redesigning and rebuilding specialized robots for each specific task, leading to inefficiency. Inspired by the diverse animal morphologies shaped by evolution, the study introduces a novel two-layer evolutionary method for generating robot morphologies. In the external evolution stage, morphological structural changes are carried out by selecting the robots with high fitness in the population. In the internal evolution phase, each robot adjusts its morphological attributes in each episode before interacting with the environment. Our approach ensures that robots can quickly adapt to environmental changes through dynamic adjustment of morphological properties while maintaining morphological diversity through changes in morphological structure. The central architectural element of this strategy is Graph Neural Networks (GNNs) and the parameters of the strategy are updated by Proximal Policy Optimization (PPO). Experimental results across four different environments demonstrate the effectiveness of the proposed method. Robots evolved through this method exhibit higher performance in terms of morphological design compared to baseline methods.

Corona Virus Disease 2019(COVID-19) is a respiratory disease whose chest X-ray (CXR) or computer tomography (CT) images are different from those of ordinary people which providing a basis for in-depth learning method detection. The article proposes a Residual Network (ResNet) pneumonia detection model based on the ResNet50 network which integrates Hierarchical Split Block (HS-Block) and Convolutional Block Attention Module (CBAM) during the feature extraction process to enhance lesion features. The datasets are collected and integrated which containing 31883 CXR and CT images tor verify the effectiveness of the model. The experimental results show that the average accuracy of the model reaches 98.8% which indicate that incorporating HS-Block and CBAM modules into the residual network can improve the performance of pneumonia detection.

Significant progress has been made in the field of Human Activity Recognition (HAR) through the application of deep learning. However, most existing studies employ supervised learning methods which require expensive and time-consuming labeled data acquisition. Therefore, contrastive learning has been applied to HAR, where data augmentation methods can effectively alleviate the problem of scarce labeled data. Furthermore, existing indoor activity recognition research based on sensors often uses multiple Inertial Measurement Unit (IMU) sensors, resulting in an obtrusive, uncomfortable, and unimodal sensor modality. Shoes, as a fundamental aspect of modern human life, offers advantages such as portability, concealment, and comfort compared to IMU sensors. Utilizing insoles as sensors for activity recognition data collection can effectively address the mentioned issues in practical applications. The fusion of insole and single IMU sensor data not only enhances robust predictions but also provides a more comfortable solution. In summary, we propose a contrastive learning framework based on the multimodal fusion of shoe insoles and single wristbands. This involves separately applying contrastive learning to the data from both sensors. Then, it computes cross-modal contrastive loss for distinct feature vectors to improve model performance. Results show that the multimodal fusion of shoe insoles and wristbands can yield superior outcomes when employing appropriate methods compared to single modality. Multiple cross-modal contrastive loss computations facilitates a more comprehensive understanding of the similarities and differences between feature vectors. Furthermore, even with scarce labeled data, contrastive learning excels.

As a promising solution, Vehicle Edge Computing (VEC) addresses the challenge of driverless technology to process computationally intensive tasks in a latency-sensitive manner by offloading the vehicle's computationally intensive tasks to MEC servers or the cloud. However, in situations with limited server resources, reducing service latency and improving the efficiency of service request processing remains a challenging task. To tackle this problem, we propose a joint task offloading and service caching framework aimed at minimizing the cost of unmanned vehicles. Initially, we formulate the problem as a mixed-integer nonlinear programming problem, subsequently transforming it into a solvable Partially Observable Markov Decision Process (POMDP) problem. We then design and introduce a task offloading algorithm framework based on DDQNL to address this problem. The performance of the proposed algorithm is validated through comparisons with other baseline algorithms.

The detection performance of focused beam-forming when the focusing depth-of-field is mismatched is analyzed. Several cases are simulated for changing processing frequency, target azimuth and distance. By analysis, the FCBF is more sensitive to the mismatch of depth-of-field when dealing with a high-frequency, near and vertical target. In application, multi depth-of-field comparison should be carried out to obtain more accurate DOA estimation results.

As blockchain technology plays an increasingly critical role in the deep integration of digital economy and real economy, it shows great potential and value in the energy industry. However, at present, there is a lack of blockchain management function support in the energy field, and the operations related to building business chains must be carried out in the way of command line and script, while there is a lack of visualization and control means, which is not friendly enough to the users and has a high technical threshold. This paper is oriented to the energy field, and for the first time proposes, designs and realizes a blockchain management platform for the energy field, which has the following advantages: 1) Rapid deployment of blockchain. The platform makes the application development process and application deployment process simple and efficient by providing various blockchain resources and components.2) Visualized supervision. The platform provides real-time supervision of on-chain data and blockchain node monitoring and other functions.

The field of 3D printing has revolutionized various industries, providing new avenues for innovation and production. However, the process of managing transactions for 3D print models can be complex and time-consuming. To address this challenge, the design and implementation of a Transaction Management System (TMS) based on the B/S (Browser/Server) architecture offer a promising solution. The TMS, rooted in the B/S architecture, brings forth an efficient and scalable platform for managing 3D print model transactions. Emphasizing real-time collaboration, the system leverages the centralized server to streamline communication and data flow, ensuring optimal efficiency. Its adaptability to emerging technologies and adherence to security measures make it well-positioned to meet the dynamic demands of the 3D printing industry. This essay explores the benefits and significance of such a system, highlighting its potential to enhance efficiency and quality in the 3D printing industry.

The rapid rollout of 5G networks has heightened the necessity for accurate forecasting of electromagnetic radiation (EMR) intensity emanating from 5G base stations. In this paper, we introduce a cutting-edge Enhanced Graph Neural Network (GNN) model specifically designed for predicting EMR intensity. This model tackles the intricate spatial and network interactions inherent to 5G infrastructure. By harnessing the capabilities of Graph Neural Networks in analyzing multifaceted data, our approach marks a substantial advancement over traditional predictive frameworks. Comprehensive empirical tests showcase the model's remarkable precision in EMR intensity prediction when pitted against standard machine learning practices and a baseline neural network. This paper not only facilitates the safe rollout and governance of 5G networks but also underscores the utilization of sophisticated machine learning paradigms in the telecommunications realm. The revelations carry profound ramifications for public health and safety, regulatory adherence, and the subsequent evolution of wireless communication networks.

There is a large amount of redundant data in the on-site installation and dismantling of electric meters, which affects the accuracy of the information. Therefore, a real-time monitoring method based on Industrial Internet of Things (IoT) was designed to monitor the dynamic information of one click disassembly of power meters. Establish a dynamic monitoring information database for one click disassembly and assembly of electric energy meters, centrally manage, store, and analyze various information during the disassembly and assembly process, to meet the real-time monitoring needs of dynamic information for electric energy meter disassembly and assembly. Building a dynamic monitoring model for onsite information of electricity meter disassembly and assembly based on industrial Internet of Things, integrating monitoring data from various sensors, electricity meters, cameras and other devices, and combining industrial Internet of Things technology to form a complete and scalable intelligent monitoring architecture. Considering the data bits, stop bits, check bits, baud rate, start bits, and communication protocol, an asynchronous serial communication format for dynamic information of one key disassembly and assembly of power meters is defined to ensure the integrity and accuracy of monitoring data. The experimental results show that this method has high monitoring accuracy and can be applied in practical life.

In this study, we developed a drowning detection method named AquaYOLO, which enhances the YOLOv5 framework by integrating the detection head mechanism from YOLOX. This involves decoupling the classification and regression prediction heads and incorporating a Dual Attention Net module that combines Channel Attention with Position Attention. Additionally, we added a small target detection layer with 4x4 pixel resolution. To improve border positioning accuracy, we replaced the conventional IoU loss function with the SIoU loss function. Our experiments demonstrate that AquaYOLO achieves a detection accuracy of 92.24% and a recall rate of 82.612% on the dataset, which is 9.98% higher than the traditional YOLOv5s. When compared to YOLOv3, YOLOv4, and Faster CNN, AquaYOLO shows superior detection accuracy, indicating its effectiveness in drowning detection scenarios. This study offers a significant advancement in marine drowning rescue solutions.

Time series data anomaly detection is to identify observations or patterns in some chronologically ordered points which are significantly inconsistent with expected patterns or normal behavior. These patterns may indicate unexpected events, unusual behavior, failures, or other unusual conditions in the system. Time series anomaly detection has important applications in many fields, including industry, finance, medical care, etc. We designed a time series anomaly detection algorithm using LSTM(Long Short-Term Memory) and attention. By adding an attention layer after the LSTM network, the network can pay more attention to more relevant features in the input multivariate time series, thereby improving accuracy and recall. In the data preprocessing process, the decision tree algorithm is used on the original data set to remove features that have little impact on the anomaly detection results, so as to reduce the computational complexity of anomaly detection, and improve the efficiency of anomaly detection. Experiments on the industrial time series data set SWaT data set show that the LSTM-Attention model proposed in this article is better to the basic LSTM network in terms of precision, recall, F1score and other indicators, and achieves good time series data anomaly detection results.

In the vacuum thermal test of communication satellites, waveguide are often used to establish satellite uplink or downlink communication links. As communication satellites become more and more complex, the use of thermal test waveguide also increases. Before the waveguide is used, the electrical performance of the waveguide needs to be tested. The measurement parameters are mainly standing wave ratio and insertion loss. In order to meet the test work of different types and different frequency bands, and to improve the test efficiency. Based on the LabVIEW integrated development environment, this paper establishes the data communication between the microwave network analyzer and the PC through the LAN interface. Based on the test requirements, the entire test system is designed in layers and modularized, which effectively realizes automatic measurement. Function, improve the test efficiency.

As the power system gradually moves towards a new energy Internet, large-scale sensing measurements are deployed in the power system. This provides data support for data-driven false data injection attack detection methods. In order to solve the problem of low classification and recognition accuracy of False Data Injection Attacks (FDIAs), this paper proposes a power grid FDIAs detection method based on S Transform (ST) and Long Short-Term Memory (LSTM) network. This method uses discrete ST to perform time-frequency analysis on the power grid measurement signal. By extracting time-frequency features of power measurement data, false features are highlighted. Then, the LSTM network is combined to accurately classify FDIAs. Simulation experiments show that the attack detection accuracy of this method reaches more than 95%, and the false alarm rate is less than 5%.

The operating condition of bearings directly determines the operating condition of the system. It is necessary to monitor the signal data of bearings to maintain the stability and reliability of the equipment. Addressing the issues of incomplete feature extraction in traditional diagnostic methods, high deployment costs in actual industrial production, and reduced diagnostic efficiency due to noise interference, a method for fault diagnosis of bearings was proposed that combines optimized Variational Mode Decomposition (OVMD) with Pyramid Transformer. Firstly, the optimized Variational Mode Decomposition parameters are obtained by analyzing the original bearing vibration signals. Subsequently, the signals processed through variational mode decomposition are preprocessed to generate two-dimensional feature maps, which are then input into the Transformer module. Passing through multiple modules of different sizes, the features extracted by different modules are fused to obtain vibration signal features with different receptive fields. After classification calculation, the final result is obtained, and a visual analysis of the results is conducted. The proposed method combines early fault diagnosis methods with deep learning techniques. Experiments have shown that this method achieves structural lightweighting while demonstrating its effectiveness and noise resistance on common bearing datasets.

With the continuous increase in the volume of Internet of Things (IoT) data, the limited number of servers deployed at the edge poses a pressing challenge for efficiently processing large-scale data streams. This study aims to explore the application of FPGA hardware acceleration technology to enhance the performance of the Apache Flink framework to meet the growing demands of users. The paper advocates the utilization of FPGA as a hardware accelerator for enhancing the performance of the Flink framework. This is achieved by harnessing the power of PCIe technology and integrating seamlessly with the OpenCL standard, which is tailored for heterogeneous systems. The JVM-FPGA communication mechanism is optimized using a data transmission pipeline mechanism. Through experiments and evaluations of two typical computationally intensive tasks, matrix multiplication and vector addition, it is demonstrated that the performance of the new framework is reduced in terms of latency, throughput is increased by 2.77 times, and CPU utilization is significantly enhanced. This study provides an innovative solution for efficiently processing largescale data streams in edge environments, successfully demonstrating that hardware acceleration can significantly improve the performance of the Apache Flink framework under resource constraints in edge environments, better meeting the continuously growing demands of IoT data.

Machine Anomalous Sound Detection is crucial for artificial intelligence automation in the context of the fourth industrial revolution. Recent approaches employ self-supervised representation learning, which combines representations extracted through training models on normal and pseudo-anomaly data in the upstream phase with anomaly detection algorithms in the downstream phase. Although effective in extracting representations and demonstrating some robustness, this approach overlooks the complexity and intrinsic relevance of the individual tasks in the upstream models. To address these challenges, this paper introduces the Self-Supervised Multi-Task Representation Learning with Artificial Fish Swarm Optimization. This network integrates two task classifiers into the upstream model, specifically designed to distinguish the sound origin of the target machine and discern attribute information related to the machines. We assign weights to individual tasks and optimize these weights using the artificial fish swarm algorithm. Our approach was evaluated on the DCASE2022 Task 2 evaluation dataset, where the method outperformed integrated models by an average of 6.8% in terms of AUC and pAUC.

In the research of lane detection based on deep learning, this paper proposes an ultra-fast lane line detection network based on hybrid attention mechanism. The hybrid attention mechanism network composed of channel attention mechanism and bidirectional attention mechanism is added in the lane detection network based on row classification. The network can effectively extract the structural features of long pixels like lane lines by capturing the global features of the image. And a loss function is introduced which can enhance the network convergence ability when the foreground target is small. This method improves the detection accuracy of the network while slightly affecting the speed but not affecting the real-time performance compared with the original network structure.

For the problem of static space and time dependencies based on traffic prediction in SDN traffic engineering, this paper proposes a dynamic network traffic prediction method, Attention mechanism for GCNGRU model (AGCNGRU), which integrates graph convolutional neural networks (GCN) with gated recurrent units (GRU) and incorporates an attention mechanism. By leveraging GCN, it captures the spatial dependency of traffic between nodes in the network, while GRU captures the temporal dependency of traffic passing through various nodes. The time attention mechanism is designed to assign weights to each hidden state, adjusting the importance of traffic information at different time points. Simultaneously, a data-driven spatial attention mechanism dynamically and adaptively adjusts the Laplace matrix, enabling dynamic extraction of spatial-temporal correlation in traffic data. This ultimately leads to accurate prediction of dynamic traffic. Experimental results on the GEANT datasets demonstrate that the proposed method significantly outperforms other approaches.

With the rapid development of the takeaway industry, the takeaway order path optimization problem has attracted extensive attention from many scholars. However, most of the existing researches only focus on customer satisfaction, ignoring the attention of takeaway workers. Therefore, this paper carries out an in-depth study on the income status of takeout workers on the basis of previous studies. First, this paper establishes a multi-objective order path optimization model, which not only considers the time delay of the order and the cost of the rider's driving path, but also takes into account the goal of the rider's income equilibrium, and fully integrates the take-out workers, merchants and customers. Secondly, an elite genetic algorithm with repair operator is designed to solve the objective function solving problem in dynamic scenarios. The arithmetic example is analyzed by real-time data of takeaway delivery. The results show that the model and algorithm proposed in this paper can effectively solve the problem, and the algorithm performs well in terms of solving efficiency and accuracy compared with the classical genetic algorithm.

Ship locks not only undertake the task of inland navigation, but also the water environment of ship locks is related to people's livelihood in inland river basins. Traditional supervision of ship pollution prevention and control in ship lock waters still adopts sampling monitoring, manual experience judgment, naked eye observation, manual recording and statistical analysis, etc., which not only affects waterway transportation, but also manually misses inspection, and the emergency response of sewage discharge lags behind, which makes it difficult to meet the requirements of modern operation and maintenance management of ship locks. Combined with the current situation of ship pollution prevention and control supervision in ship locks and the characteristics of digital twin technology, this paper uses digital twin technology from the perspective of operation and maintenance management to build an intelligent supervision system for ship pollution prevention and control in ship lock waters, which provides reference for promoting the intelligent supervision and construction of ship pollution prevention and control in ship lock waters.

Mechanical and electrical equipment on highways may affect traffic smoothness and safety after faults occur. By predicting sudden changes in fault signals, potential problems with mechanical and electrical equipment can be identified in advance, accelerating fault diagnosis and repair speed, and ensuring the normal operation of mechanical and electrical equipment on highways and road safety. Therefore, a wavelet threshold decomposition and mutation prediction method for fault signals of highway electromechanical equipment is proposed. The wavelet threshold decomposition method is introduced to denoise the fault signals of highway electromechanical equipment. Based on the singular value decomposition method, the feature vectors of the denoised fault signals of highway electromechanical equipment are extracted. On this basis, the Grey Wolf Algorithm Support Vector Machine model is used to predict sudden changes in fault signals of highway electromechanical equipment. The experimental results show that the proposed method has good performance in predicting sudden changes in fault signals of highway electromechanical equipment, and can effectively improve the efficiency of predicting sudden changes in fault signals of highway electromechanical equipment.

Unmanned aerial vehicles (UAVs) are widely used in search problem due to their portability and high performance. In UAV-assisted search problem, the path planning is considered as the coverage path planning problem, which is usually converted to a traveler's problem through the grid decomposition method. To solve this problem, this paper has designed an improved ant colony algorithm, which combines Q-learning based adaptive strategy, elite strategy and other methods to enhance the exploration and convergence ability. Simulation results show that the method can effectively improve the coverage efficiency of the multi-UAV multi-area coverage search problem and reduce the UAV flight energy consumption.

Emergency decision-making for unexpected events at civil transportation airports involves dynamic situations, time constraints and information scarcity. Rapid and effective post-emergency decision implementation remains a critical concern for relevant departments and emergency academia. With the development of graph neural networks, their application in recommendation systems has become a hot research topic. To improve recommendation quality, more and more researchers model data as an information network with two types of nodes: users and items. This approach facilitates more accurate knowledge discovery. However, existing studies often do not fully utilize the comprehensive structural and rich semantic information within the network. Therefore, this paper proposes an emergency recommendation model incorporating multiple types of entity-present learning networks to alleviate the pressure on commanders in responding to accidents. The experimental dataset was obtained on a simulation platform, and the experimental results show that our method has made significant improvements compared to advanced recommendation methods. Further research has demonstrated the effectiveness of the emergency decision-making method approach this paper proposed.

To guide the deployment of antennas in subway tunnels, the tunnel channel parameter prediction method based on Radial Basis Function is proposed. The channel parameter database for straight tunnel and curved tunnel with different turning radius at 1.4GHz is established based on Ray Tracing. The accuracy of the constructed channel parameter database is validated using measured data. By inputting coordinates of transceiver, distance, and turning radius, RBF neural network outputs key channel parameter predictions for communication system design including received power, delay spread, angular spread, and the Rician K-factor. The results demonstrate that the prediction model based on RBF can accurately predict tunnel channel parameters, making it suitable for the setup and optimization of tunnel wireless communication system.

As the aging population continues to grow, wheelchairs have emerged as crucial assistive tools in the elderly's everyday routines. Yet, the prevalent safety risks in traditional electric wheelchairs, mainly due to human operational errors, underscore the need for advancements. Addressing these concerns, this study introduces an innovative road segmentation approach utilizing an enhanced U-Net model tailored for intelligent wheelchairs. This technique adeptly segregates road surfaces, pedestrians, and obstacles, significantly bolstering the safety of wheelchair navigation. To curtail computational demands, the paper integrates a lean feature extraction mechanism inspired by GhostNet. Moreover, it presents a novel feature fusion strategy that marries coordinate attention mechanisms with skip connections, boosting the model's capacity for synthesizing information. The inclusion of pruning strategies effectively diminishes the model's parameter count, streamlining its efficiency. Empirical assessments reveal that our refined U-Net-based road segmentation method attains a mean Intersection over Union of 77.45% and a mean pixel accuracy of 85.62%, marking improvements of 4.58% and 4.82% over the traditional U-Net benchmarks, respectively. In real-world deployments within intelligent wheelchair systems, the proposed solution exhibits exceptional accuracy and robustness, heralding significant implications for enhancing mobility and safety for the elderly demographic.

The stability and reliability of vehicle sensor data such as high-precision integrated inertial navigation system (INS), millimeter wave radar (MMW), laser radar (LIDAR) are very important for the functional realization of vehicle ADAS. Due to the influence of natural environment and working environment, the perception data of the sensor may be biased. C-V2X uses the characteristics of interconnection to collect and analyze the information of traffic participants such as vehicles and pedestrians in the surrounding environment and the roadside information such as map information, and broadcasts it to the surrounding vehicles to provide traffic environment reference data for the test vehicles. In general, the calibration of sensors requires additional calibration equipment. In this paper, the C-V2X transmission data of vehicle networking is used as the truth value system to provide reference data for the real-time calibration of vehicle sensors, which can reduce the test hardware equipment, and the truth value system works steadily, providing the location information of objects with high accuracy. This calibration method can improve the accuracy of vehicle sensor, and provide reliable data support for multi-sensor fusion and system calibration.

Network security analysts in the daily monitoring and research work, will deal with a large number of security events, through the behavioral analysis of the details of the security event analysis and context correlation analysis to determine whether the real attack, and then according to the results of the research and judgment to be disposed of. These security events often originate from different brands of security equipment, and in the face of massive security event logs, it is difficult for analysts to consider all aspects, and omissions and misjudgments will inevitably occur. In this paper, we provide a security event clustering method based on the attention vector mechanism, which can automatically correlate the context of security events, extract the event sequences, and group the sequences with similar features into the same cluster, and the analysts only need to judge and dispose of the clusters, so as to achieve the purpose of enhancing the monitoring capability and improving the monitoring efficiency.

Aiming at recommending a personalized route, a route recommendation method based on Double-Tower Latent Factor Model (DTLFM) is proposed. First, the road network is separated by Density Peaks Clustering (DPC), together with a road-scoring formula concentrated on trajectory’s score. Then, DTLFM is employed to predict the road scores, and the Simulated Annealing algorithm (SA) is used to determine the optimal path by considering both the roads’ length and score. Experimental results demonstrate that the road network segmentation reduces running time by approximately 30% when the number of trajectories is increased to 200 or more, while the error of road’s ranking remains stable at 2.5 or lower. Compared with classical algorithms such as LFM and UserCF, this method can still maintain an accuracy of about 80% in small-scale data. What's more, taking the shortest path as standard, the optimal path obtained by this method can increase the average road satisfaction by at least 10%. Therefore, the route recommendation method based on DTLFM provides a new way to solve the problem of data sparsity and personalized route recommendation.

To address the dynamic characteristics of current network traffic demand, we first introduce a multilevel sleep mode for gNBs to enhance the flexibility and responsiveness of ultra-dense networks. We synthesize the global load of the network and propose a Cell association strategy based on load and energy efficiency (CALAE). We combine CALAE algorithms into the Dynamic sleep strategy based on multilevel sleep modes (DSBM) proposed in this paper. The DSBM algorithm can perform the cell association through the CALAE algorithm, and dynamically adjust the gNBs' operating state and sleep depth to minimize system power consumption and delay. The simulation results show that the DSBM algorithm proposed in this paper can effectively reduce the system power consumption and the transition delay of gNBs to meet the service demand of users.

With the popularization of smartphones, mobile applications and mobile Internet, mobile devices have an increasing demand for real-time and low latency. However, MDs constrained in their computational power and resources cannot entirely dependent on cloud computing for their processing needs. In order to reduce network latency and improve user experience, Mobile Edge Computing has emerged, and the research on computation offloading lays the foundation for the realization of MEC. In this work, in scenarios involving multi-users and multi-edge servers, we adopt the double deep Qnetwork strategy to address the issue of task offloading. Our primary objective is to reduce the total system latency while considering device mobility, task urgency, and the heterogeneous tasks. We extend the DDQN algorithm by adding a prioritized experience reaply mechanism. Experimental results indicate that the improved DDQN method enhances the convergence speed and effectively reduces the task latency relative to other baseline algorithms.

Based on the global principal component factor analysis and BCC-DEA model, this paper constructs an evaluation system for research efficiency of higher education institutions, and obtains the evaluation scores of each university based on the relevant data of Liaoning Province in 2020. The results show that: the overall scientific research efficiency of higher education institutions in Liaoning Province is high, but there are still some inefficient colleges and universities, and the management mechanism of scientific and technological projects and funds set up independently by universities should be improved in order to improve the scientific research efficiency.

Collaborative downloading, as a promising application in Vehicular Ad-hoc Networks (VANETs), can effectively address the demand of vehicles for high-quality resources in VANETs. However, there are still security and privacy issues in collaborative downloading. Therefore, in this paper, we propose a reliable, secure and privacy-preserving collaborative downloading scheme which effectively achieves message authentication and privacy protection. Besides, in the proposed scheme, we also use a homomorphic network coded signature to ensure sequential verification of file blocks and design a new vehicle selection algorithm, a reward and contract mechanism to improve the reliability of collaborative downloading. Experiments show that our scheme is reliable and practical.

In the last decade, due to the rapid development of GPS, electronic payment, and other technologies, as well as the popularity of smartphones, ridesharing has gradually emerged among people. This paper aims to investigate a new pathbased ridesharing user equilibrium model to research ‘ridesharing’. The travelers in this model include three travel modes: solo drivers, ridesharing drivers, and ridesharing passengers. In our model, a ridesharing driver can share a ride with passengers from multiple different OD pairs during the entire journey, and these passengers are not meet in this car. Meanwhile, each passenger is carried by only a ridesharing driver throughout the entire travel. Moreover, we optimize functions of travel cost to make it more realistic in our model. Finally, some numerical experiments are provided to illustrate the effectiveness and characteristics of our model. It is shown that: (1) Compared with other models, our model has lower generalized cost. (2) By inducing drivers to drive on the new roadways, the network may create a paradox.

In recent years, machine learning related techniques are widely used in edge computing scenarios. This has led to an increase in privacy concern due to the huge amount of data in end devices. To resolve this issue, federated learning (FL) has been proposed. Multiple participating devices perform distributed training locally, keeping training data local to train global neural network model. However, Limited network connectivity restricts parallel model updates and aggregation in federated learning across all devices. Additionally, the presence of non-Independent and Identically Distributed (non-IID) on diverse devices further slow the convergence of FL due to increased local and global model discrepancies. In this paper, we propose Federated learning on non-IID data with Reinforcement learning and Quantifying historical contribution (FedRQ), a non-IID FL framework. The framework aims to intelligently select clients to participate in each round in order to balance the bias introduced by non-IID data and promote model convergence. We evaluated this FedRQ under various FL tasks and verified that this framework outperforms other benchmark methods.

Compared to traditional urban delivery systems, the last-mile delivery has become a complex issue due to lower daily customer demands, significant delivery time variations, and a widespread distribution of locations. The delivery time of customer orders is associated with the choice of delivery locations, where considerations of capacity and operating hours arise, especially when utilizing shared delivery spots such as parcel lockers. To address these issues and emphasize the advantages of shared parcel lockers, we propose a delivery problem with delivery options and preference levels. We define a routing scheme for the coordinated delivery of vehicles and drones, integrating various types of delivery locations. We solve this problem using an adaptive large neighborhood search algorithm with simulated annealing, introducing specific ruin and repair operators. These operators, combined with various operators from the literature, undergo numerical experiments. In the specific context of this problem, our proposed method outperforms conventional algorithms, such as large neighborhood search and a single-parent genetic algorithm.

The tunnel project as an important part of urban transportation, its intelligent construction management is crucial to the development of the smart city. Tunnel construction scheduling is an important issue, as the construction conditions of tunnels often change due to construction techniques, geological conditions, and so on. Although previous studies have made great contributions to this, most of them usually ignore the changes in construction sequences and resource supplies, and do not establish a complete dynamic optimization model to make dynamic adjustments. In order to cope with the changes of construction conditions, a dynamic optimization model for construction scheduling is constructed by adopting the idea of dynamic planning and combining the W-RBS and S-LSM methods. The scope and resourcing of construction activities were defined and analyzed for logical, temporal, spatial, work continuity and objective constraints between activities, subject to allowable resource changes. The Genetic Algorithm is compiled by python, and its effectiveness is verified by combining with the specific arithmetic example. The results prove that through the model has the capability to optimize the duration for the changing construction conditions and developing a feasible optimization strategy.

Highway tunnels are one of the important infrastructures in urban highway and play an important role in ensuring safe and smooth traffic flow. In order to keep the tunnel in good operating condition during the use of the tunnel, as well as to ensure the travelling experience of users, the regular inspection and maintenance of highway tunnels is crucial. Aiming at the traditional manual inspection of highway tunnels with low efficiency, high time and manpower costs, and the difficulty of controlling highway tunnels, a highway tunnel inspection system based on inspection drones is designed. The system uses a UAV equipped with 3D lidar technology inside the tunnel to implement a comprehensive scan through a preset flight path and carries an optical camera, aiming to efficiently and accurately collect information inside the tunnel. The resulting data is processed by deep learning algorithms to accurately identify problems such as wall damage, dusty lighting fixtures and faded reflective strips, thereby improving the effectiveness and accuracy of tunnel inspections.

Traditional cloud computing is limited by bandwidth and transmission delay, and it has been unable to meet the current growing computing needs. As an extension of cloud computing, edge computing is a promising computing model that allows users to enjoy lower computing latency than cloud computing. This paper studies the partial task offloading problem in multi-user edge computing scenarios. Firstly, the problem is established as an optimization problem, and then the Markov decision process is used to describe the problem. Finally, an improved soft Actor-Critic algorithm is proposed to jointly solve the offloading rate, offloading node selection and resource allocation decision. The experimental results show that the algorithm can maintain good optimization performance compared with the baseline algorithm.

This paper presents an urban meteorological sensing and prediction system that employs augmented reality and MQTT technology. The system aims to address the issues of delayed data collection and inconvenient information transmission that are present in traditional urban meteorological detection methods. The system comprises three modules: a data acquisition module, a server module, and an AR client module. The system monitors environmental information in realtime using meteorological sensors deployed at city intersections. The data is uploaded to the server using the MQTT protocol and displayed in real-time on the server side. Additionally, meteorological prediction data is displayed to users through an augmented reality interface that combines Unity3D and VuforiaSDK technologies. The system brings a new experience and convenience to the smart city management and future residents' life, helps the city development towards the direction of intelligence, and provides new technical support and development direction for the future smart city development.

There are automated guided vehicles(AGVs) in automated logistics systems. Without effective on-line supervisory control strategies, there will be many conflicts and deadlock problems. Many researches concentrate on “one-shot” problem. The number of the agents is the same as the number of the tasks. But in the automated logistics system, an agent can be assigned many tasks at any time. The agent has to first move to pick up the materials and then unload the materials at another location. To solve these problems, this paper proposes a lifelong multi-agent path planning method in automated logistics system. This method doesn't entail selecting the shortest path for each agent, but instead focuses on maximizing the factory throughput by allocating tasks, choosing paths, and employing other methods. The method includes three main parts: assignment of tasks, candidate path determination and selection of the deadlock-free paths. It searches the candidate paths using the depth-first search algorithm, and avoids deadlocks and conflicts with the advanced time window algorithm. By adjusting the executing time of the task, the occupation time of path resources is reduced, which allows more AGVs travel on the path and is better than the method of AGVs wait at the nodes and edges. At the same time, the method allows the tasks executed as soon as possible, which also improves the throughput of the system. The proposed method is high-efficient by means of simulation, and the software based on the proposed method has been successfully put into practice in the material handling and distribution plant.

With the development of human economy and society, elevators have been regarded as one of the essential means of transportation in modern urban daily life. The usage of elevators is increasing in urban living, leading to a growing number of optimization issues regarding elevator operation schemes. This paper proposes an optimization scheme for segmented elevator operation based on simulated annealing algorithm, targeting the allocation of elevators in high-rise office buildings. The analysis and solution are conducted using MATLAB based on the simulated annealing algorithm. The obtained solution represents the optimal allocation scheme for segmented elevator operation sought in this paper. Subsequently, this scheme is implemented for circuit simulation in Proteus, which requires the functionality of the elevator serving specific floor intervals. Based on the calculation formula for elevator operation time in the simulated annealing algorithm and the "proportion" principle—where the actual running time of the elevator in a cycle compared to the total running time is equal to the ratio of the number of passengers transported by the elevator in a cycle to the total number of passengers—several elevator operation schemes are compared and analyzed. Through comparing the running times of elevators under different modes, it is concluded that the optimal operation scheme is indeed the segmented operation scheme.

With the advancement of technology, China is developing rapidly in the construction of emergency communication systems. Currently, emergency communication networks have an urgent demand for multimedia services and higher requirements for network bandwidth. When there are multiple different services in the network, traditional universal routing protocols are no longer applicable and cannot simultaneously meet the transmission requirements of different business flows. There is a common problem with the existing encoding aware routing in the power emergency communication MESH network. In order to improve network performance, a large amount of data flow is concentrated on certain nodes to create more encoding opportunities. When the data flow increases to a certain extent, these nodes will cause local congestion or even paralysis due to heavy load. A new routing metric has been proposed, and based on this metric, a coding aware load balancing routing strategy has been proposed. This routing strategy not only utilizes network coding to save network resources and improve network throughput, but also considers the load situation of nodes. It selects nodes with relatively light load as the next hop node to forward, achieving a balance between network coding and node load balancing. The simulation results show that in scenarios with different data transmission rates, this routing strategy can effectively alleviate network congestion, with higher network throughput and lower end-to-end latency.

In order to study the influence of different helmets on the braking reaction times (BRTs) of college student electric bicycle(e-bike) riders, a real-vehicle experiment is conducted to test the BRTs when riders wore different helmets in a closed road section. In addition to helmet type, riders’ physiological characteristics, riding workload, riding expectations and speed were also explored by ANOVA and correlation analyses, on the BRTs s of college student e-bike riders. The results showed that (1) wearing full and three-quarters half helmets significantly increase the BRTs of e-bike riders but half helmets do not significantly increase BRTs; (2) the faster the speed, the longer the BRTs for the e-bike riders; (3) driving e-bike every day of the week leads to a significant reduction in BRTs; (4) the higher the safety expectation the shorter the BRTs. A linear regression model for the BRTs of e-bike riders was established with the speed and dummy variables of whether to wear a full helmet, whether to wear a three-quarters half helmet, whether to drive an e-bike every day of the week, whether to drive an e-bike 2-1 days of the week, whether to have a very high expectation of safety, and whether to have a low expectation of safety as the independent variables. The goodness of fit of the model was 0.693, which is better predict the college student rider BRTs. The goodness of fit is commonly used as a goodness of fit indicator for multiple regression analysis, the larger the goodness of fit, the better the model is fitted. Finally, some suggestions for the types of helmets to be worn by riders with different perceptions have been put forward, and it also provides a theoretical basis for e-bike stopping sight distance and time to collision calculation.

The digital guidance system is widely used in urban planning, public transportation, and other fields. The lack of interactivity and the level of intelligence of this application need to be improved to become the focus of current research. Emotion recognition technology is a kind of emotion analysis technology based on artificial intelligence. It was presented in order to address these issues. This paper discusses the application of emotion recognition technology in digital navigation systems and its impact on user experience. Emotion recognition technology can better understand user needs and provide personalized recommendation services by analyzing user's facial expressions, voice tones, and other information, to improve the interactive and intelligent degree of the digital guidance system. Using the analytical hierarchy approach as a foundation, this paper determines the key role of emotion recognition technology in improving the interactivity and intelligence level of the system through weight analysis. The research results show that by analyzing the user's emotional state, the system is able to comprehend use's needs more effectively, provide more personalized and intelligent services, and thus improve user satisfaction and experience. Emotion recognition technology makes the guidance system more personalized and intelligent. At the same time, the technology can continuously optimize the system design and function through the analysis of user emotion, improve the innovation and adaptability of the system, and provide strong support for the maintenance and update of the system. The practical significance and application value of this research are significant in advancing the sustainable development of digital guiding systems.

Performance modeling and evaluation are crucial aspects of deploying 5G and future networks. This paper introduces a compositional method, specifically the Markovian process algebra PEPA (performance evaluation process algebra), to model and quantitatively assess the performance of UE (user equipment) mobility in the RRC (radio resource control) inactive state for 5G networks. The system’s response time is subsequently derived from the PEPA model under various scenarios. Particularly, experimental results demonstrate that this performance metric scales linearly with the number of UEs.

The objective of this study is to construct a simulation platform based on OMNeT++ to validate the potential performance bottlenecks in large-scale compute system networks. Networks composed of thousands of interconnected nodes may encounter unforeseen issues. During the validation process in a real environment, hardware fluctuations can lead to inconsistent results between actual and expected outcomes. Additionally, some servers may be leased and continuously running, which makes it impossible to perform performance tests on them and identify potential performance bottlenecks. Therefore, a simulation platform is needed to simulate and discover these underlying issues. To fulfill the aforementioned requirements, a conversion tool was developed initially to enable the simulation platform to recognize the network's topology. By integrating with OpenSM, the simulation platform acquired routing capabilities identical to those of a real network. Furthermore, the parameters of the simulation platform were adjusted to match the hardware capabilities of real network devices, successfully simulated the delay, bandwidth in point-to-point communication and bandwidth under congested communication scenarios. Finally, the simulation included the collective communication latency of the network. By comparing the simulation results with the data collected from the real environment, it was found that the error between the simulated and actual results was within 10%, thus validating the accuracy of the simulation results. The collective communication results obtained from the simulation were analyzed to identify potential network performance bottlenecks that may arise when running large-scale jobs in a real system.

In order to automatically classify questions related to diabetes raised by patients, enhance the performance of search results, and promote the development of automatic diabetes question-answering services, this study adopts the BERT-TextCNN model (Bidirectional Encoder Representation from Transformers-Convolutional Neural Networks) for classifying diabetes-related health queries. The model is compared with BERT model, TextCNN model, BERT-TextRNN model, and BERT-TextRCNN model. The results show that, compared to individual models and other fusion models, the BERTTextCNN model achieves the highest precision, recall, and F1 score, with values of 0.8532, 0.8530, and 0.8506, respectively. In the validation phase, the BERT-TextCNN model also demonstrates higher accuracy, indicating superior classification performance.

Aiming at the problems of insufficient light, slow extraction speed of human eye image and low accuracy of pupil center location in complex environment at night, a NIR pupil center location method based on machine learning was proposed. The algorithm was optimized and improved based on Yolov5s. The C3 module of Backbone network fused with the lightweight ECA attention mechanism module and replaced the CIoU loss function with the SIoU loss function to effectively reduce redundant boxes and accelerate the convergence and regression of the prediction box. The improved Yolov5s machine learning model was used to detect the human eyes and obtain the coarse location of the pupil. Then, the acquired image is preprocessed, and the Canny operator is used to extract the pupil edge contour. The improved ellipse fitting algorithm is used to fit the ellipse, and the ellipse is corrected to accurately locate the pupil center. Experimental results show that under the condition of insufficient and uneven illumination at night, the proposed method can quickly extract the human eye image and accurately locate the pupil center, which has good robustness and real-time performance.

The detection capability for low radial velocity targets is an important measure of radar performance. Facing the challenges that low radial speed targets cannot be detected in the frequency domain and the limitations present in time domain detection, this paper proposes a waveform parameter design method to improve this issue. Under the premise of ensuring a safe distance between the target and the main lobe in the time domain, this method first determines the possible range of values for the pulse repetition frequency. Then, considering the limitations of the radar's own hardware conditions, specifically the restrictions on the transmitted waveform pulse width, the method maximizes the pulse repetition frequency and pulse width based on these conditions, to achieve the best effect in time domain detection of low radial speed targets and maximize radar detection performance. Finally, this paper uses this method to provide a suitable selection scheme for radar waveform parameters at any moment in the dynamic process of missile targeting.

Transportation mode identification (TMI) is an essential component of intelligent transportation systems, which provides guidance for understanding travel behavior preferences. Existing studies primarily focus on motion attributes of trajectories, neglecting the auxiliary function of urban road-network information in TMI. Inspired by subgraph representation, we propose a novel TMI framework using road-network subgraph representation (TRSS) to tackle this issue. The framework consists of a graph attention network with labeling trick to capture the topology relationship between road segments, and a Bi-LSTM network to encode the motion characteristics. Finally, we combine road-network semantics and motion semantics to jointly train the model. Experiments on two trajectory datasets show that TRSS significantly outperforms existing TMI methods. And ablation analysis demonstrates that road-network subgraph representation can effectively enhance the performance of identifying model.

Analyzing course evaluation data is a key means to improve the quality of course construction for e-commerce majors. To this end, a course evaluation system for e-commerce majors based on convolutional neural network (CNN) and latent Dirichlet allocation (LDA) is constructed to realize the emotional tendency analysis of course evaluation and the acquisition of evaluation subject words. In terms of data collection, python crawler technology is used to obtain data. In terms of data preprocessing, Python's deduplication method and regular expression operations are used to complete data cleaning, and jieba is used to implement Chinese evaluation word segmentation. In terms of emotional tendency analysis, Word2Vec is used to convert text data into Word Embeddings, train and test the CNN model, and finally apply it to the emotional tendency analysis task. In term of sentiment topic analysis, TF-IDF is used to calculate the keywords of the evaluation data, and the LDA model is constructed to obtain the topics and subject words of the evaluation. Through experiments, it is found that the course evaluation system constructed in the article can realize the course evaluation of any e-commerce major group courses.

This paper explores the realms of face recognition through the lenses of head pose estimation and gender prediction using deep learning architectures such as ResNet and InceptionResnetV1. Our investigation into head pose estimation involved training a model to predict the three-dimensional orientation of human heads. Concurrently, we delved into gender prediction, constructing a model that accurately discerns the gender of individuals depicted in images through feature extraction and clustering techniques. Our findings contribute to the advancement of Facial recognition techniques, with implications for various applications such as human-computer interaction, demographic analysis, and targeted advertising.

Reconfigurable intelligent surface (RIS)-assisted received signal strength (RSS) positioning technology has attracted widespread attention due to its low deployment cost and ability to customize the wireless environment. Through regulating the RIS’s electromagnetic responses, the transmission channels are altered to amplify the differences in RSS values between neighbouring locations, thereby enhancing the accuracy of positioning. However, existing works only consider reflective RIS, and thus can only improve the positioning accuracy of users on a single face of the surface. To offer positioning services to users on either sides of the surfaces, we propose to utilize Intelligent Omni-Surface (IOS), a special kind of RIS that is capable of reflecting and refracting incoming signals concurrently, to assist the RSS-based positioning. Specifically, by modulating the RSS distribution on the two sides of the IOS through adjusting the reflection and refraction phase shifts induced by the surface, positioning accuracy for individuals on either sides of the surface can undergo effective improvements, so as to realize full-dimensional localization. However, since the reflection and refraction phase shifts of the IOS are coupled, it is challenging to jointly optimize them so as to minimize the positioning error. To address this challenge, within this letter, we investigated an IOS-aided full-dimensional localization system based on RSS, in which IOS is placed in the centre of the room to adjust the RSS distribution of users on both sides, and designed an IOS-aided localization protocol to coordinate IOS and users. Based on this protocol, we formulate a phase shift optimization problem to reduce users’ positioning error, and develop a phase shift optimization algorithm based on gradient descent for iteratively to solve it. Simulation results confirm the efficiency of the suggested IOS-aided positioning method.

In the medical field, the rapid development of cloud computing has brought many advantages to medical record data sharing. However, there are still problems such as lack of patient control over personal medical record data, privacy leakage and inefficient sharing. To address these challenges, an electronic medical record sharing scheme based on searchable attribute encryption in blockchain is proposed. Based on the attribute encryption algorithm, the scheme allows patients to independently set their medical record access policies according to their personal needs, realizes fine-grained access control of medical record data, and combines blockchain and searchable encryption technology to realize the multi-keyword search function, which improves the security and search efficiency of medical record data. Secondly the scheme designs an attribute revocation algorithm so that users can update the sharing privileges in time to ensure the security and privacy of medical record data. Finally, after security analysis and experimental verification, the scheme is able to realize the secure sharing of patients' personal electronic medical records with low computational overhead under the premise of protecting patients' privacy.

In response to the issues of missing context semantics and imbalanced datasets in Chinese toponym recognition tasks, this paper proposes a Chinese toponym recognition method based on Bi-Char2Vec and Non-Flat-Lattice Transformer. Initially, a new Chinese character embedding model, Bi-Char2Vec, is designed to capture the semantic representation of text in long sequences, mitigating the problem of missing context semantics in Chinese toponym recognition. Then, by utilizing Inter-Attention to interact between "character-word" attention, followed by encoding contextual information through the Self-Attention in Transformer; a global optimal tagging sequence is finally obtained by a Conditional Random Field layer. On public datasets, BIC2V-NFLAT improves the accuracy of Chinese toponym recognition, achieving an F1 score of 95%.

At present, the Internet and smart cities are developing rapidly, and all social services. Small target detection is widely used in these fields. This paper focuses on small target detection based on YOLOv5. The feature enhancement module is proposed to solve the problem of incomplete extraction of small target features by the model. An attention mechanism is added to the model, so that the detection model pays more attention to the region in the image that contains the target to be detected, increases the weight of the feature information in this part, enriches the feature space, and further improves the performance of small target detection. Finally, by combining the use of network pruning and knowledge distillation, the model is compressed to compress the model size and improve the detection speed under the premise of ensuring that the model detection accuracy is not affected. The experimental results show that our optimisation effectively improves small target detection in terms of both detection accuracy and detection speed, with FPS improved by 31.84 and AP improved by 1.7%.

Enhancing signal control efficiency through deep reinforcement learning constitutes a focal point of research in the field of traffic control. Existing methods primarily rely on copious state information and complex networks to attain precise timing schemes, which can result in decision biases and protracted training cycles. Addressing these shortcomings, a two-stage control algorithm is proposed. The algorithm employs discrete encoding of traffic states to optimize the exploration of the state space and enhances performance by utilizing a competitive Q-network structure. Concurrently, control is divided into two stages: in the first stage, traditional methods are employed to select a subset of traffic state information for phase decision-making, while in the second stage, reinforcement learning techniques are used to select another subset of state information for phase duration prediction. Experiments are conducted on the traffic simulation platform CityFlow, and results demonstrate that the proposed algorithm surpasses traditional control methods and reinforcement learning-based approaches, exhibiting higher traffic flow efficiency and reduced travel times.

To address the challenges encountered by wireless sensor networks (WSN) in processing complex computational tasks due to limitations in energy and computational resources, this paper has devised a three-tier architecture that integrates device, edge, and cloud layers. By incorporating Mobile Edge Computing (MEC) technology, the architecture aims to reduce the overall network energy consumption and enhance data processing efficiency. Within this framework, sensor nodes have the flexibility to offload computational tasks to MEC or cloud servers, or execute them locally, adapting to various scenarios and demands. Considering the increased energy consumption associated with transmitting large volumes of task data to the cloud when MEC opts for cloud offloading, this paper employs compressed sensing technology to compress the task data, thereby reducing the data transmission volume and energy consumption. Finally, a comprehensive model based on compressed sensing is constructed and analyzed, taking into account factors such as task volume, compression rate, and transmission distance. Through a variety of experimental scenarios and parameter settings, the paper comprehensively evaluates two critical performance metrics: task accomplishment time and energy consumption. According to the experimental results, specific suggestions are proposed to guide the arrangement of the timing and location of actual task offloading, with the aim of maximizing network performance and energy efficiency.

The solar energy harvesting communication system is the basic support layer that is associated with the Internet of Things (IoT). However, the inherent stochasticity of solar energy brings about extensive research on energy allocation in the system. In this paper, we propose an energy allocation algorithm for maximizing the data transfer (MDT) of a solar energy harvesting communication system while jointly considering the practical constraints of battery capacity and link capacity restrictions. We can solve it through the convex optimization problem, but the solution is extremely complicated. This complexity is the result of the coupling of energy allocation problems caused by restricted battery capacity and link capacity. To simplify the solution process, we propose a decoupled energy allocation (DEA) algorithm. Simulation experiments are conducted through real energy harvesting data sets, and the results show that DEA outperforms the other two heuristic schemes.

The purpose of this paper aims to study an intelligent cargo sorting technology based on multi-axis mechanical arm, which plays an important role in improving the working efficiency of logistics and manufacturing industries. The system designed in this paper is based on the STM32F103ZET6 microcontroller as the control core, utilizing color sensors and pressure sensors to obtain information on the color and weight of goods. It automatically controls a multi-axis mechanical arm to achieve tasks such as grasping and placing goods based on the obtained goods information and sorting requirements. Also it employs a WIFI module for communication with the host computer, enabling remote real-time monitoring and control of sorting operations. Experimental testing was conducted on 1,000 sets of goods samples with different colors and weights, for color recognition under unobstructed natural daylight conditions and in an indoor environment, yielding accuracy rates of 87.6% and 97.3% respectively. The weight recognition accuracy exceeded 99%. Compared to manual sorting, the average sorting speed of the designed system increased by 32%.

Intelligent fusion terminals are the core components of intelligent distribution stations in low-voltage distribution Internet of Things. To enhance the practical application of intelligent fusion terminals in distribution Internet of Things, a neural network-based intelligent fusion terminal testing method is proposed. The focus is on the neural network-based intelligent fusion terminal testing method and the construction of a neural network model based on modified incentive functions to support communication protocol recognition between intelligent fusion terminals and different devices, improving the detection efficiency of intelligent fusion terminals.

In light of the data exchange challenges inherent in collecting and processing heterogeneous network data within the Internet of Things framework, a thorough investigation was undertaken into multi-protocol parsing technology. Subsequently, a protocol semantic parsing gateway was conceptualized and implemented. This gateway orchestrates a semantic parsing system by defining parsing scripts and executing programs in two stages, thereby facilitating real-time semantic parsing of diverse protocol packets. Moreover, the gateway's extensibility allows for the incorporation of an unlimited number of protocols within the parsing framework, thereby enhancing the flexibility of data collection and processing. Experimental findings corroborate the stability and reliability of the gateway in executing timely protocol semantic parsing operations.

A digital-based method was proposed to address challenges in large-scale segmented hull automated spraying workshops, such as the production of toxic gases and solid particles during operations, hindering real-time monitoring of spraying activities. The method focuses on the mismatch between site data and process system performance, leading to poor film quality. The intelligent monitoring method of twin technology establishes a system architecture for a digital twin five-dimensional model, exploring twin model establishment and data-driven technology. Specifically, a new digital twin modeling method for spray film thickness was studied and validated through experiment-simulation comparisons, showing high accuracy in restoring film thickness and uniformity. The system leverages the iterative nature of the spraying process to accumulate and process on-site data, allowing for iterative optimization of spraying process parameters to enhance efficiency and quality.

In the domain of autonomous and assisted driving technologies, the precise detection and identification of road traffic signs are of paramount importance within the perception layer of such systems. To counteract this, our approach involves the selective identification of categories with more than 50 instances, using the associated street scene images as a foundational dataset. Through deliberate data augmentation techniques, we effectively combat the challenges posed by data scarcity and class imbalance. Subsequently, the YOLOv8n neural network algorithm and Slicing Aided Hyper Inference are employed to construct the traffic sign recognition model. The efficacy of the proposed method is tested on images acquired through in-vehicle dashcams and direct on-site photography. The empirical outcomes demonstrate that our YOLOv8n-based traffic sign intelligent recognition algorithm attains an exceptional accuracy rate exceeding 93%. This method significantly augments the accuracy and stability of driving assistance systems, thereby substantially improving vehicular safety.

With the progress of wireless communication technology and software technology, a large number of new intelligent applications and services are emerging, which puts higher demands on the latency and processing performance of mobile user devices. Traditional centralized cloud computing has bottlenecks in dealing with real-time and highly interactive services. As a result, edge computing has emerged to offload computing tasks to the network edge near users to reduce latency and enhance service experiences. This work proposes an Optimization Algorithm of Data Unloading and Energy Awareness Based on User Mobility (UM-OADU&EA) for mobile users' data offloading and energy-aware optimization issues. By considering the time-varying channel characteristics of users and their quality of service requirements, the model achieves effective offloading of task data and dynamic adjustment of system energy consumption. Simulation results indicate a significant improvement in user utility and a reduction in data offloading failure rates compared to traditional models. Specifically, as the average data and computation load increase, the UM-OADU&EA model improves user utility by at least 45%, with a maximum data offloading failure rate not exceeding 20%. The research on the joint optimization model of data offloading and energy awareness for user mobility has practical significance in the field of communication and computing fusion.

Wireless physical layer key generation is a remarkable technology for physical layer security. The security of key generation is not guaranteed in existing wireless physical layer key generation schemes if the eavesdropper is less than half a wavelength away from the legitimate node, which called non-secure zones problem. To address this problem, this paper proposes a Wireless physical layer key generation scheme based on unscheduled forwarding signal (WKG-UFS). Specifically, the legitimate communication nodes send random signals to each other within a coherent time. Then, the received signals are forwarded back within the coherent time after a random time interval. The legitimate nodes estimate the equivalent channel based on the final received signal and generate the secure key. Since eavesdropper does not know the original signals sent by the legitimate communication nodes, the channel feature cannot be effectively estimated. The results of theoretical analysis and experimental simulation show that the proposed scheme can effectively reduce the threat of eavesdropping on secure communication. Besides, since the time intervals for receiving and forwarding signals are variable, it can also effectively resist active attacks which by manipulating the channel.

In the era of digital transformation and data-driven decision making, mall operators seek innovative ways to attract and retain customers. The retail industry faces significant challenges in understanding and targeting customers effectively. This paper proposes a new approach to address this challenge by proposing an intelligent application that utilizes facial recognition technology to determine marketing strategies for shopping mall customers. This paper comprehensively studies the application of facial recognition technology in the marketing strategy of shopping centers to identify customers. This innovative solution leverages the power of artificial intelligence to analyze customer demographics, emotions, and behaviors, enabling shopping centers to customize their marketing efforts, increase customer satisfaction, and drive revenue growth. The importance of using FDD research method to develop and study face recognition system for shopping malls. At the same time, it will be concluded that by analyzing customer data and customizing marketing efforts, shopping malls can provide personalized experiences to attract and retain customers in an increasingly competitive market.

In urban arterial intersections, unreasonable signal coordination can lead to a decrease in the efficiency of arterial traffic flow and an increase in vehicle start-stop frequencies. Phase difference is a crucial parameter for arterial signal coordination, as the phase difference between intersections affects the delay experienced by traffic flow. This paper focuses on the phase difference of consecutive intersections and proposes an optimization algorithm for arterial phase difference. It introduces a utility function based on the correlation analysis of multiple phase differences, utilizes neural networks to describe the relationship between phase difference and arterial delay, and employs genetic algorithms to find the optimal solution for the model. Simulation results demonstrate that the proposed arterial phase difference optimization algorithm outperforms traditional uncontrolled and induction-controlled methods, yielding superior signal control effectiveness.

This study explores a method integrating BERT and BiLSTM technologies aimed at enhancing the accuracy of sentiment recognition in online shopping reviews. Through detailed experimental validation, the fused model significantly surpasses traditional approaches that solely rely on either BERT or BiLSTM across various key performance metrics. This work not only confirms the potential of deep learning techniques in understanding complex textual emotional expressions but also provides new research directions and practical application possibilities for sentiment analysis of social media data.

In order to study the applicability of contraflow left-turn lane design, this paper analyzes the traveling trajectory and signal phase of contraflow left-turn lane, establishes a delay model for contraflow left-turn lane, explores the influence of different arrival rates and contraflow left-turn lane lengths on the delay, and obtains the range of contraflow left-turn lane lengths that minimize the delay in the range of different arrival rates, and finally analyzes the delay results through the VISSIM simulation software, which is secondarily developed through COM interface using Python programming language. Finally, through the VISSIM simulation software, we use Python programming language through the COM interface to carry out secondary development, build a contraflow left-turn lane design intersection environment to carry out simulation experiments and analyze the delay results. The timing of contraflow left-turn lane opening is obtained under different lane capacity of contraflow left-turn lane design.

The rationally layout of connecting line is of significance for improving network efficiency and network planning. Under the background of advocating “network integration” and resource sharing, An urban rail transit connecting line planning model considering maintenance resource sharing is proposed, which minimizes the total construction cost of connecting lines and the total transfer time of trains require maintenance. A hybrid algorithm combining CPLEX and NSGA-II is applied to solving the model. The results of case study show that compared with the existing method, the proposed model achieves better effect in maintenance resource sharing based on realizing connection of all lines. Besides, key parameters of the model are discussed. The obtained results provide a consult for the layout determination of connecting lines and urban rail transit heavy repair sharing depots.

This paper analyses and summarizes a variety of underground space demand forecasting methods, and selects an appropriate underground space demand scale based on the calculation of multiple methods and the results of underground space resource assessment. Based on the functional coupling theory, a multi-factor demand assessment model is established to delineate the priority construction zones of underground space; it promotes the practice of the basic technology of smart planning, and facilitates the realization of the goal of smart planning for urban underground space.

This paper provides a comprehensive overview of the development and implementation of smart home system. It delves into the structural design of a smart home control system, explaining its various components, such as power supply, controller core board, and functional sub-modules. It also discusses the working theory of the system, emphasizing the importance of user interaction, data transmission, and sensor-based data collection. The paper also covers the software aspects, discussing the programming and operational algorithms that govern the system's behavior. Additionally, it explores the challenges and solutions in data transmission and management within the smart home network. The practical applications and potential future developments of the system are also addressed, highlighting its efficiency, scalability, and adaptability in modern home automation scenarios.

The present study explores the application of Internet of Things (IoT) technology in the environmental quality and safety monitoring system of construction sites. Environmental pollution has become a global challenge, and long-term exposure to unhealthy environments can induce various diseases. The IoT technology can provide real-time monitoring of environmental parameters such as temperature, humidity, and air quality, which can help identify potential hazards and prevent accidents. The data collected from the monitoring system can be analyzed using machine learning algorithms to develop predictive models for environmental quality. The study highlights the benefits of using IoT technology in improving environmental quality and safety in construction sites and discusses the challenges in implementing the technology. Overall, the study provides insights into how technology can be leveraged to address environmental challenges and promote sustainable development.

The reliability of bus network is an important factor affecting the choice of travel. It is also an important index for the adjustment of bus network and the optimization of operation and scheduling strategy, which is of great practical significance to enhance the attraction of bus. In the actual traffic system, the bus network will face various challenges, such as weather changes, traffic accidents and so on. People will change their original travel plans, leading to fluctuations in demand, causing road congestion, vehicle delays, increased travel time, and ultimately reducing reliability. In order to improve the reliability of bus network, the randomness of demand will be considered in this study. The stochastic demand is depicted through different scenarios, and the transit assignment model based on line node strategy is established. According to the results, the reliability of the bus network is analysed from two aspects: travel time and waiting time, and the reliability improvement strategy is proposed. The example analysis shows that adjusting the bus frequency based on the weighted passenger flow can effectively improve the reliability and the operation efficiency of bus network.

Current traffic flow forecasting is a crucial component within intelligent transportation systems (ITSs). However, most existing studies still cannot be employed on large-scale data flow forecasting in real-world scenarios. In this paper, we design a spatiotemporal neural network called SINN for large-scale traffic flow forecasting tasks in real-world scenarios. First, we propose a transformer-based neural network to integrate the historical traffic flow into further traffic flow. Second, we propose a structural information-based hierarchical graph neural network module suitable for large-scale network structures, which can capture relevant node features within large-scale road networks for forecasting future traffic flow. Finally, we collect a large-scale traffic flow dataset and execute evaluation experiments. The experimental results indicate that SINN is capable of processing datasets with thousands of nodes. Compared with the methods that can execute the same scale datasets, SINN can make significant improvements by 44.29%, 26.54%, and 25.23% in the metrics of MAE, RMSE, and MAPE.

CORS-RTK network based on BeiDou navigation system enjoys a wide application in fields of infrastructure construction, but its applicability in urban road reconstruction and expansion with complex road conditions calls for further exploration. According to the precision analysis results drawn from a number of horizontal and vertical comparison experiments based on the real urban road conditions, the CORS-RTK network delivers a better performance than other measurement methods in terms of precision, convenience, and reliability under various urban road conditions. The precision of vehicle-mounted CORS-RTK at a normal speed (the speed that bears no influence on the traffic) allows alignment measurement in urban roadway reconstruction and expansion. In addition, the CORS-RTK network based on BeiDou Navigation System put in the same or even better performance than CORS-RTK based on GPS under most urban road conditions. As practice indicates, the CORS based on BeiDou Navigation System gives a satisfactory performance in precision, convenience, reliability, and coordinate secrecy, which enables its wide application in urban reconstruction and expansion.

Microalgae cultivation is a promising next-generation food production industry to produce sustainable and nutrient-rich food sources. However, the existing cultivation techniques are mostly designed for large-scale industrial applications, with minimal automation to monitor and control growth conditions for small-scale production. There is a need to explore the feasibility of implementing a small-scale indoor and home-based cultivation system. Therefore, this paper proposes the smart incubator, a novel automated solution which integrates a variety of sensors, actuators, controllers, and cloud communication for effortless cultivation of Spirulina. It utilises the Internet of Things (IoT) concept to monitor and control the vital factors affecting Spirulina’s growth, such as growth medium, salinity, temperature, light intensity, pH, aeration, and agitation. Additionally, the proposed system is designed with several key functionalities including automated growth monitoring and controlling, auto-harvesting, and sustainable medium recycling. A comparative analysis of the growth rate of Spirulina cultivated with the proposed system and manual cultivation method highlighted that it is a potential solution that optimises Spirulina’s growth. However, it should be brought into further investigation and refinements to increase its feasibility as a commercial product. The discovery of machine learning integration with the system is also vital to make it more flexible in keeping optimum growth conditions.

Cloud computing and digital information technology provide strong support for achieving efficient and sustainable ecological community governance. This article explored the application and advantages of cloud computing based digital information technology in ecological community governance through empirical research. Firstly, the development background of cloud computing and digital information technology, as well as their applications in other fields, were reviewed, providing a theoretical basis for studying the application of ecological community governance. Secondly, the specific application methods of cloud computing and digital information technology in ecological community governance were analyzed, including data collection, analysis, and visualization, in order to provide guidance for practical operations.

Following operational interruptions in the urban rail transit system, affected trains are typically organized to perform turn back operations at stations with turn-back capabilities, aiming to minimize the impact of the interruption on the network. Considering potential interruptions in the urban rail transit network, this paper presents a method for selecting turn-back stations during the network planning and construction phase to optimize the network structure and enhance resilience. With the uncertainty of interruptions in the urban rail transit network in mind, a stochastic optimization model is formulated in this paper, with the objective of maximizing the expected network performance under various interruption scenarios. The aim is to select the positions of a limited number of turn-back stations in the network so that the network can maintain a high level of resilience in possible interruption scenarios. The number of passengers transported on the shortest path is utilized as the performance metric for the urban rail transit network. Finally, a case study is conducted using part of the Beijing subway network as an example. The result indicates that the resilience of the urban rail transit network is increased by 7.12% compared to traditional methods by the proposed approach in this paper.

The demonstration application of freight automated driving is increasingly widespread, and it has been used in production and operation in some areas. Automatic freight driving aims to complete effective cargo transportation, and has clear economic requirements such as reasonable input and improved transportation efficiency. Meanwhile, it also needs to ensure transportation safety. At present, the effect evaluation of the demonstration application of freight automated driving has not been carried out quantitatively. The demonstration application of freight automated driving only has "yes" and "no" problems, and there is no "good" and "bad" problems, especially the lack of systematic and reasonable evaluation index system and evaluation methods. This paper first through the research, hierarchical analysis method to establish freight autopilot demonstration application overall architecture, based on the construction of the overall architecture, with safety, efficiency, economy, using hierarchical analysis, multiple analysis, according to the "evaluation object extraction, evaluation element extraction, evaluation index construction" steps, construct covering autopilot function, transportation service efficiency, infrastructure, social acceptance of multi-dimensional multi-level index system, and gives the definition of index, and index range. There are some shortcomings in this paper. The index system of the demonstration and application of freight automated driving is only constructed. The data collection of the specific demonstration application is still in progress, and the evaluation and verification of practical engineering has not been carried out. Follow-up studies can continue to build the evaluation model and carry out the engineering evaluation verification.

With the continuous improvement of residents' living standards and the continuous development of artificial intelligence technology, community APP has become an important carrier to meet the needs of the community. By improving the functions of the existing community APP, combining artificial intelligence technology, and using collaborative recommendation algorithm and clustering algorithm, this paper builds an intelligent community APP that matches the needs of community residents. Put forward an AI-based intelligent community management model, enhance the digitalization and intelligence of the new model of community service, and make the AI functional community model more responsive to the needs and trends of The Times.

In the current network landscape, operators deploy network functions on physical devices (such as switches and middleboxes) as well as on Network Function Virtualization (NFV) devices (e.g., virtual machines on general-purpose servers), forming service function chains that meet specific requirements. This approach efficiently leverages the processing capabilities and deployment flexibility of both types of devices. However, the heterogeneity between devices may lead to the independent generation of multiple conflicting solutions within the network, making it challenging to identify an optimal deployment strategy. To address this challenge, this paper introduces a Taboo Search Algorithm Based on N-jump Search Solution Space (TSABNS). This algorithm aims to minimize the average latency cost of Service Function Chain (SFC) deployments while ensuring that most SFCs are successfully deployed within the network. Comparative simulations demonstrate that TSABNS significantly outperforms other algorithms, such as HOPE, in terms of deployment success rate and latency performance.

In order to meet the travel needs of passengers from residential areas far away from subway stations to subway stations, a time-windowed first-mile feeder DRT route optimization method has been studied. Firstly, based on k-means clustering algorithm, the passenger travel coordinates are clustered to obtain DRT stations. Secondly, a demand-response bus model with the minimum total cost of the system as the objective function, passenger time window, rated passenger capacity of the vehicle, and constraints on boarding and unloading stations is constructed. Then, an improved genetic algorithm for customized cross operation is designed. Finally, an example is given to verify the effectiveness of the model and the algorithm. The results show that the optimized model can output a reasonable multi-DRT route vehicle scheduling scheme, which not only improves the attractiveness and operation efficiency of the bus system, but also satisfies the diversified travel needs of passengers.

The root cause of traffic congestion and reduced capacity is uneven traffic distribution. To study the operation mode of mixed traffic flow composed of CAV and manual driving vehicles, an allocation model for mixed traffic of CAV and manual driving vehicles is established, and an algorithm is provided to solve the model. With the goal of optimizing system travel costs, using Sioux Falls classic network as example, the impact of CAV on system travel costs with or without guidance is analyzed. The research results indicate that the addition of CAV has a positive impact on urban road traffic, especially when the demand is high. Under the penetration rate of medium to low CAV (0-0.6), when guided, CAV can significantly reduce travel costs and traffic congestion. With the continuous increase in the penetration rate of intelligent connected vehicles, the traffic capacity of the transportation system has become better. Even if intelligent connected vehicles choose a balanced path selection method for users, the transportation system will still improve and approach the optimal system.

In order to address the challenges of urban traffic management and pollution control, particularly within the construction industry, this study focuses on analyzing construction waste truck movements using GPS sampling data. By analyzing the travel patterns and routes of these trucks, we aim to understand their impact on traffic flow and pollution levels in urban areas. Utilizing statistical analysis and clustering techniques, such as the Dynamic Time Warping (DTW) algorithm, we identify key characteristics of construction waste truck movements and their implications for traffic control and environmental sustainability. This research contributes to the development of intelligent transportation systems and strategies for mitigating urban pollution.

Due to the increase of emergency situations, in order to make the city gymnasium respond to the wartime state quickly, it is necessary to consider the construction of the micro energy network system combining peacetime and wartime. In order to avoid excessive waste of resources, the stadium should be prepared at the early stage of construction to meet the two optimal configurations for peacetime and wartime use. Two sets of optimal configurations in peacetime and wartime scenarios were obtained through the double-layer optimization algorithm, and the configuration capacity and operation scheduling were analyzed. The optimized system was rationally allocated and scheduled in different scenarios, and air source heat pump and energy storage equipment were reserved to ensure real-time energy supply during the peacetime transition period.

With the vigorous development of intelligent transportation systems, data-driven traffic flow forecasting has played a crucial role in alleviating traffic congestion and reducing traffic accidents. In recent years, the completion algorithm based on tensor has been widely used as an effective traffic prediction method. However, most of the methods only consider the global low rank structure of data and ignore the nonlocal spatio-temporal self-similarity prior characteristics, resulting in low prediction accuracy. This paper proposes a traffic flow tensor prediction framework(NSS-TENSOR) that integrates nonlocal self-similarity features and global consistency of traffic data. A novel traffic flow tensor with strong spatiotemporal correlation is reconstructed by using the dual advantages of nonlocal spatiotemporal self-similarity prior and tensor structure can automatically learn the global consistency of traffic data. Firstly, the nonlocal self-similarity prior characteristics of traffic flow data are captured by clustering algorithm and K-nearest neighbor (KNN) algorithm. Then the pattern information blocks are matched based on the prior features of nonlocal self-similarity, considering the global consistency information simultaneously, a new brief tensor is reconstructed. Through experiments on two real-world datasets, the results show that the proposed method in this paper has higher traffic flow prediction accuracy compared to the original tensor algorithm model.

Automated driving is nowadays considered as a solution to solve traffic congestion and increase driving comfort. However, it will be commercialized only until its functionality and safety are validated. Due to the complexity of traffic situations billons of stochastic test kilometres are required for the validation, which makes the validation in real world expensive. Simulation-based test is a way to reduce the cost. Traffic flow models can generate the various test scenarios required in simulation-based test. However, the traditional modelling methods are normally aimed at realistic representation of a real traffic flow and can hardly be proven that the generated traffic flow can have a high coverage over the corner cases required by automated driving validation. This paper introduces a method for microscopic modelling of stochastic traffic flow for automated driving validation. Its quick coverage over corner cases for automated driving is proven.

With the continuous improvement of urban critical infrastructure systems, the interdependent effect between different urban infrastructure systems becomes more obvious, that is, when a critical infrastructure system fails, the system coupled with it will also be affected by it. Therefore, how to measure the interdependent relationship between systems, determine their influence scope and influence path, and reduce the failure probability of critical infrastructure systems are particularly important for improving urban stability. In this paper, the interdependent mechanism between multi-layer critical infrastructure systems is analyzed and combined with the topology diagram, the interdependent topology mechanism model of multi-layer critical infrastructure systems is proposed. On this basis, combined with graph neural network and expert experience, a dynamic multi-layer critical infrastructure interdependent system prediction model is proposed. This model introduces time sequence information to update the relevant formulas of edge weights and nodes in the static network, so that the operating state of critical infrastructure can change with the change of time and space, so as to build a two-layer dynamic coupling network of critical infrastructure systems, which makes the model more close to the actual physical failure process and provides a basis for the subsequent prediction of the failure probability of critical infrastructure.

With the passage of time, China's current policies strongly support the development and application of the Internet of Vehicles. As a result, the application scenarios of the Internet of Vehicles are becoming increasingly significant and efficient, and the level of intelligence is constantly improving. However, the development of the Internet of Vehicles is accompanied by the diversity of traffic environments, and a lack of unified standards prevents the realization of different services. This paper primarily analyzes the utilization of digital twin technology to project the characteristics, objects, and behaviors of physical entities and scenes into virtual cyberspace. This allows for multiple verifications and, when combined with Internet of Vehicles technology, can address this issue. Additionally, the paper explores how to provide users with relevant one-to-one intelligent services to effectively resolve common problems faced by road owners in the new era. Finally, it proposes the potential of combining the digital twin model and Internet of Vehicles in various fields in the future.

Daily long-time traffic flow prediction is a crucial urban computing problem that aids in rational planning of traffic routes and efficient allocation of traffic resources. Existing models, while improving the accuracy of long-time traffic flow prediction, also increase computational complexity and additional space overhead. To address this, an optimization method named SUTDGCN is proposed, leveraging spatial upsampling and s graphs to enhance the real-time performance of long-time traffic flow prediction. In the spatial domain, spatial upsampling involves augmenting the original road network with K virtual nodes for upsampling, thereby constructing an upsampled road network to adequately capture both local and global spatial correlations. In the temporal domain, a simple directed graph is first constructed to represent local and global dependencies of time slices and then gated graph convolution is employed to further learn the underlying temporal dependencies. Experimental results on real datasets PeMS04 and PeMS08 demonstrate that, the SUTDGCN outperforms the STUGCN，it reduces MAE by 4.5% and 5.8 %, RMSE by 2.2% and 2.3%, and MAPE by 2.8% and 6.8%, respectively. Meanwhile, the SUTDGCN model reduces the time required for ssstraffic flow prediction by half compared to the STUGCN model.

In order to improve the accuracy of the evaluation of the development level of smart city, an evaluation model based on the hierarchical analysis method and BP neural network is proposed. Firstly, the evaluation index of smart city development is extracted and analyzed by hierarchical analysis method, forming a scientific evaluation index system. Then, the evaluation index data is obtained by using the index score as the input vector of BP neural network and the development level as the output vector, thus establishing the evaluation model of AHP-BP. This has certain reference value for the development evaluation of other smart cities.

As a new way of sports, road cycling is welcomed by people, but also brings a lot of traffic safety problems. Traditional cycling gestures are no longer suitable for road riding. Safe and reasonable transmission and expression of riding intentions are urgent problems to be solved. In order to increase the interaction and safety of riding, this study proposed a scheme of binding keys to link taillights. The taillights were used as the carrier to redesign the riding gesture symbols and solve the problem of missing information symbols of traditional taillights. Through sound and taillight symbols to convey the intention of riding, can improve the safety of riding. As a new way of movement, road cycling has been sought after by people, but it also brings a lot of traffic safety problems. Traditional cycling gestures are no longer suitable for road cycling, and safe and reasonable transmission and expression of riding intentions are urgent problems to be solved. In order to increase the interaction and safety of riding, this study proposed a scheme of binding keys to link taillights. The taillights were used as the carrier to redesign the riding gesture symbols and solve the problem of missing information symbols of traditional taillights. Through sound and taillight symbols to convey the intention of riding, can improve the safety of riding.

Facing the actual demand for the development and application of automated driving technology in the field of freight transport, for the new mixed flow environment of highway multi-driving mode and multi-vehicle, automation technology will be effectively combined with the real economy, and intelligence can better improve the living standard of the people, and also improve the national ability to resist major risks and the level of integration of big data in various fields of integration and computing fusion and innovation. In this paper, we study how to set up the problem of self-driving freight corridor under the limited road resources taking into account the efficiency and safety objectives. It proposes an analysis method of traffic flow operation characteristics under mixed flow environment, and establishes a capacity and safety characterisation model as well as a mixed distribution model of passenger and freight flows.

Urban traffic vehicle detection is a key component of smart city transportation systems, aimed at improving traffic management and safety through modern technologies and information methods. In response to the characteristics and challenges of vehicle detection in smart cities, this paper proposes a vehicle detection method based on drone aerial images, employing an object detection algorithm based on YOLOv4, namely the Adaptive Cropping YOLO algorithm. Through training and optimization on a large-scale dataset, this method can accurately detect and identify different types of urban vehicles. Experimental results show that this algorithm can effectively detect large-sized image targets that traditional YOLO algorithms may miss, providing reliable technical support for traffic safety monitoring and management.