Formulating policies tailored to urban low-carbon development phases and resident characteristics is essential for optimizing incentive structures and promoting green mobility. This study evaluates new energy vehicle (NEV) incentive strategies across four city categories, considering factors such as air quality, NEV penetration, and charging infrastructure maturity. It analyzes social media data using the Latent Dirichlet Allocation (LDA) model and designs user surveys. A Latent Class Ordered Logit Model (LCOL) is employed to assess different urban populations' preferences for vehicle electrification incentives, identifying key impacted groups. The results indicate that immediate incentives, such as driving ban exemptions and significant fiscal subsidies, effectively enhance the purchasing intent of NEVs among less receptive residents. Conversely, more receptive residents respond better to regular, smaller subsidies. Cities with low NEV penetration exhibit a higher probability of purchasing under incentives, highlighting greater potential for improvement. Enhancing charging infrastructure significantly boosts purchasing intentions in infrastructure-deficient cities, with a 1% increase in likelihood for every minute reduction in charging time. However, this effect diminishes in cities with extensive charging networks. In metropolises with vehicle access restrictions, exempting NEVs from these increases purchasing probabilities by 3.5%. These insights guide NEV promotional strategy development in diverse urban settings.

A high-quality grain distribution system is critical to ensure the balance of grain supply and demand and food security. This study considers the perishable nature of grain types and aims to minimize the total costs of transportation, carbon emissions, and loss. An integrated optimization model is proposed to consider different loading methods (bagged-bulk-container) and various transportation modes (road-rail-sea). A case study was performed using the heuristic genetic algorithm in the "grain transport from North to South China" scenario in Northeast China. The results indicate that compared to bagged grain and bulk grain transport, multimodal transport of grain containers by rail, road, and water has clear advantages in terms of lower total cost and reduced loss. The proportion of grain loss cost in container transport, bagged grain transport, and bulk grain transport is 9.86%, 42.29%, and 29.82%, respectively. In the "grain transport from North to South China" process, roads are primarily used for local collection and distribution, while railways and waterways handle long- distance trunk transportation. When the delivery time requirements increase, the proportion of railway transportation would gradually increase, and the proportion of waterway transportation would decrease. When the total delivery time reaches 71.5 hours, the optimal transportation scheme would shift from container multimodal transport via road, rail, and sea to container multimodal transport via road and rail only. In the composition of total costs, the transportation costs and carbon emission costs of the optimal routes for the three loading methods are essentially the same. The study result also serves as a reference for the government regulatory agencies and logistics service providers that reducing grain transportation losses is an effective way to lower the overall logistics transportation costs.

This paper proposes a parking reservation allocation optimization method based on the relationship between the direct revenue of the service platform and its service level in the parking allocation process, as well as the diversity of users' travel characteristics. To maximize the platform's operational service revenue, an optimized function is established with the operator's maximum revenue and the minimum comprehensive benefit of user travel costs as the objective. A periodic optimal parking reservation and allocation model (POPA) is developed in consideration of the time-effectiveness of parking allocation. An adaptive heating simulated annealing-particle swarm optimization algorithm is designed to solve large-scale parking allocation problems. The experimental results show that, considering the time-effectiveness and platform revenue of multiple factors, the optimal reservation period length for the reservation platform is 1 hour. The improved algorithm improves the solution effect by 6.14%. Sensitivity analysis proves that the introduction of punishment factors can improve the platform's user request acceptance rate by 2.25% to 18.17% without affecting the user's time cost and parking lot utilization rate. The proposed model has a 38.11% higher actual revenue than the user optimal model and a 15.31% lower average user travel cost than the platform optimal model. The expanded numerical test proves the applicability and effectiveness of the proposed model in large-scale complex scenarios.

China's high dependence on imported crude oil, coupled with the escalating risks of disruptions in the maritime transportation network, necessitates urgent attention to optimizing recovery strategies and enhancing resilience. Based on actual transportation data, a model of China's crude oil import maritime network has been developed. A resilience assessment method using resilience curves and network efficiency indicators, has been proposed. An optimization model for recovery strategies is established with the objective of maximizing resilience and then applied to five simulated disruption scenarios. The results reveal that the optimal recovery strategy significantly accelerates network recovery across all scenarios, reducing resilience loss by up to 79.14% compared to traditional strategies. The model identifies crucial nodes that significantly impact the network under various scenarios, emphasizing the importance of prioritized recovery to enhance overall efficiency. Furthermore, relaxing detour cost constraints decreases resilience loss and alters the optimal node recovery sequence. The findings provide a foundational basis for decision-making in emergency recovery and resilience enhancement of China's crude oil import maritime network.

Traffic state recognition research is of great significance for the prevention and mitigation of traffic congestion. It provides decision support for traffic management and also effectively enhances the operational efficiency of roads. Traditional traffic state identification methods typically take into account one single macroscopic characteristic parameter, while overlooking the impact of vehicle lane-changing behaviors and the consequent mutual interference among vehicles. This leads to a relatively coarse granularity in the state division space and insufficient refinement in state identification, thereby hindering in-depth analysis of traffic congestion causes. In response to this, this study proposes a traffic state identification method based on vehicle dynamic behavior characteristics from an Unmanned Aerial Vehicle (UAV) perspective. Firstly, the method combines a vehicle detection algorithm (YOLOv8-OBB) based on rotated bounding boxes and a vehicle tracking algorithm (BoTSORT) to detect and track vehicles, addressing redundant background pixels and overlapping vehicle bounding boxes within horizontal bounding boxes, extract more accurate vehicle trajectory data such as vehicle spatial direction angle and four-point rotation coordinates, and calculate microscopic traffic flow parameters. Secondly, by utilizing the obtained vehicle driving direction angles and rotated position information, this study proposes vehicle dynamic behavior characteristics parameters: lane change interference rate and vehicle direction fluctuation index. Combined with macroscopic average speed and traffic density parameters, a multi-dimensional state feature space is constructed and applied to traffic state identification in actual road scenes. The ultimate experimental results demonstrate that the method achieved an mAP@0.5 of 0.987 in the rotated vehicle detection, with stable and continuous vehicle trajectory data output. In traffic state recognition, by introducing the lane change interference rate based on the average speed and traffic density as macroscopic feature parameters, the state recognition precision reached 0.983. Moreover, incorporating the direction fluctuation index, the state recognition precision reached 0.987. Additionally, according to the state characteristic space representation, the traffic state enables accurate classification into four states: smooth state, steady state, crowded state, and blocked state. This allows for quantitative analysis of the impact of vehicle dynamic behavior on traffic state, and provides novel theoretical insights for traffic state recognition from a UAV perspective, offering advanced fine-grained perception capabilities for intelligent transportation systems

In order to scientifically and effectively set up dedicated lanes for connected and automated vehicles (CAV), reduce conflicts between human-driven vehicles and CAV, and improve the safety and efficiency, this paper proposes a bi-level optimization model for urban road dedicated lanes for the CAV. The upper model has two optimization objectives, which aim to reduce the total travel time cost and reduce the traffic accident rate. The model also considers whether to set up a dedicated lane for CAV as a decision variable and uses the Strengthened Elitist Genetic Algorithm (SEGA) to solve the problem. The distribution of traffic flow in the lower model follows the user equilibrium principle. The lower model is solved using the Frank-Wolfe algorithm integrated with the golden section search. The Nguyen-Dupuis network serves as a case study to evaluate the model's effectiveness and the SEGA algorithm, and the effects of different penetration rates, capacity and total origin-destination (OD) demand on total travel time and traffic accident rate are analyzed. The results of the example show that the establishment of dedicated lanes for CAV will lead to an increase of total travel time regardless of the penetration rate of 10% or 80%. The total travel time decreases by 2.68% when the permeability is 40%. The accident rate would increase to different extents, when the penetration rate is below 20%. When the penetration rate is 40%, the accident rate decreases by 14.30%, and when the penetration rate is 80%, the accident rate decreases by 5.77%. With the increase of total OD demand, the total travel time would decrease by 0 to 6%, and the trend is insignificant. The accident rate reduces by 13.45% at low traffic demand (i.e., 14400 vehicles per hour), and reduces by 3.16% at medium traffic demand (i.e., 21600 vehicles per hour), and reduces by 10.35% at high traffic demand (i.e., 31200 vehicles per hour). The most suitable condition for establishing the CAV dedicated lanes is when traffic demand is low and the penetration rate is around 40%. The findings of this study can provide a theoretical foundation for establishing urban roads dedicated lanes for CAV.

The uncontrollability inherent in human-driven vehicles poses challenges to the efficient utilization of bus lanes with intermittent priority (BLIP). To address this issue, a control method for reusing BLIP in connected and automated vehicles (CAVs) is proposed. The lane-borrowing control takes into account the time-space constraints of bus movements, while the lane-returning control focuses on coordinating with adjacent CAV platoons to handle scenarios where the safe distance for lane-return is inadequate. The proposed method is simulated via an open-boundary cellular automaton model. Simulation results showed that: (1) At a given traffic flow, reusing BLIP with CAV can significantly improve road traffic efficiency. Notably, a moderate CAV penetration rate yields the most substantial improvement, with the average road speed rising from 6.67 km·h^-1 to 30.53 km·h^-1. (2) Irrespective of the CAV penetration rate, collaborative lane-changing within CAV platoons is more effective in boosting road traffic efficiency compared to single-CAV collaborative lane-changing, with an increase in the average road speed by 8%~19%.

To address the congestion and secondary accidents on highways involving both Connected and Automated Vehicles and Human-driven Vehicles after abnormal incidents occur, this paper focuses on a single lane and proposes a lane-based speed regulation method. This study utilizes the controllability of Connected and Automated Vehicles by controlling the passing speed to indirectly guide the driving behavior of Human-driven Vehicles. The area near the bottleneck is divided into a speed limit area and a coordination area. In the speed limit area, the Connected and Automated Vehicles speed limit values for different lanes are determined based on real-time traffic flow at the bottleneck, and the number of vehicles flowing into the coordination area is controlled to alleviate the formation and propagation of congestion waves. In the coordination area, the Connected and Automated Vehicles movement on the incident lane is controlled to ensure that vehicles could pass through the bottleneck area safely and efficiently. A set of simulation experiments are conducted, and the effectiveness of the proposed method is verified from two aspects: efficiency and safety. The simulation results show that compared to an uncontrolled baseline scenario, the average traveling time of vehicles can be increased by 2.4% and the improvement rate of TET (Time Exposed time-to-collision) can reach 14% in a 50% Connected and Automated Vehicles market penetration rates environment. And in a 90% market penetration rates environment, the average travel time can be increased by 18.5%, and the TET improvement rate rises to 51%. This paper could provide strategic recommendations and methods to control the traffic flow when an incident happens under the mixed flow environment.

To address the significant deviations observed between simulated and actual pedestrian trajectories in urban rail transit stations with traditional pedestrian simulation models, this study proposes a data-driven pedestrian simulation method based on a recurrent neural network (RNN). The approach integrates interaction mechanisms between pedestrians and obstacles, perspective mechanisms, and destination attraction mechanisms derived from the traditional social force model to enhance the alignment between simulation results and real trajectories. Additionally, a residual network structure is employed to account for the impact of complex station environments on pedestrian movement. A conditional variational autoencoder (CVAE) is used to incorporate local path endpoint predictions, further improving the accuracy of simulation outcomes. The proposed method is validated by using several public datasets and video surveillance data from subway stations in a Chinese city. Simulation results demonstrate that, compared with the traditional social force model and other methods, the proposed approach reduces the average displacement error and endpoint displacement error by at least 11.9% and 10.2%, respectively. Moreover, the incorporation of scene information and endpoint information corrections decreases the up two errors by an additional 28.1% and 25.9%, respectively, confirming the effectiveness of the proposed method.

Car-following states reflect the risk level of vehicle following behavior. To predict the car-following state changes on the curve sections of mountainous two-lane roads, this paper uses video data collected by drones to develop a car-following state transition prediction model based on higher-order Markov chains. First, from the preprocessed data, the car-following trajectory characteristics are extracted, and factor analysis is used to refine common factor characteristics that represent car-following states. Then, the K-Means++ algorithm is utilized to cluster the common factor characteristics. The car-following states are categorized into three states: strong car-following state, weak car-following state, and a transitional zone between strong and weak car-following states. A higher-order Markov chain model is then proposed to predict the car-following state transitions on mountainous two-lane roads. The results show that the transition between strong and weak car-following states involves a state transition process. During strong car-following, the leading car behavior significantly constrains the following car state, causing the following car's speed to change with a delay in response to the leading car. As the car-following state transitions from strong to weak, this constraint gradually decreases. The seventh-order Markov chain model achieves a prediction accuracy of over 97.6% for car-following state transitions. The self-transition probabilities for the three car-following states are respectively 97.57%, 98.90%, and 96.74%. In terms of state transitions, the direct transition probability between strong and weak car-following states is low, with the transitional zone playing an important role in the transition pattern. This proposed method demonstrates good performance in predicting car- following state transitions, and the research results can provide a methodological foundation for the development of active safety pre-warning systems for collisions.

The identification of key nodes in a rail transit network is critical to evaluate the network robustness and develop risk resistant plans and therefore ensure efficient operation of the transit network. This paper considers the mutual influence between nodes in the rail transit network and selects the Degree Centrality (DC), Betweenness Centrality (BC) and Closeness Centrality (CC) as comprehensive measurement indicators of node importance. The real rail transit network is converted as the corresponding topological network. The key nodes of the rail transit network are identified through the gravitational influence model, and the differences in network performance under different influencing factors are analyzed to obtain the optimal gravitational influence radius and attack strategy. The study assesses the robustness of the rail transit network from a gravitational perspective, and proposes relevant improvement recommendations. The results indicate that the importance of nodes is composed of the gravitational attraction generated by the target node and other nodes. When the gravitational influence model has a gravitational radius R=8 and a dynamic attack strategy is selected, the relative size decrease rate of the largest connected subgraph is respectively 13.25% and 10.39% higher than that when R=7 and R=9. The relative size decrease rate of network passenger flow efficiency is respectively 5.12% and 6.71% higher than that when R=7 and R=9 . Compared with the FGM, GC, KSGC, CI recognition models, the gravitational influence model has obvious advantages in identifying key nodes in rail transit networks. In addition, after attacking the top 30 nodes, the relative size of the largest connected subgraph in Beijing's subway network decreases by 91.68%, and the relative size of network passenger flow efficiency decreases by 86.17%. The results show that the gravitational influence model is applicable and effective in Beijing's subway network. The proposed method provides a new perspective for analyzing network robustness and provides an effective basis for decision makers to create network risk prevention plans.

To reduce carbon emissions during train operations and construct a sustainable and green urban rail transit system, this paper presents an optimization approach for the timetables and stop plans of express and local trains in urban rail transit, integrating carbon trading. First, the travel path selection mechanism of different passenger types is characterized and a calculation method for passenger travel time is proposed. Next, by computing the carbon emissions from train traction, ventilation and air conditioning, lighting, and signal systems, and incorporating the carbon emissions of the stop plans of express and local trains with carbon trading, the impact on operating costs is determined. A bi-objective nonlinear optimization model is established to minimize both passenger travel costs and enterprise operating costs. A two-stage solution algorithm is designed, which involves determining overtaking stations and partition- based calculation using Gurobi. Finally, a case study is provided to validate the model and algorithm. Through the case analysis, it is evident that different positions of overtaking stations yield different results. Compared with the all-stop mode, the operation mode of express and local trains considering carbon trading reduces passenger travel time by 5.8%, carbon emissions by 17.4%, and enterprise operating costs by 5.3%. The research findings indicate that the organization of express and local trains considering carbon trading has remarkable effects in reducing passenger travel time, carbon emissions, and enterprise operating costs. With the approaching of carbon peaking and the gradual rise in carbon prices, the carbon trading mechanism will effectively lower enterprise operating costs and encourage enterprises to actively reduce carbon emissions, thus promoting the sustainable development of urban rail transit.

As the "dual carbon" goal of Carbon Peak and Carbon Neutrality rises to the national strategic level, the establishment of a green transportation system has become increasingly urgent. This paper proposes a train operation plan for urban rail transit that focuses on the core principle of "efficiency enhancement and carbon reduction" in green transportation. The operation plan is based on a multi-route, multi-type formation configuration, taking into account the benefits of resources, the environment, passengers, and enterprises. To investigate the impact of different routes on passenger flow distribution, passenger flows are classified based on their travel characteristics, and an analysis of the associated travel costs for each passenger group is conducted. A multi-objective optimization model is established with the objectives of maximizing train transportation resource utilization, minimizing carbon emissions during train operations, and reducing both the operational expenditures of enterprises and the time costs associated with passenger travel. The model is subject to various constraints such as line capacity, departure frequency, and the number of vehicles in operation. To solve the model, an improved Sparrow Search Algorithm (SSA) was proposed, with a comparative analysis conducted against a full-length route, single-type formation operation plan. Furthermore, the solution results were compared with those obtained from the traditional SSA and Particle Swarm Optimization (PSO) algorithms. The results demonstrate that the multi- route, multi-type formation operation plan performs better than full-length route, single-type formation plan in terms of capacity utilization, carbon emissions reduction, enterprise operational costs, and passenger travel time costs. Moreover, the improved SSA shows significant advantages over traditional algorithms in terms of solution efficiency and quality. Therefore, the method proposed effectively balances the interests of enterprises and passengers, and also enhances resource utilization and reduces carbon emissions, providing strong decision-making support for the green operation of urban rail transit systems.

In the urban rail transit crew plan, limiting the deadheading position of the crew can reduce the number of crew shifts and increase the flexibility of the plan. First, with the minimum number of crew shifts, the shortest non-essential labor time and the shortest convenience time as the optimization goals, this paper proposes the 0-1 integer programming model of the crew scheduling plan with limited deadheading positions considering the constraints of the coverage of the multiplying section, the intermittent continuation, the limitation of working hours, the time window for departure and departure, and the continuation of meals. Then, the column generation algorithm is designed to solve the model, and the decomposition model is divided into the main problem of linear relaxation covered by the set and the shortest circuit of the pricing sub-problem. The bi-directional labeling algorithm for solving the pricing sub-problem is designed by constructing the spatio-temporal axis network based on the train timetable. The effectiveness of the model and algorithm is verified by a case study in a city rail transit line 1. The results showed that the number of morning, day and night shifts was respectively 47, 53 and 48. After limiting the deadheading position, the working time of each crew shift showed a downward trend, which could effectively improve the efficiency of the crew shift. Compared with the labeling algorithm, the solving efficiency of the two-way labeling algorithm is improved by 40%.

The operation mode of full-length and short-turn routings is the main method to solve the imbalance of passenger flow in urban rail transit system. The simulation study of train operation for full-length and short-turn routings is helpful to improve the efficiency and ensure the safety of train operation. Considering the characteristics of two typical routing modes of single-routing and full-length and short-turn routings, this paper proposes a method to calculate the minimum safe tracking distance of trains and develops a simulation model of urban rail transit train operation based on cellular automata. The influence of train braking performance, signal system and line conditions on train operation acceleration are also analyzed. The train tracking and turn-back operation are simulated and verified using an urban rail transit line as an example. The influence of train speed and braking performance parameters on line operation capacity is analyzed, and the variation law of the carrying capacity and the number of rolling stocks is discussed under different operation ratios of full-length and short-turn routings. The results show that considering the change of acceleration in the process of train operation can significantly improve the simulation accuracy of the system. The simulation model of urban rail transit train tracking operation based on dynamic safety interval can effectively shorten the minimum safe tracking distance of trains. When the ratio of full-length and short-turn routings is 1∶1, the carrying capacity is about 33.33% higher than that of single routing. The increment of the number of rolling stocks shows a periodic trend with the increase of the ratio of the turn-around times of the full-length and short-turn routings. When the number of rolling stocks is constant, the carrying capacity is the largest under the operation ratio of 1∶2. This study can provide a theoretical basis for the urban rail transit timetable optimization and train operation management.

To analyze the spatiotemporal coupling relationships of passenger flow and the passengers travel patterns in urban rail transit, this paper proposes a spatial-temporal double-hypergraph neural network model considering the urban built environment, which is called ST-DHGNN (Spatial Temporal-Double Hypergraph Neural Network). The model is consisted of a double hypergraph neural network module and a time series module. The double-hypergraph neural network module is designed to uncover high-order connectivity among rail transit line stations and clustering relationships among neighboring stations within similar built-up areas. The time series module is utilized to represent the temporal dependencies in historical passenger flow data. Additionally, a new loss function is developed with the built environment and lines as variables, aiming to dissect the impact of the built environment and enhance the model's predictive performance. An empirical study is conducted using Wuhan Metro data as an example. The results indicate that: (1) Considering the built environment and high-order connectivity among rail stations significantly improves passenger flow prediction accuracy. The proposed model achieves Root Mean Square Error (RMSE) of 52.04 and Mean Absolute Error (MAE) values of 29.32. The performance was improved by over 22% compared to baseline models. (2) The ablation experiments have verified the contribution of integrating high-order connectivity relationships of movement trajectories and the built environment to the model's performance. Specifically, in the single-step prediction task,considering these two factors improves the model's performance by 6% and 9%, respectively. In the multi-step prediction task, the improvements are 4% and 12%, respectively. (3) The constructed interpretable loss function incorporating built environment factors enhances the model's predictive capability while imparting better scientific rigor and interpretability. The study can provide technical support for passenger flow management and train scheduling in urban rail transit.

To enable delayed high-speed trains to quickly resume normal operation while satisfying requirements such as the exactness of train stop, punctuality, energy efficiency, and real-time scheduling, this paper proposes a joint optimization method for dynamic scheduling and speed curves, aiming to addressing the complexities of the integrated models in balancing nonlinear constraints and the limitations of solving multiple independent models separately in non-integrated models. First, based on the constraints of the reference system, the Intrinsic Curiosity Module Prioritized Experience Replay Dueling Double Deep Q-Network (ICM-PER-D3QN) is applied to optimize the train speed curve model, ensuring the exactness of train stop, punctuality, and energy efficiency. This data is used as the foundation for joint model training. Second, the ICM-PER-D3QN algorithm is utilized to solve the dynamic scheduling model of trains, alleviating delays and ensuring the real-time performance of scheduling. Utilizing train operational data in sections, a convolutional neural network integrated with a long short-term memory network is used to jointly optimize train speed curves and dynamic scheduling. The experimental environment is based on downward line of the Beijing Shanghai High-Speed Railway, with three delay scenarios designed. Simulation results show that, under the joint optimization model, the average scheduling time of trains is 0.92 s, the average matching accuracy between dynamic scheduling results and speed curves reaches 98.89%, and the average matching time is 0.0014 s. In addition, compared to the unoptimized speed curves based solely on the dynamic scheduling model, the average traction energy loss is reduced by 9%, and the average total delay time is reduced by 6.38%.

Based on vehicle travel time calculation under different traffic congestion conditions, the time-dependent crowdshipping pickup and delivery problem (TD-CPDP) is proposed, and a mixed integer programming model is developed by considering joint optimization of routing and speed for crowd-shipping vehicles. An improved adaptive large neighborhood search (IALNS) algorithm, incorporating a free-flow speed optimization mechanism, is designed to integrate route and speed decisions. The algorithm features a novel returned optimal searching strategy to avoid local optima, destroy-repair operators for exploring the solution space, and an adaptive mechanism for operator selection to improve search efficiency. Comparative validation with perturbation-based, ant colony, variable neighbourhood search, and adaptive large neighbourhood search algorithms demonstrates the superiority of IALNS algorithm. Sensitivity analysis reveals that, compared to constant high or low vehicle speeds, free-flow speed optimization reduces distribution costs by 3.97% and 20.91% , respectively, while maintaining low carbon emissions. Distribution costs and emissions are sensitive to occasional driver detours and compensation pricing, but both can be kept low when vehicle travel time limits and compensation pricing are within reasonable ranges.

Existing research on emergency vehicle route planning typically treating the deployment of dedicated lanes as a predetermined condition. This study proposes a bi-level programming model that concurrently optimizes emergency vehicle route and dedicated lane deployment. In this model, the deployment of dedicated lanes on each road segment is defined as a decision variable in route planning. Prospect theory is introduced to evaluate the impact of these dedicated lanes on traffic saturation. The objective function of the upper-level model comprises two components: emergency vehicle travel time and the prospect value of emergency vehicles travel smoothly, with the latter serving as the decision criterion for dedicated lane deployment. The lower-level model performs traffic assignment based on Wardrop's Equilibrium Principle. An innovative GA TS algorithm, combining tabu search and genetic algorithm, is proposed to solve the model. Numerical experiments on the Nguyen-Dupuis simulation network verify the effectiveness of the model and algorithm. Results show that this model reduces emergency vehicle travel time by 10.69% without increasing traffic saturation. Multiple experiments under different traffic demands further confirm that the proposed model can effectively shorten the travel time of emergency vehicles with insignificant impact on traffic saturation, and this reduction becomes more pronounced with increasing traffic demand. Additionally, the GA-TS algorithm improves solving efficiency by reducing the average solution time by 87.02%.

To address the problem of untimely capture of expressway network topology changes, this paper proposes a method for dynamically generating expressway network topology based on vehicle trajectory information. This paper first analyzes the abnormal topology patterns by mining the massive Electronic Toll Collection (ETC) data and combing the expressway network topology structure validity constraints. Then, the topology optimization rules are defined according to the characteristics of abnormal topology patterns. Based on the optimized topology set, the LightGBM model is used to realize the dynamic generation of expressway network topology structure. Taking the actual vehicle trajectory data of Fujian provincial expressways as an example, the study verifies the effectiveness and feasibility of the propose model and demonstrates the accurate generation of the provincial expressway network topology structure can be achieved through training and learning of local regional expressway network data. The results show that the expressway network topology structure generation accuracy of the model reaches 98.3%, and it can be extended to expressway nodes such as toll stations, ETC gantries, and service areas that have the ability to collect vehicle trajectory information, providing strong support for the refined management of expressways and personalized travel services based on the road network topology analysis.

To investigate the impact of cognitive distraction on drivers' car-following behavior on expressways, this study conducted driving simulation experiment with various distraction tasks. The study dynamically collected vehicle kinematics characteristics, driver manipulation, and eye movement parameters, and analyzed the influence mechanism of the secondary task state and speed interval on car-following performance. A set of cognitive distraction state representation parameters was developed for car-following behavior in different speed intervals. Methods such as the Support Vector Machine (SVM), Random Forest (RF), and Extreme Gradient Boosting (XGBoost) were introduced to identify drivers' cognitive distraction states in real-time. The findings indicated that immersive computing imposed a higher cognitive load on drivers compared to conversational secondary tasks. Cognitive distraction reduced drivers' control over the steering wheel and throttle pedal, more focused gaze on the road ahead, and suppressed visual transfer. The cognitive distraction representation parameters varied across different speed intervals. The XGBoost model outperformed both the SVM and RF. By calibrating the optimal sliding window width and step size under different speed intervals, the XGBoost model achieved recognition accuracies of respectively 85.98%, 87.98%, 88.45%, and 92.21% for the overall interval and the speed intervals of I ([60, 80) km·h-1), II [80, 100) km·h-1), and III [100, 120] km·h-1). Up to the risk threshold moment, the recognition rate for cognitive distraction samples reached a maximum of 90%. The findings provide references for recognizing cognitive distraction and optimizing early warning systems on expressways.

Electric bikes (e-bikes) are experiencing rapid development and growing popularity in China. In order to investigate the influencing factors of crash severity between e-bikes and vulnerable road users under the technical standard update, and to analyze the possible heterogeneity and temporal instability. This study aims to investigate the factors influencing the crash severity between e- bikes and vulnerable road users under the update of technical standards by analyzing potential heterogeneity and temporal instability. Using a dataset of 6022 e-bikes crashes collected in Yancheng, China, potential contributing factors to injury severity were selected from five aspects: rider, accident, road, time, and environment. To explore the potential heterogeneity, a random parameter Logit model with heterogeneity in means and variances was employed. The log-likelihood ratio was utilized to test the temporal instability of the factors influencing crash injury severities. The average marginal effects of each variable were calculated to measure their impacts on the injury severity of e-bike crashes. The results showed that the random parameter Logit model with heterogeneity in means and variances outperforms both the standard model without considering heterogeneity and the binary Logit model in terms of fitting and accuracy. Moreover, under the influence of the 2019 national standards, there is significant temporal instability in the factors affecting e-bike crash severity, resulting in substantial changes in the important influencing variables before and after the update of the national standards. In the old-standard model, the variables of e-bikes and no controls are random variables, and the factors of rural roads and sand roads increase their mean and variance respectively. In the new-standard model, the variables of e-bikes and sign markings are random variables, and the factors of autumn and hit-and-run increase the mean of the e-bikes parameter. The research findings offer valuable insights for developing interventions for road traffic crashes involving e bikes and provide a theoretical support for updating the current technical standards for e-bike safety.

This paper studies the slot allocation problem for container liner shipping. Traditional slot allocation models have not fully taken into account demand liquidity and revenue fluctuations. This paper considers demand transfer and demand market segmentation, i.e. spot market, short-term contract market, and long-term contract market, where demand transfer occurs between the spot market and short-term contract market. To reduce revenue fluctuations caused by demand uncertainty, this paper introduces risk aversion theory into the studied problem. Then, a mixed-integer nonlinear slot allocation model is constructed. Based on model transformation and linearization, we obtain a linear programming model that can be efficiently solved by Gurobi. The case study is based on an Asia-Pacific-Europe shipping route operated by Maersk Line. Numerical results show that: (1) compared with the scenario without considering risk, the revenue fluctuation reduction in the scenario considering risk aversion can be approximately 80%; (2) the establishment of short-term contract market can effectively promote revenue growth and reduce revenue fluctuation; (3) sensitivity analysis on risk coefficients indicates that shipping companies need to actively monitor the spot market and promptly adjust their slot allocation strategies based on market conditions to balance the trade- off between revenue and revenue fluctuation.

Aircraft taxiing routing planning is an important issue in airport traffic management, which requires the balance of efficiency (optimal routing for the airport system) and fairness (equitable service among airlines). This paper investigates a bi objective aircraft routing planning problem that incorporates both efficiency and fairness, along with considerations such as safety separation time between aircraft and time windows. The problem is formulated as a bi-objective mixed-integer nonlinear programming model, aimed at minimizing total aircraft delay and minimizing the maximum average fairness deviation among airlines. Given the characteristics of the model, an ε-constraint method coupled with an initial solution algorithm acceleration strategy, is proposed for the model solution. The generated solutions offer information into the trade-offs between the competing objectives without requiring decision-makers to articulate specific preferences. Numerical experiments based on Guangdong Baiyun Airport are conducted to validate the applicability and effectiveness of the model and algorithm. Results show that the proposed ε-constraint method, enhanced by the acceleration strategy, outperforms the classical ε-constraint method in solving the problems, with the acceleration becoming more pronounced as aircraft numbers increase. Achieving fairness among airlines often entails a sacrifice in efficiency, suggesting that taxiing routing adjustments should be made during less congested periods.

Traditional trajectory optimization methods often encounter challenges in enhancing local flight efficiency while maintaining the overall stability of airspace operations. To address this issue, this paper introduces a multifaceted complexity calculation method specifically designed for airspace grid assessment and explores its application in trajectory search algorithms. Initially, interaction complexity is computed based on two motion tendencies: "approaching" and "converging", while background complexity is derived from airspace structure and meteorological conditions. Subsequently, these complexities are allocated to airspace grids, forming a complexity map of each grid. Finally, this method is integrated to an improved trajectory optimization approach to assess the impact of optimization results on airspace operational pressure. Validation is conducted through simulations in a modeled airspace and actual data from the Shanghai Terminal Area. Simulation results indicate that the operational pressure of air traffic scenarios can be quantified by multifaceted complexity. Compared to the original data, the complexity-assessed improved A* algorithm results in a 20.10% reduction in flight distance, an estimated 30.00% decrease in flight time, and a 16.67% reduction in maneuvers. Furthermore, when comparing with original data and traditional A* algorithm optimized trajectories, the optimized results demonstrate a decrease in local airspace operational pressure.

Automated electric container trucks (AECTs) have demonstrated outstanding economic and environmental benefits in port internal logistics. However, their application in external drayage operations still faces challenges such as complex and dynamic road environments and unstable energy supply. To extend AECTs to semi-open areas near ports, this paper proposes a multi-attribute complexity learning-driven approach for optimizing AECT drayage routes. Considering the dynamic and uncertain characteristics of port drayage scenarios, a large-scale road network complexity assessment method is designed by integrating multi-attribute decision-making and machine learning. Then, the operational design domain (ODD) constraints and driving range limitations are introduced into the drayage route optimization model, comprehensively considering the autonomous driving capabilities, energy consumption characteristics, and transportation efficiency of AECTs. Case studies show that: (1) roads of different grades exhibit significant differences in complexity levels and spatial distributions, with lower-grade roads generally having higher complexity than higher-grade roads; (2) when the ODD boundary complexity reaches 0.55 and above, the complexity learning-driven route optimization model can reduce the manual takeover rate of AECTs by 5.04%~16.83%, achieving “fully autonomous driving”in semi-open scenarios; (3) as the ODD boundary expands, AECT's autonomous driving performance and energy- saving effects gradually improve. When the ODD boundary complexity reaches 0.55 and 0.70, AECTs achieve autonomous driving on designated routes and arbitrary routes, respectively, with transportation costs reduced by 24.03% and 29.26% compared to traditional container trucks.

As the government gradually reduces subsidies to China Railway Express (CRexpress), the growth of train departures slows down, and the cargo structure on some CRexpress platforms experiences significant fluctuations. This paper extends the Hotelling model from customer spatial distribution to cargo value distribution, and proposes a game model that includes local governments, CRexpress companies, shipping companies, and shippers to determine the optimal government subsidy level and discuss how railway and shipping companies can clarify their respective cargo markets and set freight pricing strategies. Based on the regional cargo value distribution and shippers' preferences for transportation prices and time, this paper develops utility functions for local governments, railway express companies, and shipping companies. The game equilibrium state under different subsidy levels is solved using the grid method. The impact of changes in cargo value distribution, marginal transportation costs, and carbon trading prices on government utility and optimal subsidy values is also analyzed. The results from case study show that when the subsidy value for CRexpress is 1200 USD · FEU^-1, government utility reaches its peak, with equilibrium pricing for CRexpress at 8969 USD·FEU-1 and 4273 USD·FEU^-1 for shipping. The cargo values exceeding 252847 USD·FEU^-1 prefer high time-efficiency railway express, while the cargo below this value prefer low-cost shipping. Canceling subsidies will lead to a sharp decline in the competitiveness of CRexpress companies and a decrease in government utility, so local governments currently need to maintain a certain level of subsidies for CRexpress companies. In the future, the key to achieve complete subsidy withdrawal lies in reducing the marginal costs of CRexpress transportation.

To investigate car-following behaviors of light vehicles and heavy vehicles in freeway tunnels, this paper first used cameras and laser radars to collect vehicle driving trajectories at the Qifu Tunnel in Guangdong Province, and further extracted car following trajectory data for both vehicle types. Then, the Real-time Deviation of Safety Margin (RDSM) was proposed to assess the real-time following risk level, and the Fuzzy C-means algorithm was used to classify the risk level into no risk or low risk, moderate risk and high risk. Subsequently, 26 potential impact factors were selected from five dimensions, including preceding vehicle types, vehicle position in the tunnel, driving environment, current vehicle driving and interaction states, and historical vehicle driving and interaction states. Multinomial Logit model and correlated random parameter Logit model were applied to analyze the effects of each factor on real-time following risk in the freeway tunnel for light and heavy vehicles, and reveal the heterogeneity of the impact factors. The results show that heavy vehicles are affected by more factors regarding their real-time following risk in the tunnel. When the following vehicle differs in vehicle type from the preceding vehicle, the real-time following risk is relatively reduced. Fluctuations in the driving states of the preceding vehicle are more likely to cause high-risk following. Average marginal effects indicate that, compared with the car-following real-time risk at the tunnel entrance, the occurrence probability of high-risk car-following behavior at the tunnel exit for light vehicles increases by 0.0413, while that at the tunnel internal segment increases by 0.0155 for heavy vehicles. Moreover, the standard deviation of the following distance in high-risk states exhibits heterogeneity in both vehicle types.

To investigate the spatiotemporal distribution characteristics of taxi carbon emissions, this study extracts parameters such as average speed and travel distance between trajectory points from taxi Global Positioning System (GPS) data and constructs the Computer Programme to Calculate Emissions from Road Transport (COPERT) micro- emission model to quantify taxi emissions. Based on this, distribution fitting is used to analyze the distribution characteristics of emissions over time, space, and among vehicles. Based on the analysis results, two traffic management measures—taxi restriction and speed control—are proposed, and their emission reduction potential is evaluated through numerical simulations. An empirical study in Changzhi City shows that the taxi industry exhibits a zero-sum game phenomenon, with emissions more evenly distributed between 8:00-13:00 and 14:00-22:00. Emissions aggregated at nodes and road segments follow a truncated power-law distribution, with the top 10% of nodes and road segments accounting for 95.59% and 74.71% of emissions, respectively. The evaluation results indicate that the taxi restriction policy can reduce emissions by up to 20.35%, maintaining a taxi speed of 15 m ⋅ s-1 results in the lowest emission factor, potentially reducing emissions by up to 21.43%. Implementing speed control on the top 10% of road segments with the highest emissions can reduce emissions by 16.23%, while randomly selecting the same number of segments for speed control can only reduce emissions by 2.37%. The results can support the development of refined carbon emission control strategies and energy-saving measures for urban transportation.

To address the limitations of the traditional manual experience-based districting of Comprehensive Inspection Train (CIT) inspection areas, this paper proposes an optimization method for the districting plans based on a mixed-integer linear programming (MILP) model with the objectives of task balance and structural compactness. The method incorporates key constraints such as coverage integrity, inspections under up-to-speed conditions, line-vehicle attribute matching, and inspection area connectivity. The adjudication constraints are defined for connectivity with complex heterogeneity and downward compatible and non-unique matching relationship based on multi-flow model. Based on the ideal point method, the inverse of each objective's optimal solution is used as a weight, converting the original multi-objective model into a single-objective MILP to maximize the overall performance. Computational experiments are conducted across various scenarios to identify the key factors influencing the districting plan's quality. The proposed method is then applied to an actual large-scale network to provide decision support for inspection resource allocation. Experimental results indicate that CIT heterogeneity significantly affects both the quality of the districting plan and solution speed. The proposed districting method effectively mitigates the limitations of manual methods by balancing inspection loads and reducing off-site travel. With 10 CITs, the total deviation from the ideal solution is 2.84%, demonstrating a marked improvement over existing manual method.

To improve the traffic operation of the city area and address the localized congestion and fiscal subsidy challenges arising from the free passage policy on the ring expressway, this paper proposes a combined traffic management scheme integrating tradable credit scheme and newly constructed connecting links. A bi-objective bilevel programming model is proposed and the interaction mechanisms is analyzed among government departments the ring expressway company, and travelers. The upper-level model outlines the joint scheme formulated by the government and the ring expressway company to achieve their respective objectives. In contrast, the lower-level model depicts traveler behavior under this joint scheme while constraining traffic flow on various links of the ring expressway to maintain its normal service levels. Additionally, an algorithm combining non-dominated sorting genetic algorithm II and Frank-Wolfe alogrithm is designed to solve this model. The results indicate that, compared to the current free passage scheme and other standalone solutions, the joint scheme achieves significant Pareto improvements, maximizing reductions in system travel time (by 18.56% ) and substantially increasing the ring expressway company's revenue (by 51.5%). A comparative analysis of urban road saturation distributions shows key differences between the standalone tradable credit scheme and the joint scheme. The joint scheme more effectively utilizes the ring expressway to alleviate urban traffic congestion. It reduces the number of severely congested links and increases the number of uncongested links. Moreover, it avoids excessively high credit price, making it more traveler-friendly. Additionally, the sensitivity analyses were conducted to examine the impacts of OD demand, upper limits of credit-charged links, number of new connecting links, and credit price on the effectiveness of the joint scheme.

To investigate the visual load characteristics of the drivers at the entrances and exits of an underground ring road, this study conducted a real-vehicle test and selected four sections, each with equal numbers of entrances and exits in the underground ring road of Jiefangbei. The sections were named as L0 to L3 and Tobii Glasses2 eye-tracking device was used to obtain the data of the visual behavior characteristics of the drivers. Utilizing traffic sign rate information density, average gaze duration, average blinking frequency, and pupil area change rate as evaluation indicators, a visual load evaluation model is constructed based on game theory combination assignment and cloud object elements to investigate the change rule of visual load characteristics of drivers at the entrances and exits of the underground ring road. This study shows that: the combined weights of traffic sign rate information density and average gaze duration are relatively significant, accounting for 31% and 35% , respectively, and the combined weights of average blinking frequency and pupil area change rate are lower, accounting for 16% and 18%, respectively. The visual load level of the drivers increases progressively from L0 to L3, shows the growth trend from low to high load. The correlation of the indexes for the traffic sign rate information density indicator demonstrates a clear trend from low to high load, whereas the other indicators show less an inconspicuous trend. In terms of sensitivity, the evaluation indicators rank as follows:traffic sign rate information density>average gaze duration>pupil area change rate>average blinking frequency. To reduce drivers' visual load, it is recommended to provide advance notice of road information or minimize visual stimuli. Possible measures include increasing the distance between traffic signs and reducing the repetition of directional signage.