This PDF file contains the front matter associated with SPIE Proceedings Volume 13422, including the Title Page, Copyright information, Table of Contents, Introduction, and Conference Committee information.

Emergencies are characterized by suddenness, complexity, and high uncertainty. Drills to respond to emergencies mostly use actual combat methods, which consume a lot of manpower and material resources. The frequency is low, and emergency personnel are not familiar with the collaborative handling of emergencies. It's hard to improve. This paper proposes a virtual collaborative emergency response method for road emergencies in flood season based on Multi-Agent. The transportation committee, highway branch, and maintenance group of the relevant units of road emergency response are used as participants in emergency response. Combined with the characteristics of road emergencies in flood season, Intelligent agents such as scene agent, monitoring agent, rescue agent, and disaster-affected Agent agents are established in the virtual scene to represent the transportation committee and other departments for collaborative emergency response, providing a more comprehensive, realistic and efficient virtual collaborative drill method for emergency response, thereby improving the emergency response department emergency response capabilities and collaborative emergency response capabilities.

With the development of urbanization and globalization, a large number of traffic lines have emerged in China. Nevertheless, they face potential earthquake risks due to the seismic activities caused by the well distributed earthquake belts. For the sake of their post-disaster operation safety, it is essential to (1) evaluate the emergency-response capacity and (2) develop corresponding recovery scheduling plans. This study proposes a probabilistic methodology for transportation networks by integrating the occurrence probability of earthquakes, vulnerabilities of components, topologies and traffic demands of networks, etc. The two main challenges, i.e. real-time post-earthquake residual functionalities of highway segments and reliable operation realization of networks, in that process are addressed by intelligent surrogates. To achieve that, this paper investigates the functionalities of transportation networks in emergencies, formulating a multi-dimensional emergency resilience vector. Moreover, a decision tree-based approach is introduced to determine potential emergency responses and their durations using historical event data. The results are expected to optimize mitigation, preparedness, response and recovery procedures for transportation systems.

At present, because shipping companies cannot provide detailed and effective information on the risk cost management of ship shipping, the risk cost control of ship shipping is basically carried out by relying on incomplete information and the relevant experience of management personnel. In order to improve the management level of ship shipping risk of shipping enterprises, this paper makes a preliminary economic analysis and discussion on the risk cost management of ship shipping enterprises, and makes a new definition and classification of the composition of ship shipping risk cost, gives the evaluation index of ship shipping risk cost management and the technical and economic analysis and control method of ship shipping risk cost management, establishes the correlation analysis model of ship shipping risk cost-benefit system, and conducts an empirical analysis of this correlation model. It is concluded that in the shipping risk management of the shipping enterprise, the size of crew training cost is the most correlated with the reduction of shipping risk loss cost, followed by the cost of ship preventive maintenance, the cost of safety asset investment, and the cost of environmental protection. By analyzing and finding out the key factors affecting the risk management of ship shipping, it is beneficial for enterprises to use limited financial resources to guide reasonable ship risk management, which is of great practical significance for seeking the best ship risk cost.

In view of the current situation that the construction goal of highway traffic safety facilities in China is not clear, according to the traffic safety influencing factors, the traffic safety facilities construction influencing factors are divided into road conditions, traffic conditions, service conditions, environmental conditions and driver conditions, through the use of hierarchical analysis and the fuzzy comprehensive evaluation method, the establishment of highway traffic safety facilities influencing factors model, based on the results of the comprehensive evaluation of the determination of the safety facility System specific construction objectives, to road section traffic safety facilities set the main factors affecting the proposed specific traffic facilities to improve the proposal, for highway traffic safety facilities system refinement to provide specific methods.

Low technical grade and poor route design are common issues with rural roadways, and certain parts even develop high accident rates. This study designs a set of comprehensive traffic safety evaluation systems with the goal of conducting a thorough and systematic safety review of rural roads in order to successfully raise the level of traffic safety on those routes. The model has been successfully implemented on a rural road, and the outcomes demonstrate its great practicability and outstanding effectiveness in assessing the degree of road safety. The evaluation model put forth in this research has broad practical implications and potential applications in addition to offering crucial theoretical support for enhancing rural highway safety. It also exerts a considerable influence on traffic management efforts.

Since GNSS signals tend to be blocked in underpass environments and general SLAM methods still struggle with cumulative errors in large-scale scenes, traditional localization methods are more likely to degrade in GNSS-denied areas. To address this problem, we design a vehicle-road-collaborative-based global positioning system, which combines the vehicle-end local positioning module and the road-end 3D target detection module to conduct global localization. Specifically, the vehicle utilizes the pre-calibrating global information of road-end system to bridge the gap between the local coordinate and the global coordinate. To further improve the positioning performance, we analyze the uncertainty of the perception results of the road-end system, and constructing a novel factor graph to fuse traditional self-motion constraints and the perception information of the road-end system. Meanwhile, to alleviate the communication delay problem, we build a dynamic interpolation module to align the observations at diverse sampling timestamps. We demonstrate the performance of our method by conducting comprehensive real experiments. Our qualitative and quantitative experimental results obviously outperform those of competing methods.

Considering the potential influencing factors such as roadside landscape, driving time, road alignment, vehicle type, etc., and combining the results of the driver questionnaire survey, we constructed a simulation scenario, designed a driving simulation experiment, collected a number of fatigue indicators such as the percentage of zero speed of the steering control of the driver, and constructed a driving fatigue evaluation model, based on which to identify a typical combination of unfavorable factors when driving fatigue occurs, and analyze its impact on the driving performance of the driver.

This article introduces the Bus Rapid Transit (BRT) system while analyzes applicability for Bus Rapid Transit (BRT) in domestic cities at macro aspects from city scale to economic status and micro aspects, including passenger flow needs and roads condition. Six cities are selected as representative cases where the constructions of BRT systems have already been finished. The article also discusses BRT system construction by conducting cluster analysis to these abovementioned cases.

Due to their environmentally friendly and energy-saving characteristics, electric freight vehicles (EFVs) have emerged as indispensable and significant modes of urban transportation. However, the limited driving range of vehicles often leads to low transportation efficiency. In this research, an eco-driving model is introduced specifically tailored for connected and autonomous electric freight vehicles (CAEFVs). This model incorporates wireless charging technology by deploying wireless charging lanes at multiple signalized intersections, allowing CAEFVs to replenish energy while moving slowly at traffic lights. At the same time, differentiated passage strategies are designed to meet the charging requirements of different vehicles while ensuring uninterrupted passage through intersections. The traffic benefits of these passage strategies are analyzed considering different initial speeds and wireless charging areas (WCA). The study reveals that the charging-optimum mode (COM) outperforms the time-optimum mode (TOM) in terms of charging efficiency, albeit requiring longer travel times. This study offers valuable insights into the implementation of wireless charging infrastructure for CAEFVs, with the potential to enhance their driving range capabilities.

To explore the acquisition, analysis, and visualization applications of subway passenger flow data in the context of big data, and to uncover the patterns behind passenger travel data, this paper utilizes subway IC card swiping data as the data source to achieve the acquisition of subway network data, extraction of inbound and outbound passenger flows, modeling of subway rail network, and visualization of passenger flow big data. Through the analysis of subway passenger flow data, this paper reveals the urban transportation travel demand and residents' travel characteristics, analyzes the impact mechanism of land use types and land development intensity on passenger flows, and holds significant importance in improving the quality of urban public transportation services and optimizing subway operation strategies.

In order to improve the efficiency and safety of airport unmanned tower scheduling system, genetic algorithm is used to design and optimize the intelligent scheduling system, integrating advanced sensors and data processing technology to achieve efficient management of airport ground traffic. The results show that the system can effectively reduce the total scheduling time and conflict rate, improve the execution efficiency of scheduling strategies, and demonstrate excellent environmental adaptability and reliability.

In order to enhance the durability of the battery and optimize the efficiency of energy utilization, this study delves into the problem of energy loss in new energy vehicles during driving. The focus is on the development of energy recovery technology in the intelligent control of the vehicle. Relying on advanced internet technology, this study combines battery state-of-charge and vehicle speed data to deeply analyze the impact of these key factors on the effectiveness of intelligent energy recovery systems. In addition, this study delves into the mechanism of the interaction between the front and rear braking forces, as well as the interrelationship between braking intensity and braking torque. A computerized control system was designed based on the correlation between braking torques. After constructing the intelligent control model, experimental validation was carried out through the simulation environment. Experimental data show that the intelligent control system designed in this study significantly improves the efficiency of energy recovery during braking.

This paper mainly introduces the development of 5G technology in the current era, as well as the status, importance, and specific applications of 5G-based vehicle networking technology, digital twin technology, and other technologies in the field of road obstacle avoidance for autonomous vehicles. Based on these technologies, a road obstacle avoidance system is designed. The specific application of vehicle networking technology and digital twin technology in the obstacle avoidance system is visualized by constructing physical models, virtual models, and other methods, and the application scenarios, potential problems, and hidden dangers of the system are proposed. The results of this paper have certain application value for the Internet of Vehicles and autonomous driving technology.

Requirement analysis is the core step in requirements engineering. The traditional method of requirement analysis is usually carried out by customer, requirement analyst and other cooperative work, and repeated manual iteration. Based on the idea of model-based systems engineering (MBSE), taking the demand analysis of metro train braking system as an example, a top-down demand analysis method is proposed by using the forward design process, and a joint simulation platform is built by combining Unity3D and AMEsim. The key functional requirements and performance requirements are simulated and analyzed to verify the effectiveness and feasibility of the established requirement model, and then the requirement itemized model is managed and released to each subsystem. The results show that compared with the traditional design, the model established by this modeling method can improve the efficiency of system design and ensure the traceability of requirements. At the same time, the simulation analysis and verification of requirements improve the efficiency and accuracy of demand management.

With the rapid development of the aviation industry, flight delays have become a global issue, resulting in significant economic losses and passenger dissatisfaction. Accurately determining the causes of flight delays is crucial for airlines, air traffic management, and passengers. This study proposes an Adversarial Defense of the Transductive Support Vector Machine (ad-TSVM) algorithm to address the problem of flight delay causing determination. The ad-TSVM algorithm generates adversarial samples through adversarial training techniques and incorporates these samples into the Transductive Support Vector Machine (TSVM) to improve the model's classification accuracy and robustness. Compared to traditional methods, ad-TSVM can more precisely attribute the responsibility for flight delays to various stakeholders, such as airlines, air traffic control, weather, and air traffic flow, while resisting adversarial attacks. Experimental results demonstrate that the ad-TSVM algorithm significantly outperforms traditional TSVM and other standard machine-learning algorithms regarding classification accuracy on multiple real-world datasets. Additionally, the algorithm exhibits excellent performance in resisting adversarial attacks, effectively enhancing the model's robustness and reliability. This study addresses flight delay cause determination challenges, offering theoretical support for further research and practical applications in related fields.

To enhance the efficiency of vehicle-cargo matching on the platform, this study delves into the issue of price heterogeneity alongside previous research on matching rates, transportation costs, and more. Addressing this problem, this study constructs a multi-objective optimization model aimed at maximizes the overall satisfaction of vehicle owners, cargo owners, and platform profits. An improved dynamic Bayesian network (IDBN) combined with a fast non-dominated sorting genetic algorithm with an elite retention strategy (NSGA-II) is proposed for vehicle-cargo matching. Firstly, the fuzzy analytic hierarchy process is employed to calculate the vehicle-cargo matching degree hierarchically. Secondly, a dynamic Bayesian network structure is constructed, mapping vehicles, cargoes, and matching combinations to Bayesian network nodes, and solving the multi-objective model for more practical results. The algorithm is compared with the IDBN algorithm in the experimental process. Results reveal that the proposed algorithm enhances the average satisfaction by 90.87% and increases platform profits by an average of 235.60 yuan per time slice, demonstrating its feasibility on the platform.

In view of the high power, high heat production and high heat dissipation of the auxiliary system of hybrid tracked vehicles, the conventional energy management strategy does not consider the battery loss cost and heat dissipation of the auxiliary system. In this paper, an energy management control strategy based on particle swarm optimization is proposed by introducing battery loss cost and heat dissipation demand power. Firstly, the battery model and auxiliary system heat dissipation model are constructed, and the energy management strategy is formulated by particle swarm optimization algorithm. The influence of battery loss cost and auxiliary system heat dissipation on the energy management strategy of hybrid tracked vehicles is studied, and an energy management control strategy for mode switching according to driving requirements is established according to the optimization results of particle swarm optimization algorithm. The results show that the PSO optimization results considering battery loss and auxiliary system heat dissipation are more in line with the use scenario of hybrid tracked vehicles, and the energy consumption and economy increase by 5.5% compared with the PSO optimization results which do not consider both.

In order to study the method of formulating differentiated charging scheme for mountainous resource transport road sections, the road characteristics and vehicle characteristics of such road sections are analysed, and the form of differentiated charging is determined on this basis. Taking Baofu Expressway as an example, a correlation matrix is constructed based on the third-generation prospect theory using the values of four attributes of different roads, namely, passage time, transport cost, safety level, comfort level and driver's expectation, to analyse the change of prospect value and driver's decision-making and their route selection. The risk preference coefficient and loss aversion coefficient are set as variables to predict the prospect value changes of the relevant routes in the range of parameter fluctuations with a preferential interval of 5%. It is found that when the degree of rate preference is increased to 20%, drivers' route choice changes from free roads to high-speed road sections; in order to ensure the profitability of the enterprise, the 20% toll preference in a single direction for six types of trucks is the best plan for the differentiated tolls of the Baofu Expressway; and the conclusions of the study provide a reference for the formulation of differentiated tolls for mountainous resource transport road sections.

Addressing the obstacle avoidance path planning problem for intelligent vehicles, this paper proposes a dynamic path planning method based on multi-factor cost pruning optimization. This method enhances planning efficiency while ensuring safety by pruning decisions using a multi-factor cost function combined with dynamic programming principles. Specifically, it selects key obstacles based on the dynamic window division according to vehicle speed and integrates factors such as obstacle cost and road boundary cost to calculate decision priority. Additionally, an evaluation function is designed to assess the environment's complexity by considering the proportion of obstacles in the dynamic window relative to the window area, distribution, dynamic characteristics, road conditions, and other factors. The pruning proportion is dynamically adjusted based on environmental complexity to maintain safety and maximize planning efficiency. Experimental results indicate that this method can promptly respond to environmental changes, optimize planning efficiency, ensure path safety and rationality, and demonstrate high practical applicability and effectiveness. Through these innovative methods and experimental results, this paper provides significant technical support and theoretical guidance for the field of intelligent vehicle path planning, highlighting its broad application prospects and scientific research value.

To improve the overall safety and comfort level of highway alignment geometric design scheme, driving simulation experiment is designed and implemented. Firstly, a subjective quantitative evaluation analysis is performed on the amenity, consistency, and comfort of the alignment design. The aggregation area and distribution state of driving speed are analyzed. The consistency of the operating speed consistency of each alignment unit is analyzed on the basis of the unit division of horizontal (H) and vertical (V) alignment. Considering the sensitivity of x-lateral acceleration (Ax) outliers in the driving process, the variation degree of Ax data for each alignment unit is analyzed based on the robust coefficient of variation (RCV), and the units with poor comfort of alignment geometric design are identified. The key range and sequence of optimization adjustment of alignment geometric design elements of alignment are presented by the comprehensive analysis of the combined unit of horizontal and vertical (HV) curves. The results showed that the subjects had higher scores on the amenity and consistency of the alignment design, but the driving comfort was relatively poor affected by the terrain; H2 speed fluctuation is more complex than H1, H1, V2, V3, and V4 units have poor speed coordination sections. Compared with coefficient of variation (CV), RCV is better at characterizing dispersion and stability of Ax data. Adjusting the design index within the alignment unit H2, V3, V4, V5 is the key to improving the overall driving comfort of the route. The RCV of the HV unit is smaller than that of the independent alignment unit. When optimizing the comfort of the design scheme, it is necessary to focus on adjusting the radius and slope of the HV1 unit, considering the combination design of the HV2 and HV3 units, and avoiding the unfavorable combination form containing multiple vertical curves in H2.

Green logistics present rapid development trend in China, this study combined with academic platforms and media platforms, using neural network word2vec word embedded algorithm to carry out comparative study of green logistics, choose China network and China logistics and purchasing network as a research object, to collect "green logistics" corpus, build academic platforms and media platforms two corpus and word2vec modeling. On this basis, compare the top 20 words with the top 20 words most similar to "green logistics" with the attention topics of the two platforms. Then, using the t-SNE algorithm, visually displaying and comparing through the matplotlib.pyplot package, and the results show that there are differences between green logistics in academic and media platforms. This study innovatively uses the word2vec algorithm to compare the differences between academic platforms and media platforms in green logistics from the semantic level, thus providing a new perspective for green logistics research.

With the rapid development of economy of globalization and e-commerce, logistics industry has become an indispensable key link in modern economic system. Logistics demand forecasting is a key link in logistics management, which plays an important role in improving logistics efficiency, reducing operating costs and optimizing resource allocation. This paper aims to discuss in detail the current situation and challenges in the field of logistics demand forecasting, select 7 influencing factors of economic indicators in Shandong Province from 2017 to 2022 as input indicators, and cargo transportation volume as the output indicators of logistics demand, and on the basis of gray correlation analysis, select several influencing factors with high correlation degree with feature series as prediction indicators, construct GM(1,7) model, use MATLAB to solve the model, and calculate the parameters. The forecast value of logistics demand in Shandong Province from 2017 to 2023 was obtained, and validity of model was tested, and the prediction of this model has high reliability and accuracy, 00with a reasonable relative error. In addition, this paper combines the present situation of logistics development in Shandong Province. Based on this, three suggestions are put forward: increasing infrastructure investment, optimizing agricultural and manufacturing logistics, and improving urban distribution network.

Transportation is an important foundation for supporting modern logistics development, and it is also an important link for logistics cost reduction and efficiency improvement. Based on the analysis of the current characteristics of logistics development in Shandong Province, this paper analyzes the mechanism of transportation promoting logistics cost reduction and efficiency improvement, systematically summarizes the achievements and problems of Shandong Province's work in promoting logistics cost reduction and efficiency improvement through transportation, and puts forward the path and specific measures for promoting logistics cost reduction and efficiency improvement through transportation from the perspectives of optimizing the network layout of transportation hubs, improving the transportation service system, strengthening the efficient linkage of transportation, and optimizing the business environment for transportation and logistics.

A path planning method combining the artificial potential field method (APF) with numerical optimization is proposed for the driverless vehicle driving problem of obstacle avoidance for moving and static obstacles in complex traffic scenes. First, the road safety potential field is constructed, and the dynamic and static potential field functions of the influence range of the ovoid potential field are established based on the braking safety distance. Then, according to the motion characteristics of dynamic obstacles, the potential field function about relative velocity and acceleration is established to improve the dynamic potential field function, and the potential field rotation coefficients are designed, which more accurately describes the dangerous trend of the motion of dynamic obstacles. Finally, the improved APF method is combined in the numerical optimization, and the path smoothing adjustment term is added to generate the path curve. Co-simulation is carried out by the hardware-in-the-loop (HIL) simulation platform, and the results show that the proposed algorithm is able to successfully avoid obstacles in various complex traffic scenarios.

The ability of autonomous vehicles to avoid accidents in complex traffic environments is a focal point of public concern. Intel has proposed a mathematical model called Responsibility-Sensitive Safety (RSS) to ensure that autonomous vehicles maintain a safe distance from surrounding vehicles. However, testing of this model has been limited so far. This study first considers the characteristics of the RSS model and designs test scenarios accordingly. Next, the collision avoidance process in the test scenarios is analyzed, and a collision avoidance control strategy is developed based on the scenarios. Finally, a test environment is built using the Apollo simulation platform, and test plans are designed with different vehicle speeds. The performance of the RSS model is then compared with three typical control models. Simulation results show that under high-speed conditions, none of the four control models succeeded in avoiding collisions. In the front cut-in scenario at low and medium speeds, the RSS model maintained safety while also achieving higher speeds with minimal acceleration fluctuations, indicating relatively high stability. Among all tests, the TTC model showed the least variation in vehicle speed, demonstrating the highest stability.

Cold-chain logistics (CCL) is a vital pathway to guarantee the modern circulation and advance of fresh agricultural products. Reasonable prediction of the demand of cold-chain logistics of fresh agricultural products (CCLFAP) is a vital safeguard to ensure the stability of CCL. For purpose of predicting the the demand of CCLFAP more accurately, this paper proposed an Improved sparrow search algorithm (ISSA) to ameliorate BP neural network (BPNN) model. Sin chaotic mapping, dynamic adaptive weights and reverse learning strategies are introduced to improve the quality of the initial population of the sparrow search algorithm, update the discoverer and the location of the optimal solution, avert the algorithm sinking into the local optimal solution, and search the optimal solution more effectively in the global scope. Improved sparrow search algorithm was used to optimize the initial weight threshold of BPNN model to reduce the risk of overfitting. The experimental consequence indicate that the ISSA-BPNN model is superior to the other two models in the aspect of convergence rate, optimization ability and prediction accuracy, thus can provide reliable prediction value for the demand of CCLFAP.

In recent years, the emergence of cutting-edge technologies such as autonomous driving, artificial intelligence, and communication has catalyzed a surge in research activity within the field of urban air mobility. This burgeoning area of study is focused on exploring the potential impact of air vehicles and urban air traffic on revolutionizing air transportation and shaping the future landscape of transport systems. The present paper undertakes an in-depth analysis of the current state of air vehicles and urban air traffic technology, offering a comprehensive summary of existing policy standards while delving into discussions regarding the critical importance of standardization research. Furthermore, it provides detailed recommendations for advancing standard development alongside pertinent research plans to address this evolving domain effectively.

With the continuous improvement of intelligence and driving safety requirements, accurately sensing the external environment and planning paths reasonably and efficiently has become one of the research hotspots in intelligent vehicles. Based on the study of intelligent vehicle architecture and predictive control schemes, this article focuses on the research of trajectory planning algorithms. This article proposes an improved RRT* algorithm based on the principle of fast random tree algorithm (RRT). This algorithm introduces a bidirectional random tree search strategy to shorten the convergence time and improve computational efficiency; by using redundant point clipping and B-spline curve fitting methods, the algorithm generates relatively smooth paths. Build a narrow channel environment and a multi obstacle environment in MATLAB software environment for simulation verification. The simulation results show that compared to the traditional RRT* algorithm, the improved RRT* algorithm outperforms the traditional RRT* algorithm in terms of planning speed, path length, and number of sampling nodes.

Under emergencies, reasonably guiding passenger flow can improve the transportation efficiency and minimize the impact of the accident on urban traffic operation. First of all, based on the psychological theory and the relevant information such as the choice preference of passengers, the passengers are classified to get the sensitivity difference of different potential mode passengers on the scheme attributes, which provides data support for path selection modeling and optimization. Secondly, the model with time index is established by using the data of the train and the passengers, and the model with comfort index is constructed by using the cubic polynomial fitting method. According to the utility maximization problem theory, the objective function including time and comfort index is constructed to provide the optimal path for different types of passengers. Finally, a case study of partial rail transit lines in Beijing is presented. The results show that, in the case of emergency, by analyzing the behavior preference of passenger's path selection, the passengers can be classified, and a more suitable route can be provided for each type of passengers, which can improve the time utilization ratio, and passengers' perception of the train's comfort.

To address the challenge of setting traffic signs on parallel divided road sections with ten or more lanes, and to enhance the safety of driving on super-multi-lane parallel divided road sections, an example is taken from the Guangao Expressway South China Sea to Zhuhai section reconstruction and expansion project. A signage layout scheme suitable for parallel divided road sections is designed, and a ten-lane road section model is constructed. Simulated driving experiments are conducted to validate and analyze the design scheme. Comprehensive analysis and evaluation are performed on vehicle speed, lateral deviation, and driver physiological characteristics. The results indicate that the addition of advance signs for parallel divided sections can effectively reduce driver reaction time, enabling smooth reception of information about parallel divided sections, leading to relatively stable vehicle operation and enhancing driving safety. However, through simulated driving experiments, it is found that the designed advance signs for parallel divided sections are prone to misleading drivers and carry a high information load, imposing a certain cognitive burden on drivers. Therefore, optimization is conducted based on the original design scheme, simplifying sign information, ensuring a high degree of alignment with the actual road alignment, and improving driving safety on parallel divided sections for drivers.

Port collection and distribution railways construction has significantly promoted the transport structure adjustment in China. The transport volume maintained a high growth rate in container rail-water intermodal transportation with the continuously policy support. However, the increase of the growth rate gradually slows down. The driving factors of the growth rate changes need a systematically understanding to better motivate the development of rail-water intermodal transportation. We explored the characteristics of rail-water intermodal transport using the infrastructure and transport data from 63 major ports and 110 important port areas. The result show that the development of rail-water intermodal transport is inhibited by factors such as geographic distribution, cost strategy and the competition with river water transport. Further, the rail-water intermodal transport volume limited by the access rate and capacity of port railway infrastructure. Therefore, both manage strategy and construction action are required to promote high-quality development of rail-water intermodal transportation, thereby enhancing logistics efficiency and reducing costs.

In community group buying, small batch delivery presents new challenges. In order to address the intricate logistics environment of multi-yard and multi-car coordination, we are engaged in a study aimed at optimizing the simultaneous pickup strategy. The objective is to accelerate the cargo collection process through the implementation of efficient scheduling and loading procedures, thereby enhancing logistics efficiency and flexibility. Additionally, the study is designed to provide practical and feasible solutions for the logistics industry and delivery problem with time windows and divisible demands (MDMCSPDTWSD), an extension of MCVRP, MDVRPTW, and SPD. Given its complexity, in this study, we propose a novel two-stage enhanced variable neighborhood search algorithm, which effectively improves the efficiency and quality of complex problem-solving through innovative strategy combinations. In the initial stage, heuristic techniques are employed to rapidly construct a robust foundation framework for the initial solution, facilitating a comprehensive understanding of the problem's characteristics. This step not only markedly reduces the running time of the algorithm, but also ensures that the initial solution has a high starting point in quality and provides a robust foundation for subsequent optimization. Subsequently, in the second stage, we introduce refined neighborhood search technology, which enables an in-depth exploration of the solution space. This allows for comprehensive and multi-level optimization of the initial solution, thereby enhancing the algorithm's global search ability and significantly improving the quality of the initial solution, which is brought closer to the optimal solution and offers an effective means of addressing complex optimization problems. To reduce local search time, we parallelize the operators in the search.

With the acceleration of social life, people's shopping habits have gradually changed, and the demand for online purchase of fresh products has increased rapidly. Fresh-food e-commerce gradually changes the pattern of traditional distribution mode of fresh-food products and adopts the multi-channel network distribution mode such as B2B, which reduces the intermediate links, but also causes the logistics network to become more complex. In the circulation of fresh products, the problems of poor preservation effect and high distribution cost have plagued the development of fresh cold chain logistics industry. In today's cold chain logistics industry, where the Internet of Things and electronic information technology have not yet been fully popularized, optimizing the vehicle distribution path of fresh agricultural product cold chain logistics still has high research significance. So how to optimize the delivery route of refrigerated trucks and reduce delivery and transportation costs remains a key issue in cold chain logistics optimization research. Therefore, the main challenge facing enterprises is how to reduce transportation costs, optimize the cold chain logistics distribution path and improve customer satisfaction. According to the analysis of concrete examples, this paper constructs the distribution path optimization model, selects the genetic algorithm, uses MATLAB to solve the model, obtains the optimal distribution path scheme.

Emergency medical supplies have challenges related to promptness and expenses associated with transportation during natural catastrophes. Utilizing drone delivery is a key approach to address the complete logistical problem. Thus, to enhance the efficient spread of emergency medication, this study suggests a cooperative distribution approach called "vehicle-drone." The model makes use of a different two-level distribution model: materials are delivered by drone from the regional disaster centers to the related disaster points at the second level, and materials are delivered by vehicles from the first-level distribution center to the regional disaster centers and the edge areas at the first level. The OPTICS-FCM algorithm groups the regional catastrophe centers together and assigns them as secondary distribution centers. The classic particle swarm technique is used to optimize the course of the vehicle, but the Adaptive Elite Genetic technique (AEGA) is employed to optimize the path of the drone, considering time window restrictions and limitations in the drone's range. Given the simulation results, the "vehicle-drone" cooperative distribution model may effectively address the challenges associated with medical emergency distributions. It achieved this by reducing both the time and cost involved compared to the traditional single-vehicle distribution approach.

Langfang International Commercial Trade Logistics Port, a key part of the Daxing International Airport economic zone, plays a vital role in supporting airport airside resources. This study aims to promote the coordinated development of the Beijing-Tianjin-Hebei region by integrating regional and international resources and driving innovation in the logistics industry, positioning the logistics port as a new engine for regional growth. The research clarifies the core concept, characteristics, and planning theory of airport logistics ports. Key factors such as policy orientation, geographical location, industrial structure, and transport networks are analyzed through detailed investigation. A high-precision logistics flow prediction model, using methods like the entropy weight method, gravity model, and GM (1,1)-Markov model, provides accurate forecasts for 2025 and 2030. Core functional areas including cold chain logistics, cross-border logistics, high-tech equipment logistics, and express logistics are identified, and their areas are estimated based on predicted flow rates. The relationships between these functional areas are also analyzed. By integrating the improved Systematic Layout Planning (SLP) method with the ant colony optimization, and considering actual land use for the logistics port, a precise layout plan for Langfang International Commercial Trade Logistics Port is developed and evaluated.

Lane line detection plays a crucial role in enabling self-driving capabilities and represents a challenging and far reaching subject. Despite the existence of numerous detection methods, many algorithms still require improvements to effectively handle dim ambient light and complex environmental conditions. Addressing these challenges, this paper presents a scheme that tackles multiple tasks in autonomous driving, such as lane detection and ground marker recognition, by performing various forms of data augmentation and utilizing the YOLOv5 model. The experimental results demonstrate the effectiveness of the proposed scheme under complex lighting conditions.

Evaluation of traffic load is an important aspect of road infrastructure analysis and planning in every country. It also has significance in the context of road safety, societal mobility design, and economic development. Therefore, the growth of global road traffic poses challenges in many areas, necessitating continuous analysis, adaptation of road infrastructure, countermeasures against air pollution, increased emissions, and congestion, which is the origin of this article. Thus, the aim of this article is to present issues related to traffic measurement using the example of one European country, namely Poland, with a particular focus on the measurement methods used and their evaluation. The research problem adopted is the question of whether the current measurements are sufficient and whether the results obtained by different methods are comparable. Additionally, the paper includes an analysis and evaluation of the possibilities of using such results in traffic management processes and proposes recommendations and guidelines for further necessary actions and measures to improve road conditions.

Carbon emissions are a global issue that restricts social development, and the carbon emissions level of the power industry directly affects the achievement of international commitments to carbon reduction. Reducing the energy consumption of rail transit can help optimize the energy structure, reduce greenhouse gas emissions, and contribute to achieving China's dual carbon goals. Therefore, this article proposes a method for monitoring the energy consumption of traction transformers in rail transit under the dual carbon target. A monitoring and control scheme based on microcontroller was studied by analyzing the losses of traction transformers in rail transit. This scheme calculates the load of the transformer through a multifunctional meter, and sets corresponding gear shifting thresholds to adjust the transformer capacity, in order to reduce losses and save energy. After experimental verification, this method has improved the energy-saving effect of traction transformers to a certain extent, while ensuring the stability and reliability of operation, providing useful reference for achieving carbon reduction goals in the field of rail transit.

To address the limitations of traditional whitelist-based obstacle detection systems in handling complex scenarios for autonomous driving, we propose a novel approach leveraging PETR (Position Embedding Transformation Representation). Our method predicts three-dimensional (3D) drivable areas by modeling occupancy information, employing a new 3D occupancy query mechanism that integrates 2D image features through an attention decoder. Extensive ablation studies identify optimal backbone networks and query sizes, demonstrating the method's ability to accurately model driving scenarios and delineate drivable areas. This innovation significantly enhances general obstacle detection in autonomous driving.

As urban landscapes and transportation infrastructures evolve, the precision of traffic flow prediction within Intelligent Transportation Systems (ITS) becomes paramount. It is vital for mitigating traffic congestion and for refining autonomous driving and route planning. Yet, traditional methods struggle with the intricacies and vastness of modern data. This paper surveys traditional approaches before exploring deep learning techniques that adeptly model the nonlinear, spatio-temporal complexities of traffic flows. It examines how deep learning harnesses spatial and temporal data correlations, offering a robust framework for ITS. The discussion concludes with the challenges in traffic prediction and opportunities for future research, charting a course for innovative solutions.

To enable safe and efficient autonomous navigation of unmanned ships in complex marine environments, particularly when encountering overtaking, head-on, or crossing scenarios, it is crucial to accurately identify and detect unknown unmanned vessels. This paper proposes a lightweight object detection algorithm tailored to these challenges. The lightweight network Ghostnet is used to improve the backbone network of the YOLOv5s algorithm, achieving lightweight target detection. By introducing the ECA attention mechanism, the model can better focus on important information in the image, enhancing the ability of feature representation, and improving the model's capability for complex scene understanding and target recognition and localization. Finally, the MPDIoU is used as the bounding box loss function for the YOLOv5s algorithm, which not only improves convergence speed but also obtains more accurate regression results. Comparative experiments show that the proposed improved algorithm outperforms models such as YOLOv5-BiFPN, YOLOv5-CBAM, and YOLOv5-SE in terms of model size, computational complexity, parameter count, detection speed, and average precision.

Traffic flow prediction plays a vital role in modern intelligent transportation systems. In fact, predicting the future state of traffic is a difficult task because of the complexity of spatial relationships and temporal dependencies. In recent years, graph convolutional networks have been widely used in traffic flow prediction tasks and have shown excellent prediction performance. Therefore, prediction methods based on graph convolutional networks and their variants are key research directions in the field of traffic prediction. In this paper, we firstly review the development of graph convolutional networks in recent years, and secondly categorize graph convolutional networks and their variants into two main groups, and subdivided into nine types on this basis. In addition, representative graph convolutional networks and their variant models were selected for in-depth analysis and discussion. Finally, based on the current development status, the future research directions of traffic flow prediction technology based on graph convolutional networks are openly discussed.

Autonomous driving test scenarios are the basis of autonomous driving virtual simulation tests. In this study, we take the right lane change cut-in scenario of adjacent lanes on highways in the highD dataset as an example, and based on the pure transformer architecture, we propose a method that combines CNN-GRU to capture the time series features and introduces MH-ESA to increase the smoothing, which is called LC-GAN, for generating the autonomous vehicles cut-in scenarios while driving on highways. Experiments show that LC-GAN is more advantageous in terms of data fidelity and usability compared to TimeGAN and COSCI-GAN, in addition to similarity improvements of 19.02% and 20.34%, respectively. As the number of generated scenarios increased from 1540 to 20,000, the coverage rate from 87.31% to 93.90%, it shows that the proposed method can effectively generalize the generation of cut-in scenarios for autonomous driving test, which lays the foundation for generating test scenarios and advancing autonomous driving test.

The management and control of the intersection at the junction of cities and towns is more difficult because of its complex flow of people and vehicles. Through the effective analysis of intersection traffic and the continuous optimization of signal timing, it is an important means to ensure the orderly passage of suburban intersections. Taking the typical Zijin intersection in the south of Hefei City as an example, the video surveillance system of the intersection carried out traffic investigation and data analysis on vehicle speed, vehicle density, delay, etc., designed a more reasonable signal timing scheme for the intersection, and planned to effectively improve the traffic efficiency of the intersection during peak hours. This research method and design scheme can promote similar urban intersection traffic research, and has certain reference significance for alleviating urban road traffic congestion.

The rapid advancement of deep learning and artificial intelligence has significantly improved autonomous driving systems, with traffic sign detection being crucial for their safety and efficiency. Challenges such as varying lighting conditions, occlusions, and diverse sign states complicate this task. Traditional methods relying on manually designed feature extractors often fall short in adaptability and generalization. To address these issues, we propose a novel approach using ResNet and transfer learning. ResNet mitigates the vanishing gradient problem through residual connections, while transfer learning leverages pre-trained models to enhance performance with limited data. Our model, based on ResNet18 and augmented with data enhancement techniques, was trained on the “Road Sign Detection” dataset from Kaggle. The results show a significant reduction in training loss and a validation accuracy of 85.8%, demonstrating the robustness and effectiveness of model in complex traffic sign detection scenarios.

This study focuses on the importance of spatiotemporal correlation when predicting the traffic status of urban expressways in the short term. Firstly, by analyzing the data of detectors in the upstream and downstream of the expressway, we can understand the changes in traffic flow and speed, and use an ordered sample optimal segmentation algorithm to divide the day into different time periods in order to identify the time periods when the traffic status is relatively stable. Subsequently, a spatial vector autoregressive model considering the impact of upstream and downstream road segments was established to predict traffic flow and speed at target locations at different time periods. The research results indicate that this is particularly true during peak hours, where the traffic status of downstream sections has a significant impact on the upstream. During peak hours, the traffic status of the target location is mainly determined by the upstream section, while the impact of the downstream section is relatively small. Therefore, in response to this situation, sufficient prediction accuracy can be achieved by solely utilizing the traffic status of the upstream section. In addition, compared with traditional ARIMA prediction models and historical mean prediction models, the VAR model proposed in this study can more comprehensively consider the spatiotemporal dynamic characteristics of traffic status, thereby more accurately predicting the traffic status of urban expressways.

Despite numerous actions taken by port authorities to enable modal shift, road transport has been playing a dominating role in serving the hinterland of many ports. To address the problem that it is difficult to grasp the operating characteristics of lorries on port roads. This study establishes a port truck travel chain identification model based on truck trajectory data and port spatial distribution. By screening the truck travel chain associated with the port, we analyse the spatio-temporal operation characteristics of road trucks and mine the distribution of port economic hinterland. The Dongjiakou port area of Qingdao port in China is used to illustrate and analyse. This study aims to help port enterprises understand the current situation of port road transport operations, which is of great significance for port strategic planning and market decision-making.

How to accurately acquire and track the positions of multiple vehicles in a traffic system is a challenging problem. Especially in the clustering detection problem, the uncertainty of vehicle states (including number, spatial location, etc.) in traffic scenarios leads to the fact that the predefined single clustering condition cannot match the variable clustering scenarios, which causes serious distortion of clustering results. As a result, this paper proposes a scene-adaptive multi-target clustering algorithm for millimetre-wave radar point cloud data, where spatial dimension is proposed to be clustered with density eigenvalues, and speed dimension fast clustering is carried out in a single frame based on sorted difference, and then fusion correction is made to the results of the two clusters by using the simulated spindle to solve the difficulties in obtaining the multi-target adaptive clustering conditions of the road as well as the problem of over-segmentation of the clusters of some large targets. The algorithm firstly, calculates the adaptive threshold of the original radar data set. Then, a density-based spatial clustering algorithm (DBSCAN) is used to perform primary clustering and outlier removal on the radar data, and secondary clustering in the velocity dimension. Finally, the spindle fusion clustering algorithm is applied to fuse the two clustering results. The proposed algorithm is experimentally validated using data collected from a 77GHz TI AWR1843 radar platform. The experimental results show that the improved algorithm improves the clustering accuracy by about 17.48% compared to the conventional DBSCAN algorithm in traffic multi-target scenarios.

Accurately assessing air traffic complexity is crucial for ensuring flight safety and efficiency. In this paper, we propose an air traffic complexity assessment method based on multi-scale spatio-temporal images. We first convert numerical air traffic data collected by surveillance radar into image data, and then interpolate the spatial and temporal scales of the images to generate multiple sets of images with different spatial and temporal scales. We found the best image set based on the Circle metric learning model by evaluating the performance of 20 image sets combined from different spatio-temporal scales. On this image set, we applied multiple deep metric learning methods to assess air traffic complexity. The results show that the new dataset performs better on several assessment metrics, confirming the validity of the approach to improve air traffic complexity assessment by adjusting the data scale. This demonstrates that the accuracy of air traffic complexity assessment can be effectively improved by adjusting the data scale. It also demonstrates the importance of choosing the right scale of dataset to improve the performance of machine learning algorithms.

The integrated driver assistance system is an important part of intelligent and connected vehicle. In order to do the research of the car-following ability and emergency avoidance ability of the integrated driver assistance system, this paper selects two scenes of cut-in and cut-out. In the meanwhile, this paper selects the test data of three representative vehicles respectively, analyzes the test results, and puts forward improvement suggestions.

With the rapid development of civil aviation, flight delays have become increasingly critical. Effectively predicting flight delays has become a research hotspot in recent years. In this paper, we propose a novel flight delay prediction method based on image recognition by applying deep learning techniques. We first rasterize the target terminal area into a grid image of size 25×25. Then, we convert the arrival and departure traffic flow data collected over some time into colors and fill them into the corresponding grids to obtain the arrival and departure traffic flow images. On this basis, we design the prediction model Attention-Complementary Hybrid Network (ACHNet), which can be used for multi-channel image feature fusion. It takes arrival and departure traffic flow images as inputs and fuses the spatial information of these two types of images through the fusion branch. We use the Convolutional Neural Network (CNN)-based prediction model to predict flight delays on arrival and departure traffic flow image sets respectively, and compare the prediction results with those of ACHNet. The experimental results show that the fusion of multi-channel image features can effectively improve the performance of flight delay prediction.

This paper proposes an automatic road marking extraction method based on vehicle-mounted laser scanning data. First, the grid elevation difference threshold method is used to segment the point cloud data into ground points and non-ground points. Then, the point cloud data is divided into blocks along the scanning trajectory, generating georeferenced images of elevation, intensity, and point density. Through a series of image processing techniques, including median filtering, Sobel operator, and maximum entropy thresholding, road marking pixels are preliminarily extracted from these images. Subsequently, the extracted pixel information is converted back into point cloud data. The point cloud is segmented into independent objects through connected component analysis, and further refined using the Otsu thresholding method and Statistical Outlier Removal (SOR). This method combines 2D image processing and 3D point cloud analysis techniques, effectively extracting and optimizing road marking information. Experimental results show that this method achieves an average completeness of 93% and an average correctness of 91% on two sets of test data, demonstrating its efficiency and accuracy in extracting road markings. This provides support for high-precision map creation and autonomous driving technology.

Accurate bike-sharing demand prediction helps relevant authorities make informed decisions on bike placement and advance scheduling. Although a great deal of work has been carried out in this field, we found that most studies have focused on modeling at a single granularity, ignoring the interactions between multiple granularities. To address this issue, we propose a Multi-level Model for Bike-sharing Demand Prediction (MMBDP). This model starts with an embedding module that determines the macro- and micro-granularity representations of bike-sharing demand from the initial study area. Then, macro- and micro-modules are built to perform spatiotemporal modeling to extract the spatiotemporal correlation information at different granularities. Finally, the fusion module is designed to integrate the information from each granularity and perform the final prediction. Experiments on real datasets show that the proposed model outperforms the baselines. Ablation experiments further reveal the superiority of our model. The model helps to address the bike supply-demand imbalance problem and promotes the sustainable development of urban transportation systems.

The road driving environment is an important factor affecting the severity of traffic accidents. In this paper, the traffic accident at traffic signs under the condition of a rainstorm is taken as the research object to explore the causes and influencing factors of the accident. Based on the traffic accident cases in the US Accidents (2016–2023) database, this paper analyzes the distribution characteristics of the accident data, uses SMOTE technology to form a balanced dataset, and then uses XGBoost machine learning method to build a traffic accident prediction model. Finally, combined with SHAP interpretable analysis, the relationship between various factors and the severity of traffic accidents is further revealed. The results show that traffic jam distance, climatic conditions (temperature, air pressure, humidity), and road conditions are the important factors that cause traffic accidents under these conditions. According to the research results, this paper puts forward some suggestions to reduce accidents from the perspective of traffic management, which provides a new perspective and information for accurate accident prevention and control.

With the rapid advancement of autonomous driving technology, lane line detection has become essential for vehicle safety and reliability. Traditional methods struggle with variations in lighting, road marking wear, and interfering objects, impacting robustness and real-time performance. To address this, we propose a lane line detection method based on YOLOv8, optimizing the YOLO series algorithms for better accuracy and efficiency. We constructed a dataset covering diverse road and weather conditions and used data augmentation techniques like blurring and adding noise to improve model robustness. During training, we refined YOLOv8’s feature extraction network and optimized detection heads for better lane line details and localization accuracy, while controlling computational complexity to maintain high frame rate detection. Post-processing steps, including geometric shape correction, dynamic confidence thresholds, and non-maximum suppression, were designed to reduce false positives and missed detections. Experimental results show our method achieves a mean average precision (mAP) consistently above 0.8, indicating high accuracy and stability in lane line detection. Future work will further optimize the model and explore additional data augmentation and processing techniques to enhance performance and applicability.

Traffic sign detection in foggy weather is a challenging task for unmanned driving systems. Fog introduces noise pollution to the images, which will affect the detection accuracy. In this paper, a traffic sign detection method based on asymptotic feature fusion and attention is proposed, to address the problem of large parameters and poor noise resistance of existing models in foggy weather. Firstly, a atrous selective kernel attention is proposed to improve the ability to extract target features by dynamically adjusting the receptive field range. Secondly, a lightweight asymptotic feature pyramid network is proposed to enhance the interaction of non-adjacent layer information in feature fusion. Thirdly, a novel lightweight convolution, GSConv, is integrated into the backbone network to reduce the model's parameters. The experimental results show that mAP50 achieves 98% on the CCTSDB 2021 dataset, respectively, which is 2.7% higher than the original model, and the parameters are decreased by about 40%. It shows that our method achieves detection results with fewer parameters and higher accuracy.

Predicting traffic patterns is crucial for smart transportation systems, as it aids in enhancing city traffic management levels, elevating citizens' travel experiences, supporting environmental conservation and promoting economic development. Due to the fact that the existing algorithm for graph convolutional model ignores the dynamic spatio-temporal changes in traffic flow data and does not extract sufficient spatio-temporal correlation features, we propose an alternative model that incorporates the dynamic spatio-temporal changes of traffic flow data, the attention mechanism is introduced to improve the GCN model, and the features of data about traffic flow are analyzed from the three dimensions of time, space, and spatio-temporal fusion. The model's performance is evaluated on two real datasets, PeMS04 and PeMS08, and the results show that it is superior to several types of baseline models based on various evaluation metrics.

Traffic flow prediction is a crucial component of intelligent transportation systems, providing scientific support for traffic management and urban planning. However, current studies are most focused on short-term prediction, with long-term prediction remaining a challenge. In addition, to control the dynamic spatio-temporal characteristics of urban traffic effectively is still difficult for traditional methods. To address these challenges, this paper proposes a K-shape-based spatiotemporal graph attention network model (KSTGAT) for long-term traffic flow prediction. The model employs a multi-head attention mechanism to extract long-term temporal features and replaces the traditional linear mappings in the attention mechanism with GRU to better capture nonlinear dynamic changes. For spatial feature extraction, K-shape is used to cluster the input sequences, and a composite adjacency matrix is constructed based on the clustering results and predefined adjacency matrix. This is followed by a graph attention network to extract dynamic spatial features. Additionally, the model also incorporates the periodic characteristics of traffic flow data. Experimental results demonstrate that KSTGAT outperforms state-of-the-art models, proving its effectiveness in long-term prediction.

The real-scene 3D model, established through oblique photogrammetry, is a surface model made up of a continuous irregular triangular mesh. While this model offers good visual effects, it lacks the ability for entity-attribute integration and database linkage, limiting its deeper application potential. To address this limitation, monomerization of feature entities is necessary. Traditional monomerization methods, such as physical, assignment, and vector monomerization, have disadvantages including the need for pre-processing, high workload, and poor presentability. This paper proposes a dynamic monolithization method based on projective geometry, using inverse ray transformation to directly extract contour point coordinates from the tilted model. This method enables real-time vector extraction, face overlay visualization, and pickup query through human-computer interaction. Using this method within a Cesium environment, the bottom edge polygon of the monolithic feature is drawn and then stretched to form the overlay model surface. This enables monolithic visualization and interactive retrieval in a 3D landscape. This vector dynamic monolithization method eliminates the need for physical segmentation and vector preprocessing of tilted model data. Compared to traditional vector monolithization, it avoids generating redundant data, circumvents the shortcomings of attribute and cutting monolithization, and enhances the spatial expression capabilities of real-scene 3D landscapes, thereby making urban space management more flexible and convenient.

In response to issues of low detection accuracy and poor robustness caused by missed detections of small objects and occlusions among dense objects in road scenes, a YOLOv8-RC object detection algorithm is proposed. First, by introducing receptive field attention convolution, the network can flexibly adjust the level of attention to targets of different scales, thereby enhancing the feature extraction capability of the backbone network and improving detection accuracy. Secondly, a lightweight CARAFA is used to replace the upsampling operator in the neck network, which reduces the number of parameters while improving detection accuracy. Finally, by adding a small object detection layer, local feature extraction is performed on larger-sized feature maps, thereby improving the accuracy of small object detection. Experimental results show that on a dataset excerpted from BDD100K, the average precision of the YOLOv8-RC algorithm model in this study increased by 5.9% compared to the model before improvement, with the model's parameter count reduced by 0.05M.

Aggressive driving is a traffic safety issue that people are concerned about. To explore the causes of aggressive driving, manual analysis methods are applied currently. However, this method may lack of objectiveness and accuracy. To address this, this paper will discusses how to identify and improve the aggressive driving behavior of taxi drivers by text features extraction (TF-IDF). It adopts a four-stage experimental process: firstly, classify and describe different road conditions, then conduct a seminar and analyze the content of the seminar in detail by displaying driving behavior cases. Thirdly, use text feature extraction technology to identify aggressive driving behavior patterns and induce preventive measures; Finally, report analysis results and give suggestions. In the process of text analysis, keywords are extracted by natural language processing technology, and lexical networks are constructed by using the co-occurrence frequency method. The results show that aggressive driving behavior is mainly concentrated in urban traffic scenes and key traffic nodes, and is closely related to drivers' subjective factors and external traffic environment. Therefore, the study puts forward a series of targeted suggestions, aiming at improving the level of road traffic safety by improving driver safety awareness and optimizing driving behavior. It demonstrates the application potential of text feature extraction in traffic behavior research and provides a scientific basis for traffic management and policy formulation.

To study the influence of the traffic environment on the near-collision behaviors of drivers of dangerous goods road transportation vehicles (dangerous goods vehicles) at the mechanism level, based on electroencephalogram (EEG) and simulated driving data, this paper investigated how the traffic environment affects the near-collision behaviors of drivers of dangerous goods vehicles using a deep learning model. We found that specific EEG features, especially in the parietal and central brain regions, have a significant impact on the prediction of near-collision events. The model achieved a prediction accuracy of 77.2% in predicting near-collisions and can better accurately determine the driving state by capturing the dynamic information of the driver's EEG and the traffic environment. The research results can provide a theoretical basis for warning of near-collisions in the road transportation of dangerous goods vehicles.

Lane change driving is a core behavior for adjusting speed and enhancing driving experience, in which drivers change lanes to find a more appropriate driving speed or a more comfortable driving environment. However, the complexity and variability of lane-changing operations, often accompanied by interactions with neighboring vehicles, exacerbate the dynamics and unpredictability of the driving environment. More importantly, inappropriate lane changing behavior has become one of the major factors in the frequency of traffic accidents and even fatal injuries, highlighting the major challenges to road traffic safety. Therefore, this paper proposes a lane-changing risk assessment method that considers the driving styles of surrounding vehicles. Firstly, the vehicle trajectory data provided in the highD dataset is selected as a sample to cluster analyze the driving styles of vehicles around the lane change. Then the lane changing risk evaluation index is established based on the headway time distance and stopping distance index, and the risk level is classified by fuzzy c-mean (FCM). Finally, the vehicle lane-changing risk is predicted using a long-short-term memory neural network (LSTM) with time series inputs, and the LSTM model considering driving style (DS-LSTM) is established. The experimental results show that the DS-LSTM model is more effective in predicting the risk of vehicle lane changing.

Aiming to address the challenge of traditional trajectory prediction methods in effectively capturing complex interaction behaviors between vehicles, a vehicle trajectory prediction model is proposed in this paper. The model integrates an LSTM and a safety interaction area to increase prediction accuracy and robustness in intricate traffic circumstances. By combining the time series processing capability of LSTM with the spatial constraint information of the safety interaction area, the model can accurately capture the dynamic behavior and interaction between vehicles. The fusion model performs better in the root RMSE, MAE, and R2 than the conventional LSTM model, according to experimental results, indicating that it has improved prediction accuracy and safety in the automated driving system. Future research will explore the possibility of incorporating more context-sensing information with complex interaction models to further enhance predictive performance.

To address the issues of low detection accuracy and poor real-time performance of autonomous driving target detection algorithms under adverse weather conditions, this paper proposes an improved vehicle target detection algorithm, MFW-YOLOv8, specifically designed for such environments. The C2f module in the backbone network is optimized by incorporating the MSBlock, which reduces computational complexity and achieves a lightweight model. The SPPF module is replaced with the Focal Modulation (FM) module to automatically focus on critical regions of the target, enhancing the accuracy and robustness of detecting hard-to-identify small objects. Furthermore, the WIOUV2 Loss function is employed to optimize the original network's loss function, thereby improving the model's generalization capability. Experimental results on the improved BW8K dataset demonstrate that the enhanced YOLOv8 algorithm achieves a 2.5% increase in mAP@0.5 compared to the original algorithm, with a 3.1% reduction in the number of parameters and a 6.8% reduction in computational complexity. Additionally, the mAP@0.5 under three adverse weather conditions (rain, snow, and fog) shows respective improvements of 1.3%, 1.7%, and 2.7%, demonstrating the effectiveness and advancement of the improved algorithm.

Based on vehicle-to-everything (V2X) edge networks, a traffic accident risk prediction method utilizing deep gradient boosting networks is proposed. Initially, massive real-time traffic data collected from the V2X edge network is preprocessed to form a new feature matrix, which is then input into a convolutional neural network (CNN) with relatively low structural complexity for automatic multi-dimensional feature extraction. Subsequently, pooling layers are used to reduce the number of feature parameters, and the output is converted into a suitably sized feature vector, which is then fed directly into an extreme gradient boosting (XGBoost) model for classification and prediction. Finally, based on the output indices from XGBoost, the severity of traffic accidents is simulated and predicted, with real-time alert information transmitted to vehicle units via edge servers. Simulation results demonstrate that this method outperforms traditional XGBoost, random forests (RF), and multilayer perceptron (MLP) algorithms in terms of performance.

Urban traffic management, a crucial component of modern urban development, plays an essential role in addressing urban traffic congestion and enhancing traffic efficiency. In this study, we explore the digital design of Urban Path Guidance Systems (UPGS) using artificial intelligence (AI) technology to meet the growing transportation demands during urbanization. Firstly, we outline the background of urban traffic data collection, processing, and application, emphasizing the importance of AI in traffic data analysis, and presents the main content and conclusions of the research. Secondly, we provide a detailed description of data preprocessing, feature extraction, and dimensionality reduction methods, as well as the construction and optimization of the Long Short-Term Memory (LSTM) network-based dynamic traffic flow prediction model. Additionally, we also discuss intelligent path planning and personalized recommendation methods, leveraging AI technology to provide users with the best travel routes based on real-time traffic conditions and user needs, thereby enhancing the travel experience. Finally, through a case study of Singapore's intelligent traffic management, we demonstrate the application and effectiveness of the proposed models and methods in actual traffic systems. This study not only provides new insights into urban traffic flow prediction but also contributes to the achievement of efficient and intelligent urban traffic management.

In densely populated regions, the identification of evacuating pedestrians remains a significant challenge, with profound implications for disaster management and public safety. Here, we present a novel approach to identifying evacuating pedestrians under such challenging conditions. Our method integrates attention mechanisms with segmentation techniques to accurately identify individuals amidst the dense smoke and haze of fires. Through extensive experimentation and validation, we demonstrate the robustness and effectiveness of our approach in real-world scenarios. This study not only advances the field of pedestrian recognition but also provides valuable insights for improving emergency response strategies in fire-affected areas. We propose a computer vision approach by augmenting target pedestrian recognition with a mask branch on top of object detection algorithms, utilizing channel attention mechanisms to enhance target pedestrians in the environment while reducing the weighting of the environment in images. We introduce a YOLOv8-Mask model which effectively enhances pedestrian recognition in hazy scenes typical of fire-smoke environments, facilitating efficient rescue operations for trapped individuals. Finally, comparative experiments conducted on self-collected datasets with five established object detection models demonstrate the superiority, robustness, and effectiveness of our proposed model in crowd recognition under such hazy conditions and real-world scenarios. The research outcomes presented herein contribute to assisting rescue efforts in densely populated venues amidst fire-smoke conditions, thereby providing technological support for the development of resilient cities. The model of this study has been made available for public access and can be found on this website https://github.com/xifangfff/YOLOv8-Mask/tree/main

In recent years, with the rapid development of China's economy and society, the number of car ownership has increased year by year, which has a negative impact on the development of urban slow traffic. This article focuses on the real needs of the urban transport system, with the goal of solving the problem of slow traffic optimization in the urban spatial scope, and constructs a set of evaluation system for the level of urban slow traffic supply oriented by demand analysis. From the level of region, street and node, this paper constructs an evaluation system by combining quantitative analysis with qualitative analysis, data investigation with empirical research, evaluates the supply level of slow traffic, and applies the model to the commercial blocks that are being renovated. The results show that the evaluation system has good practicability and effectiveness and can further enrich the theoretical research in the field of slow traffic planning in China.

In the context of the evolving domestic and international maritime environment, this study conducts a comprehensive analysis of the development status of Tianjin Port, Shanghai Port, and Shenzhen Port using the Analytic Hierarchy Process (AHP) and SWOT analysis. The research identifies five key dimensions for evaluating the three major ports: regional strategic position, industrial structure, cargo handling capacity, innovation in development models, and policy support intensity. Expert scoring questionnaires were used to rate these indicators, with importance scores ranging from one to nine. Quantitative results from the AHP model revealed that the location factor (A) had the highest weight at 40.12%, indicating its significant impact on port development, while innovation and development (D) had the lowest at 9.65%. Within the location factor, the regional strategic position (A2) was the most critical sub-indicator with a weight of 25.34%. Other crucial sub-indicators included the industrial structure of the three major sectors (B1) at 61.54%, cargo handling capacity (C1) at 58.33%, development model innovation (D1) at 54.55%, and policy support (E2) at 70%. The SWOT analysis leading to tailored development strategies. Tianjin Port was recommended to adopt a turnaround strategy, Shanghai Port a growth strategy, and Shenzhen Port a strategy leveraging its strategic location and technological advancements. This study offers strategic insights and recommendations to enhance the international competitiveness of these ports and contribute to the overall advancement of China's port industry.

6G vehicle-to-vehicle (V2X) communication and independent driving innovations are pivotal for independent driving. Be that as it may, the high-speed development of vehicles in 6G V2X frameworks postures a challenge in keeping up steady and dependable remote association quality and making strides channel pick up. Tending to this issue, this paper presents a novel multi-user planning calculation utilizing square diagonalization (BD) precoding for 6G ultra-high capacity multiple-input multiple-output (MIMO) frameworks. This calculation misuses BD precoding's affectability to channel relationship, utilizes the Pearson coefficient to degree client channel relationship post-matrix vectorization, characterizes a arranging coefficient based on the client commotion intensification calculate to evaluate channel quality, and considers the relationship affect between clients. It organizes client channels and channel quality to choose prevalent channels whereas diminishing channel relationship. Reenactment comes about show that the proposed calculation outperforms multi-user planning calculations based on subspace relationship, condition number, and geometric point by accomplishing more noteworthy client channel pick up and altogether bringing down the framework bit mistake rate, making it appropriate for 6G-V2X communication applications.

Understanding pedestrian behavior in urban environments is essential for advancing autonomous driving technologies and intelligent transportation systems. This study investigates the motion and trajectories of pedestrians across three distinct urban settings: free walk, shared space, and urban road. We analyze motion characteristics such as speed and acceleration, employing statistical methods, and examine pedestrian trajectories including average curvature, average speed, and proportion of stop points. Our findings reveal significant variations in pedestrian behavior across these environments. Urban road exhibits lower speeds, higher curvature, and frequent starts and stops, whereas shared space shows a lower proportion of stop points. These insights underscore the importance of adapting autonomous driving systems to diverse environmental contexts to enhance safety and efficiency.

With the acceleration of urbanization and the development of economy, China's urban population is growing rapidly, and various social problems have become increasingly prominent. The development and utilization of urban underground space is the inevitable result of the improvement of urban modernization level, and it is an indispensable part of urban sustainable development. In addition, the urban ground space and the sky are occupied by high-rise buildings and viaducts, resulting in the shortage of urban land, but also has a great impact on the environment. Therefore, the development and utilization of urban underground space has attracted more and more people's attention. The rational development and utilization of underground space will be an effective way for modern cities to achieve carbon neutral peak of carbon neutrality and solve the shortage of urban land area.

In order to promote the development of the appliance recycling industry, this paper proposes an evolutionary game model consisting of a manufacturer and a recycler, and obtains three sets of evolutionary stable strategies by setting the parameters to obtain the payoff matrix and calculating the replicated dynamic partitions of each game party. The model was validated using MATLAB R2022b, and numerical simulations of the main influencing factors were carried out sequentially to observe the effects on the strategies of the game parties when the parameters were changed. The results show that (1) the value of used appliances, the recycling capacity of the manufacturer, and the social benefits harvested by the manufacturer are positively related to the probability of proxy, and the proxy fee is negatively related to the probability of proxy. (2) The labor and publicity costs of the recycler's proxy recycling and the storage costs of proxy failure are negatively related to the proxy probability.

As traffic noise pollution intensifies due to the rising number of vehicles, there is an increasing need for effective noise reduction strategies. Traditional methods, such as installing sound barriers, creating buffer zones, and using soundproof materials on buildings, have been extensively studied. This study focused on evaluating the effectiveness of plants as sound barriers for traffic noise reduction. The research was conducted on a road segment near a residential area in Zhengzhou City, chosen for its suitability for the study. Measurements were taken 30 to 40 meters from the traffic noise source. Results indicated that dense plantings could achieve a modest noise reduction, ranging from 3 to 10dB(A). However, factors such as the distance of noise transmission and the volume of traffic were significant considerations. Consequently, the study concluded that while plants can contribute to noise reduction, their overall effect is limited.

The air traffic demand surpasses the capacity of most of the busiest airports worldwide. The mismatch between airport capacity and air traffic demand leads to serious congestion and delay problems. Since airport capacity is difficult to expand in a short time, in order to reduce flight delays, we need to adjust and optimize flight slots for flight applications, and allocate flight time resources scientifically and efficiently under the premise of meeting airspace resource constraints such as airport capacity. In this paper, we proposed an airport flight slots allocation method based on the Q-Learning algorithm in reinforcement learning. Firstly, we preprocess the original data set. Data cleaning processes data with missing attribute values and removes irrelevant attributes. Through data transformation, the data of string attribute is transformed into discrete numbers, and the time attribute is processed. After that, we construct a mathematical model of single-airport flight time allocation, which is a combinatorial optimization problem and determines the optimization objectives and constraints. Then we apply the reinforcement learning algorithm Q-Learning to solve this optimization problem, and the algorithm updates the strategy through interaction with the environment to get a better solution. Finally, we take Guangzhou Baiyun International Airport as a study case to train and verify the proposed single-airport flight time allocation model. The experimental results show that, the optimized flight slots satisfy the airport capacity constraints while minimizing the adjustment of flight schedules as much as possible, and adjust the flights at some times to alleviate the congestion.

As urbanization accelerates, the rapid increase in urban populations and vehicle numbers poses unprecedented challenges to city traffic. The demand for intelligent transportation systems, a key component of smart city construction, is growing day by day. This system is a highly integrated interdisciplinary field that combines complex computer algorithms, which to some extent limits its extensive application. The emergence of large language middleware has reduced the deployment barriers for related applications. This article proposes an innovative intelligent transportation system that combines a large language model (LLM) with Amap API through LangGraph, providing strong support for urban traffic. This system not only assists users in making travel decisions through natural language dialogue based on the superior reasoning and planning capabilities of LLM, but also enhances the semantic understanding ability of LLM by constructing a graph structure through LangGraph, it improves the capture capability for urban traffic tasks and implements a self-feedback mechanism for the large language model. This innovative approach offers a more convenient and efficient method for the application of large language models in the field of Intelligent Traffic System.

In response to the post-COVID-19 economic recession, City Y, as part of its smart city initiatives, issued electronic consumption vouchers (ECVs) to boost the market economy. However, some individuals exploited these vouchers, affecting their fairness and effectiveness. Thus, this paper presents an unsupervised deep learning framework to detect fraudulent merchants and anomalies in voucher redemptions. We use Long Short-Term Memory (LSTM) units combined with an autoencoder (AE) to learn long-term dependencies in time series data. A smoothing unit is also used to reduce prediction bias. The fraud scoring module assesses merchants based on the ratio of anomalous redemptions to total redemptions. Our experimental results show that this approach significantly improves fraud detection accuracy and efficiency, making it a promising tool for maintaining the fairness and effectiveness of ECVs within the smart city ecosystem.

High-definition maps (HD maps) are crucial for the development of autonomous driving and intelligent transportation technologies. However, the numerous and complex regulations related to HD maps present significant challenges for map production and review personnel when querying and understanding these regulations. Additionally, different interpretations of the same regulation by various individuals can affect the efficiency of data processing and application. To address these issues, this study is based on relevant Chinese standards and regulations, and establishes a Knowledge Graph for High-Definition Map Review Standards (KGMRS) through entity extraction and knowledge integration processes, incorporating expert experience. KGMRS is stored in a Neo4j graph database, enabling structured management and rapid querying of regulatory texts. Based on this research, a knowledge-guided, algorithm-based automatic map review application framework for HD maps is proposed, aiming to enhance the automation level of map data review. The results show that KGMRS significantly improves the query efficiency and utilization rate of high-definition navigation electronic map standards, markedly enhancing data standardization. The outcomes of this study provide strong support for the development of intelligent transportation systems.

This paper proposes a comprehensive method for evaluating the vulnerability of integrated airspace networks, considering both structural and functional vulnerabilities. First, a weighted network of air segments is constructed, where weights are assigned based on segment capacity, traffic flow, distance, or average transit time under varying ratios of manned to unmanned aircraft. A weighted network efficiency index is defined. Next, critical waypoints are identified by analyzing changes in the weighted network efficiency index under network disruptions and the operating cost index under flight scheduling adjustments. Finally, the network's vulnerability is assessed using the total delay cost as an indicator, based on deliberate attack scenarios. Using the integrated airspace flow network in East China as a case study, the analysis demonstrates that the weighted efficiency index effectively identifies functionally vulnerable nodes and can be used for efficient assessment of the network vulnerabilities.

This study constructs a fault tree with "fire during urban rail transit operations" as the top event, including nine intermediate events such as "unsafe human behavior," "electrical circuit failure," and "fire protection equipment failure," and 28 basic events such as "flammable materials and oxidizers," "carelessness with fire and electricity," and "insufficient firefighting skills." Based on the fault tree, a qualitative and quantitative fire risk analysis for urban rail transit operations was conducted. The analysis identified 176 possible fire occurrence pathways and quantified the influence of various factors on fire occurrence. Among these, "flammable materials and oxidizers" had the greatest impact on fire risk during urban rail transit operations, while "fire protection equipment failure" and "inadequate emergency response" were also significant factors. Based on the risk analysis results, targeted fire risk control measures for urban rail transit operations were proposed. These measures include strict control of flammable materials and oxidizers, improving the reliability of fire protection equipment, enhancing emergency response capabilities, increasing the reliability of electrical circuits and equipment, improving the operational reliability of rolling stock, and eliminating unsafe human behaviors. The research findings provide scientific evidence and technical support for the prevention and control of fire risks in urban rail transit.

Ethnic minority architecture in Guangxi serves as a vital medium for cultural heritage preservation. This paper analyzes the characteristics of decorative components, including woodcarving, stone carving, painting, and weaving, and summarizes the challenges they face in intelligent construction. It proposes a parametric design approach based on BIM technology, enabling rapid addition and flexible adjustment of decorative components through BIM model construction. The study emphasizes the importance of color and material handling in intelligent construction, facilitating vibrant facades and efficient production of decorative components. It offers new insights and methods for the preservation and development of traditional architecture, fostering the inheritance and development of ethnic minority cultures.

Ride-sharing can improve the utilization rate of space in the car and reduce the amount of road traffic, which is of great significance to alleviate congestion and carbon emissions. By introducing carbon cost, this paper analyzes the traffic assignment problem under the demand of private car travel and carpooling travel, defines the travel cost of different travelers (self-driving drivers, ride-sharing drivers, ride-sharing passengers), constructs a ride-sharing user equilibrium model based on low-carbon travel orientation, and proposes strategies to promote users to participate in carpooling. The effectiveness of the model and strategy is verified by an example. Based on the sensitivity analysis method, the influence of preferential strategies on carbon reduction effect and congestion degree is analyzed. The results show that the preferential strategy has a positive effect on promoting shared travel, reducing carbon emissions and alleviating traffic congestion, which provides a reference for traffic managers to formulate relevant policies.

The debugging of high-speed trains is a process to ensure the stable operation of trains, and traditional debugging equipment has some problems in terms of comprehensiveness, reliability, and incomplete testing. In order to solve these problems, this article designs an intelligent debugging platform. Not only can it solve the problem of requiring a large number of personnel to operate on site, but it can also improve transmission efficiency and have a wider range of applications.

Cross-domain collaborative monitoring is an important means to ensure flight test monitoring in a large airspace. Aiming at the cross-domain telemetry signal transmission and multi-network collaborative monitoring requirements brought about by cross-domain collaborative monitoring, this paper proposes a cross-domain collaborative monitoring technology based on multi-network integration. A cross-domain flexible networking architecture based on air-ground integration is designed to solve the problem of rapid return and flexible deployment of telemetry signals by combining "operator private network+satellite communication"; the collaborative monitoring of multiple monitoring locations in a wide area is realized by the joint monitoring mode of "flight test commercial network+monitoring local area network." The cross-domain collaborative monitoring system based on multi-network integration effectively ensures the stable transmission of telemetry signals and multi-network collaborative monitoring during cross-domain flight of the test aircraft, expands the capability baseline of flight test telemetry monitoring, and provides effective technical support for subsequent flight test telemetry in a large airspace.

The integration of smart grids in super high-rise buildings offers significant potential for optimizing renewable energy utilization. However, various barriers hinder successful implementation. This paper aims to identify and analyze these barriers through a structured questionnaire distributed to industry experts. The study presents the findings from the collected data, highlighting the major challenges and their respective weights in the context of smart grid integration in super high-rise buildings. The results aim to inform stakeholders and guide future efforts in overcoming these barriers to promote sustainable energy solutions.

This research explores the integration of Artificial Intelligence (AI) and the Internet of Things (IoT) to enhance energy efficiency and power quality management in prefabricated intelligent buildings. The study examines the potential benefits of leveraging advanced technologies in monitoring, controlling, and optimizing energy consumption and power quality. By implementing AI-driven algorithms and IoT-enabled devices, this research aims to provide a framework for sustainable building management, highlighting significant improvements in energy conservation, cost reduction, and overall building performance.

The braking pressure control device of this rail transit braking system sets the required pressure for common braking through the main regulating valve. When loaded, the auxiliary reducing valve outputs pressure to the main regulating valve through an electromagnetic valve as a carrier to supplement the pressure and increase the braking pressure; by controlling the connection and pressure holding function of the solenoid valve through the controller, the final output pressure of the main regulating valve can be infinitely changed. When the brake pressure needs to be quickly released, the maximum rapid exhaust can be achieved through the internal one-way valve air passage, avoiding internal throttling of the main regulating valve; the remote relief air control valve and brake restriction air control valve are equipped with piston type valves, which are safe and reliable to use, and have fast functional response; simplifying the internal structure of the brake pressure control device not only improves vehicle operation safety, but also facilitates user maintenance and repair of the vehicle; to meet various usage environments of vehicles, braking restrictions and remote release functions can be implemented.

Urban public transport is an important part of urban infrastructure and plays an important role in economic development, urban construction and social life. In order to improve the living environment of urban residents and promote the healthy development of public transportation, a five-dimension evaluation index system of safety, convenience, comfort, reliability and economy is constructed from the perspective of passenger research. AHP and fuzzy comprehensive evaluation methods are used to study the status quo of public transportation service quality in Hohhot urban area by using AHP-FCE model. The optimization of the bus service quality evaluation system can give priority to the development of bus service quality, improve the perceived service quality of passengers, and provide optimization direction and decision-making basis for planners and decision makers.

The rapid growth of international trade has promoted the development of container transportation, with container terminals playing a crucial role in the process. Multiple factors, including throughput, energy type, and the energy consumption of individual equipment, influence the energy consumption and CO₂ emissions at container terminals. It is essential to comprehensively assess these terminals' energy efficiency and CO₂ emission levels. This study integrates the influencing factors of energy efficiency and CO₂ emission intensity at container terminals, establishes evaluation indicators and applies the super-efficiency data envelopment analysis (DEA) method to develop a comprehensive evaluation model. The model is used to assess container terminals' energy efficiency and CO₂ emission intensity levels, and its effectiveness is validated through case studies. This study innovatively integrates the evaluation of energy efficiency and CO₂ emission levels, using objective data on energy consumption, throughput, and the energy consumption of operational equipment, thereby overcoming subjective influences. The proposed method provides a technical foundation for assessing the energy efficiency and CO₂ emission levels of container terminals.

Three dimensional mesh data is the main data structure for building real-life 3D models. To address the pressure of processing massive models and visualization rendering problems, it is necessary to optimize mesh simplification algorithms to support the construction of smart city information infrastructure. In this regard, based on the QEM algorithm, this article introduces a measure of vertex distribution uniformity in the model details, strengthens the focus of folding operations on removing irregular triangles, and designs a new error that considers the impact of vertex offset on local volume changes. The results indicate that the simplified model of the improved algorithm preserves details while ensuring mesh regularity. In order to improve the model simplification of the algorithm for subsequent texture mapping and rendering efficiency enhancement, and contribute to the improvement of visualization performance.

Visualization of urban traffic flow is beneficial to traffic management and urban planning. This paper proposes a method for visualizing urban traffic flow based on Unreal Engine, which realizes traffic flow simulation at three levels: macro, meso, and micro, shows realistic traffic flow dynamic effects, improves the loading efficiency of large-scale traffic flow, and carries out system design and testing. The results show that this method meets the multi-level display and rendering requirements of urban traffic data, and can provide support for urban traffic management and urban planning.

The growth of passenger flow in Qingdao Metro has led to the rise of customer service demand. The existing customer service system is constructed based on NLP and knowledge graph, which has the problems of inaccurate understanding of questions, low correlation of answers, and low passenger satisfaction. With the development of large language model technology, customized large language models combined with domain-specific knowledge are increasingly applied to practical application scenarios. After comparing the mainstream open-source large language models, this paper selects Qwen2-7B model as the base model, combines the special knowledge of Qingdao Metro customer service field, generates the Qingdao Metro special customer service model through pre-training and fine-tuning, and develops the Qingdao Metro intelligent customer service agent by means of retrieval-augmented generation and prompt engineering. The evaluation test shows that the score of the model reaches 87 points, which is greater than the 80 points of the Qwen2-7B general model and the 45 points of the existing customer service system. Compared with the original customer service system of Qingdao Metro, the answer accuracy of the model is increased by 38%, and the answer accuracy is increased by 16% compared with the general model. The proprietary large language model of metro intelligent customer service based on open source large language model can effectively solve passenger problems and improve passenger satisfaction.

In response to the complexity of fire safety in urban areas, this study proposes a theoretical method for the accurate diagnosis of urban fire risks. By reviewing the current state of research, the study presents theoretical solutions in the areas of risk factor identification, safety capability evaluation, and demand analysis. Taking into account the actual conditions, usage characteristics, operating environments, and management statuses of urban buildings, as well as the overall fire safety landscape, the study addresses the challenges and shortcomings in fire prevention and control within the context of urban fire risk management.

With the development of China's national economy, the demand for dry bulk cargo has continuously increased. Dry bulk transportation, which involves large volumes and long distances, is primarily handled by water transport, accounting for about 90% of the import and export of dry bulk cargo due to its energy efficiency and economic benefits. However, the energy consumption and CO₂ emissions of dry bulk terminals are influenced by various factors such as throughput, energy types, and the energy consumption of individual equipment. A comprehensive evaluation of energy efficiency and CO₂ emission levels is necessary. This study establishes evaluation indicators for energy efficiency and CO₂ emission intensity by considering the influencing factors of dry bulk terminal energy efficiency and CO₂ emissions. An evaluation model is developed using the Super-Efficiency data envelopment analysis (DEA) method. Through this model, the energy efficiency and CO₂ emission intensity levels of dry bulk terminals are determined and validated with case studies. This study introduces a novel integration of energy efficiency and CO₂ emission rating evaluations, using objective data on energy consumption and throughput, as well as the energy consumption of individual operational equipment, overcoming subjective influences and providing technical support for measuring the energy efficiency and CO₂ emission levels for dry bulk terminals.

With the continuous advancement of in-vehicle network technology, Controller Area Network (CAN) technology has become a key component of automotive control systems. As a standardized automotive diagnostic communication protocol, the Unified Diagnostic Services (UDS) protocol plays a crucial role in fault detection and vehicle maintenance. As the complexity of in-vehicle network systems increases, communication faults between Electronic Control Units occur frequently. In response to this issue, a diagnostic strategy based on the UDS protocol has been developed. This strategy can effectively differentiate communication interruptions caused by external interference from those caused by internal faults, enabling precise diagnosis of internal faults in in-vehicle network systems and significantly improving the accuracy of fault detection.

With the rapid advancement of computer virtual reality technology, the demand for simulation-based intelligent driving systems in the domains of roads, traffic, vehicles, and safety has become increasingly prominent. However, the closed nature of current technologies, high costs, limited application scenarios, and significant discrepancies between simulation results and real-world outcomes have constrained the progress and application of driving simulation technology. In response to these challenges, this paper presents the development of virtual reality intelligent driving simulation software using the open-source Unity3D physics engine. By integrating engineering design and 3D modeling, we have constructed an open-source driving simulation system that offers realistic simulation effects, flexible application scenarios, and the potential for further development.

With the rapid development of urban rail transit, there are efficiency issues in the assembly, commissioning, daily operation, and maintenance of urban rail air conditioning systems. In addition, there is a shortage of professional commissioning and maintenance personnel, which makes it difficult for existing railway colleges and manufacturers to meet the growing demand for training modes and equipment. Therefore, this article proposes and designs a comprehensive and practical urban rail air conditioning simulation training system. The system adopts a real air conditioning unit combined with a fault injection system, and uses Unity3D software for dynamic demonstration. Through 3D animation simulation of the working principle and operation process of the air conditioning system, it enhances the intuitiveness and practicality of the training. The system realizes the switching between centralized control mode and local control mode of train air conditioning, as well as fault simulation and diagnosis functions, comprehensively demonstrating the actual operating status of the air conditioning system. This simulation training platform not only effectively enhances students' practical operation and fault handling abilities, but also provides advanced technical support for improving the maintenance efficiency of rail transit air conditioning systems, which has important practical significance for teaching and practice in the rail transit industry.

This study aims to plan the urban development and land resource utilization, focusing on ensuring economic sustainability. The analysis encompasses several natural and social factors. Through geographic information system kernel density analysis and 1000m² grid sampling, data is integrated and applied to a multi-objective optimization model based on principal component analysis, genetic algorithm, and back propagation neural network. This model is composed of four sub-models, each addressing different dimensions of independent variables such as land cover type and urban area size, in relation to dependent variables such as gross product values of various industries, population density, and the remote sensing ecological index. The prediction results indicate that the northeastern part of Ningbo (Hangzhou Bay) and the central area (Zhoushan Port) have higher remote sensing ecological index ecological indices and larger industrial output value vectors, making them suitable for prioritized development of primary and secondary industries. In contrast, other regions, due to their relatively lower values, are more suitable for the development of tertiary industries.

This study addresses the challenge of fault diagnosis in photovoltaic inverter circuits by designing a method that combines Variational Mode Decomposition (VMD) with wavelet energy analysis. The method initially employs the VMD algorithm to perform multi-scale decomposition on signals such as current and voltage during the inverter's operation, revealing the intrinsic modal characteristics at different frequencies. Subsequently, by calculating the wavelet energy of each mode, it extracts the anomalous energy distribution characteristics of the inverter circuit under fault conditions. This method effectively isolates subtle changes related to faults from complex non-stationary signals and enhances the accuracy and real-time performance of fault diagnosis. It provides a powerful tool for the early detection and precise identification of inverter circuit faults during the operation and maintenance of photovoltaic power stations.

Digital twin cities are important carriers for realizing urban digital governance and developing the digital economy. In response to the unclear definition of the scope of urban transportation digital twins, the urgent need to deepen the architectural model, and the difficulty in integrating cross-domain heterogeneous information. Analyzes the technical needs of transportation digital twins, defines the concept of transportation scenarios, proposes a technical architecture for urban transportation digital twins, and studies key technologies such as digital identification and coding of transportation twins, unified coordinate systems, and construction of transportation scenario information models. It also conducts application demonstrations combined with urban expressway traffic control, promoting the implementation of digital twin cities in the transportation field and improving the level of intelligent transportation construction.

As the construction of smart cities deepens and local data platforms become increasingly sophisticated, a large amount of urban data is gathered and integrated, effectively promoting the joint construction and sharing of urban data. However, the issue of a lack of correlation between cross-domain data still exists, making it difficult to provide highly integrated and correlated data services for the construction of urban intelligence. Based on the multi-source spatiotemporal data after governance on the smart city integration platform, this paper proposes a spatial correlation method based on StreamSets task flows to achieve the correlation of urban multi-source data with spatial objects. The application practice results show that this method is conducive to providing city decision-makers with correlated data and information services for urban objects quickly.