Northwest China's Gansu Province situates at the heart of the Belt and Road corridor. In recent years, taking geographical advantages, the province is experiencing the transformation from the corridor to a hub of the new Eurasian Land Bridge. While serving the national strategy, Gansu focuses on building distinctive industrial clusters along the corridor, thus converting connectivity advantages into industrial competitiveness. With the subject of "The Belt and Road and Regional Economic Development", it is discussed in this session how the province can actively serve and be deeply integrated into national strategies, how to promote the inter-connectivity between strategic corridors and the reinforcement and supplementation of chains within corridor, and how to facilitate the efficient transformation of corridor resources into a multimodal transportation and logistics system, continuously boosting regional economic development.

Comprehensive transportation is the development direction of intelligent, green and safe transportation in recent years. Through comprehensive development and utilization of aviation, railway, highway, waterway and pipeline transportation modes, it builds a transportation system with advanced transportation technology, reasonable layout and structure. Air-rail intermodality is one type of comprehensive transportation, which not only reduces the waste of resources through the reasonable use of existing transportation infrastructure, but also improves transportation accessibility through low cost of transportation mode shift. Air-rail intermodality is an effective strategy to ease airport congestion, expand the scope of airport radiation, reduce carbon emissions and meet the diversity of travel demand. In recent years, related research on air-rail intermodality has included research on competition and cooperation of air-rail intermodality, air-rail intermodality network, and evaluation methods of air-rail intermodality at macro level, then research on of air-rail intermodality at the middle level, and research on passenger travel behavior of air-rail intermodality at the micro level. This paper summarizes and the representative studies in domestic and international level of air-rail intermodality from 2000 to 2024, and sorts out the research results under different operation backgrounds and different transportation network scales. It is found that domestic scholars are more concerned on transport passenger travel behavior, emphasize passenger travel demand, and scholars in other countries tend to focus on air-rail intermodality social benefits and environmental benefits, emphasize the sustainable development of air-rail intermodality. The reasons of the research differences at home and abroad are analyzed from the social background, technical background and policy support. In the future, to explore practical collaborative research, including conflicts of interest between operators and between operators and passengers, the conflict between social and environmental benefits is a valuable research direction. Systematic research on the construction of air rail intermodal network, the location of transit nodes to network optimization, construct the evaluation method system of air-rail intermodality, it has guiding significance to the development of air-rail intermodality. Expanding the study area for air-rail intermodality with the considerations of the diverse needs of travelers, coping strategies that incorporate uncertainty factors, such as providing flight delay insurance, is a very meaningful research topic. It is a practical work to deeply integrate the theoretical research and practical application of passenger travel behavior of air-rail intermodality. Atlast, the future development of air-rail intermodality should use advanced science and technology to reduce the data acquisition cost of the whole travel chain of air-rail intermodality passengers and build a travel service platform of air-rail intermodality including travel path planning, one-stop ticket purchase, connection process visualization, connection time prediction and other services.

Accurately quantifying the competitive and cooperative relationships between rail transit and bus services is a prerequisite for rational public transport route planning. From the perspectives of passenger competition and complementarity, this paper analyzes the suppressive and facilitative effects of rail transit on bus ridership. The concepts of Competition Degree, Cooperation Degree and Coopetition Degree are developed within a unified dimension, and an evaluation model is constructed across three hierarchical levels: station pair, bus station, and bus route. On this basis, the discrete choice model of residents' travel mode is designed to quantitatively calculate the indicators. A case study on four typical bus routes in Hangzhou was conducted, comparing the distribution changes of these indicators before and after incorporating passenger flow factors. The results showed that incorporating passenger flow factors significantly altered the evaluation outcomes for some routes. Specifically, the relationship between bus route 143 and subway line 2 has shifted from cooperation to competition, with the coopetition degree rising from 0.069 to 0.999. In addition, this paper verifies the rationality of this change through the visualization results of travel demand distribution, indicating that the proposed competition and cooperation model can effectively quantify the competitive and cooperative relationships between rail transit and bus transit at three levels, and obtain more realistic and reasonable evaluation results. The research findings can be used to identify bus routes that exhibit significant competition or low cooperation with rail transit, providing a scientific basis for coordinated optimization of the urban public transportation system.

The Northeast China Sea-Land Transport Corridor emerges as a novel international intermodal transport route, seamlessly integrating maritime and China Railway Express railway systems. This paper addresses the intermodal route choice problem for intermodal transport operators between China and Europe, specifically focusing on the Northeast China Sea-Land Transport Corridor. First, accounting for intermodal transport risks stemming from regional conflict events, an evaluation framework for intermodal transport risk indicators is developed from the perspectives of general risk and regional conflict risk. It incorporates the probability of risk occurrence and the severity level of risks. A multi-objective model for the intermodal transport route choice of China-Europe containers is constructed, aiming to minimize both total transport costs and transport risks. A NSGA II algorithm, based on topological sorting path chromosome coding, is devised to identify Pareto-optimal intermodal transport route choice schemes that align with the requirements of intermodal transport operators. Ultimately, three risk scenarios were established for numerical experiments based on the different stages of the regional conflict and its scope of influence. The findings indicate that the Northeast China Sea-Land Transport Corridor outperforms traditional corridors in complex and dynamic risk environments, demonstrating competitiveness for both high-value, time-sensitive, and time-insensitive cargoes. The competitiveness of Northeast China Sea-Land Transport Corridor increases and then stabilizes as the delivery time limit extends. It demonstrates strong competitiveness for goods with delivery time requirements within the medium range.

This paper addresses the route optimization and traffic allocation in multimodal transportation networks with complex structures and environments. An openness coefficient is established to reflect the impact of transportation market barriers on the actual structure of transportation networks and the transportation environment. This coefficient specifically reflects the accessibility of freight lines for different types of multimodal transport operators within the physical network. To optimize the paths and flow in multimodal transport networks, we comprehensively consider factors such as network structure and environment, flow balance, mode selection and conversion, as well as path capacity, and consequently develop a mixed-integer programming model. Taking a regional multimodal transportation network of China as an example, the model is solved using the GUROBI optimization solver to verify its effectiveness. The research findings indicate that when the openness level of the multimodal transportation network is either low or high, the total transportation costs decrease by 3.03% and 5.05%, respectively, compared to the original network, and the corresponding revenues from dedicated lines have increased. Additionally, with the enhancement of path capacity, the freight volume of low-cost paths that are capacity-saturated increases. The transportation volumes and path utilization rates under different path capacities provide various optimization solutions for the expansion and reconstruction of multimodal transportation networks.

The point-to-point direct transport is the major operation mode of the China-Europe Railway Express. However, the direct transportation mode needs long assembly time for vehicles grouping, which leads to decreased operational efficiency. This paper proposed the combined transportation modes of direct and assembly for the China-Europe Railway Express, and conducted the joint optimization for operation plan and train diagram. The ε-constraint method was used to obtain the Pareto front with the dual-objective of low transportation cost and short transportation time. The heuristic algorithm framework was embedded with CPLEX. The feasibility and effectiveness of the proposed model was verified through the case studies with 7 domestic stations and 20 cargoes in the China-Europe Railway West Line. It was found that reducing the transportation cost of China-Europe Railway Express would be at the expense of its transportation time. The direct and assembly transport modes would help to reduce the cost and shorten the time. The transportation cost of assembly and direct mode increased by 9.6% compared to only using the direct mode, while the transportation time decreased by 20.3%. Compared to the results of joint optimization, the train diagram obtained through operation plan optimization alone cannot meet the connection constraint and time limit for the train operations.

The rational layout of integrated transportation corridors is crucial for the coordinated development of multiple transportation modes within urban agglomerations. Utilizing travel characteristic data to identify integrated transportation corridors is a key problem that urgently needs to be addressed. Based on the mobile signaling data from urban agglomerations, this paper proposes an integrated transportation corridor identification framework which is suitable for various transportation modes. The framework consists of four components: data preparation, transportation mode classification, shortest path search, and corridor identification. This paper proposes a multi-modal division algorithm for high-speed rail, conventional rail, and road, with decision variables based on average transport speed and station POI locations. Additionally, a bidirectional A* algorithm is designed for shortest path search. Based on the obtained shortest paths and the spatial features of the road network, this study proposes an integrated transportation corridor identification algorithm through mobile signaling data. An empirical analysis is conducted using the Beijing-Tianjin-Hebei urban agglomeration as a case study. The corridor identification algorithm identifies 6 integrated transportation corridors to verify its effectiveness and availability. The identified two transportation modes: road and rail，account for 81.87% and 18.13%, respectively in terms of transportation volume. Compared to rail transport, road transport handles a larger share of short-distance passenger flows. Considering holiday travel characteristics, the presence of holiday factors significantly increases the passenger flow in the integrated transportation corridors, with average transportation volume rising by 62.6% and average passenger turnover increasing by 61.2%.

The transportation of perishable cargoes through the existing intermodal freight network has increased significantly. The focus on efficient transshipment of refrigerated containers has been driven by the strict quality requirements of perishable goods. Therefore, an intermodal transportation path optimization model is proposed aiming to minimize the average cost and quality degradation. Considering the uncertain of railway loading demand, the impacts of refrigeration supply and failure on quality degradation are explored. The non-dominated sorting genetic algorithm (NSGA-II) is adopted. A numerical experiment is conducted for the import of apples from the Port of Antwerp to Lanzhou. Results indicate that, although refrigeration failure time is brief, it can lead to the quality degradation of up to 40% compared to the supply state. The research provides robust transportation solutions for perishable products by recommending that the duration of single stops at nodes be limited to less than 11% of the total time to preserve freshness. For transfer station operators, shortening the duration of refrigeration failures and enhancing service levels within stations become as effective methods to attract shippers.

To enhance the accuracy of vehicle trajectory prediction on two-lane mountainous highways, this paper proposes a trajectory prediction model that considers the impact of lane-occupying overtaking. First, based on the Unmanned Aerial Vehicle video trajectory data, the lane-occupying overtaking process is divided into four states according to the heading angle: following, lane-occupying, overtaking, and returning. Second, a multivariate feature dataset is constructed, containing lane-occupying states, vehicle motion characteristics, spatial position attributes, and traffic conditions. The Gradient Boosting Decision Tree (GBDT) algorithm is then used to fit the complex relationships between lane-occupying states and vehicle motion characteristics, spatial positions, and traffic conditions. The SHAP (SHapley Additive exPlanations) method is used to identify key factors affecting lane occupying state changes. Then, the lane-occupying states, key factors influencing these states, and historical trajectory datasets are input into the Informer model in the form of a sliding time window to predict lane-occupying overtaking trajectories on two-lane mountainous highways. The effectiveness of this model is verified by comparing it with traditional overtaking trajectory prediction models that do not consider lane-occupying impacts. The results indicate that time headway, lateral speed of the main vehicle, and lateral offset are key factors affecting lane-occupying state changes. The proposed model demonstrates good applicability and prediction accuracy in the context of lane-occupying overtaking on two-lane mountainous highways. Compared with the trajectory prediction model without considering the effect of borrowed lane overtaking, the mean square error and mean absolute error of the model in this paper are reduced by 53.05% and 38.11%, respectively, and the mean value of the coefficient of determination is improved by 23.58%.

In complex multi-vehicle dynamic interaction scenarios, intelligent vehicles need to accurately perceive and predict the driving trajectories of surrounding vehicles to ensure safe and efficient driving. To address the issue that existing models fail to fully consider the dynamic coupling relationships among multi-dimensional features, this paper proposes a vehicle trajectory prediction method based on a spatio-temporal cross-attention mechanism. First, a spatial attention module is adopted to extract the dynamic interaction features between the target vehicle and surrounding vehicles from their historical trajectory data. Then, the obtained dynamic interaction feature parameters are input into a long short-term memory (LSTM) neural network encoder to capture cross-time dependencies from the time domain perspective. Subsequently, the hidden state of the encoder is input into a cross-attention module that combines Fourier transform and a learnable router to capture cross-time dependencies in the frequency domain and further extract the coupling features among multi-dimensional features. At last, the future trajectory of the target vehicle is generated through an LSTM neural network decoder. The model is trained, validated, and tested using the next generation simulation (NGSIM) dataset. The results show that the model has a root mean square error of 0.74 meters in the 5-second prediction time domain, which is a 10% improvement in accuracy compared to the best results of other prediction models (0.82 meters).

To accurately characterize the microscopic lane-changing behavior of vehicles in ramp merging areas under mixed passenger and freight conditions, this study constructs and validates a microscopic lane-changing model that considers the influence of trucks, using the quantification of car-truck interactions as a starting point. First, the concept of "oppression" is introduced, and an improved Morse potential function model is developed to describe the impact of surrounding trucks on the lateral and longitudinal driving behavior of passenger cars. Then, the changes in lateral and longitudinal "oppression" during the lane-changing process of passenger cars are analyzed, leading to the proposal of a vehicle lane-changing decision model that incorporates truck "oppression" effects: the Modified Morse-Based Lane-Changing model (MMBLC). Finally, numerical simulations based on real trajectory data are conducted to verify the stability and effectiveness of the model. Additionally, Python and SUMO (Simulation of Urban Mobility) are used for joint simulations to compare the performance of the MMBLC model with existing lane-changing models in traffic flow. The results show that, in stability analysis, the MMBLC model has a smaller impact on traffic flow and recovers stability more quickly. In a three-lane main road ramp merging area with a truck proportion of 30% and a traffic volume of 3600 vehicles per hour, the MMBLC model improves lane-changing success rates by 11.9% and 53.1% compared to the LC 2013 and MOBIL (Model of Optimal Control Based on Interacting Trajectories) models, respectively, while reducing the proportion of hazardous scenarios by 10.5% and 52.8%.

Existing research on abnormal group detection overlooks the characterization of individuals' latent behavior patterns, latent behavior patterns within neighboring groups, and the temporal variations in behavior patterns. In this context, this paper proposes a Spatio-temporal Dynamic Hypergraph Clustering (STDHC) model to address the above limitations. The study first extracts sequences of travel feature matrices for individuals across continuous time slices, based on which corresponding sequences of behavior pattern hypergraphs are constructed, with an aim to depict high-order correlation characteristics of multiple individuals at various time periods. Subsequently, the Transformer is used to capture latent behavior patterns underlying individuals' explicit travel features from the temporal dimension. The hypergraph convolutional network is utilized to model high order correlations of latent behavior patterns within neighboring groups from the spatial dimension. Additionally, the changes in the hypergraph topology structure under the bidirectional temporal propagation are measured to understand the temporal variations in individuals' behavior patterns. These three types of features are integrated using an attention mechanism to update the hypergraph convolutional network, enabling automatic detection of associated groups. The proposed model is applied to the detection of pickpocket gangs in public transit. Through a series of comparison, ablation, and robustness analysis experiments, it is demonstrated that the model achieves a performance improvement of 2% to 6% over six baseline models. The research findings provide theoretical support for the intelligent detection of abnormal groups in public transit and enhance safety and operational efficiency.

Path redundancy is one of important dimensions to evaluate the resilience of transportation networks, which has been widely recognized for reducing the impact and economic losses of road interruptions. However, we are still unclear about the value and role of path redundancy in daily trips of travelers. This study attempts to verify the impact of path redundancy on the travel time reliability among different origin-destination (OD) pairs from the perspective of travelers within a regional scope. Based on real-time travel time data collected in the road network of Xuhui District, Shanghai from April 26, 2023 to May 9, 2023, this study compares the travel time reliability of ODs having high path redundancy with that of ODs having low path redundancy determined by propensity score matching (PSM) method, and then conducts regression analyses. The results show that: (1) the high path redundancy can effectively improve the OD travel time reliability; (2) the improvement effect of high path redundancy on OD travel time reliability has significant regional differences. The improvement effect in some regions is as high as 7.32%, while it is only 1.94% in other regions; (3) the heterogeneity of path redundancy within the same region has a differential impact on the OD travel time reliability, and the improvement effect of OD travel time reliability shows a marginal decreasing trend with the increase of path redundancy. This study provides a method and basis for evaluating the value and role of path redundancy in daily trips from the perspective of travelers, and also helps guide planners and managers to develop targeted, effective, and moderate road redundancy investment plans within limited resources.

Driving behavior recognition is a core challenge in intelligent driving assistance systems, and classifying driving behaviors based on Electroencephalography (EEG) signals is crucial for achieving human-centered intelligent driving assistance. To enable five-class classification of EEG signals under common driving behaviors, this paper proposes a method based on recurrence plots and a convolutional neural network (CNN) with channel squeeze enhancement (RP-CS). The RP-CS method extracts nonlinear features from EEG signals by embedding one-dimensional time-series signals into a higher-dimensional phase space and constructing recurrence plots using Euclidean distances. These recurrence plots, which integrate both nonlinear and temporal features, are used as input to a CNN enhanced with a channel attention mechanism, enabling accurate five-class classification. Comparative experiments on a public dataset demonstrate that the RP-CS method achieves superior performance, with a maximum classification accuracy of 95.84% under subject-dependent conditions and an average classification accuracy of 71.92% under subject-independent conditions. The results indicate that effectively combining nonlinear EEG features with deep learning models can significantly enhance the efficiency of EEG-based classification. This approach offers a viable solution for driving behavior monitoring and safety assistance, contributing to improved performance in driving assistance systems and enhanced driving safety.

To address the limitations that traffic signal control cannot ensure the smooth passage of emergency vehicles (EMVs), a two-layer optimization strategy is proposed for emergency vehicle (EMV) to enable EMVs to borrow the green time of conflicting traffic flows. The strategy is based on the construction of spatial right-of-way for EMVs at intersections using dynamic bus lanes, in order to address the serious encroachment of social vehicle interests by the preemptive signal control strategy of EMV at intersections. The upper-level optimization takes the advantages of inducible or controllable trajectories of connected vehicles (CVs). With the optimization objective of minimizing the product of vehicles' average delay at intersections and reciprocal of the expected vehicle speed attainment, a joint optimization model for the trajectory of CVs and the signal timing is established. Based on the optimized signal timing, the lower-level optimization makes decisions on collaborative conflicting traffic flow and green time borrowing strategies, constrained by the achievement of EMV travel speed. Taking phase green time and CVs acceleration as decision variables, and considering constraints such as conflicts between intersection phases and vehicle trajectories, a dynamic programming model is established to solve the objective function. Simulation results show that, compared with only implementing signal optimization control, the proposed method can significantly reduce the travel time of EMV at intersections. Compared with preemptive signal control strategies, it can greatly improve the encroachment of social vehicle interests. It indicates that the hourly traffic volume of bus is a key factor restricting the rapid passage of EMV.

In a mixed heterogeneous traffic environment consisting of human-driven cars, trucks, and both types of connected and autonomous vehicles (CAVs), the implementation of CAV-dedicated lanes can effectively mitigate mutual interference among different vehicle types. This study investigates traffic flow management method specifically for freeway CAV-dedicated lanes. By analyzing the spatial distribution characteristics of various vehicle types, with considering factors such as the CAV penetration rate, truck mixed rate, CAV platoon mode, and platoon size, a Markov chain-based method is proposed to calculate the capacity of car truck mixed heterogeneous traffic flows. Subsequently, the traffic capacities of CAV-dedicated lanes and mixed lanes are calculated under the conditions of CAV dedicated-lane deployment. By considering weighted lane saturation and lane saturation balance, a bi objective optimization-based traffic flow management method for freeway CAV-dedicated lanes is developed. The validity of the proposed method is demonstrated through case studies. Results indicate that CAV penetration rate, truck mixed rate, aggregation factor, and CAV platoon significantly affect traffic capacity. Under connected and autonomous car (CAC) dedicated lane deployment conditions, with a traffic demand of 6000 veh·h^-1, a truck mix ratio of 0.1, and a CAV penetration rate of 0.6, the proposed method outperforms the conventional approach of prioritizing CAV allocation to dedicated lanes in terms of both weighted lane saturation and lane saturation balance. This research provides theoretical support for the optimal deployment and traffic flow management of freeway CAV-dedicated lanes.

To elucidate the stability evolution mechanisms of heterogeneous traffic flows composed of Human Driven Vehicle (HDV) and Connected and Autonomous Vehicle (CAV), the study investigates their dynamic behaviors under mixed and separated configurations. Considering CAV functional degradation, we developed improved car-following and lane-changing models suitable for heterogeneous traffic flow, establishing an expected expression for the relative quantity ratio of different vehicle types. Fundamental diagram models for both states were derived, with analytical characterization of flow-density curve disparities between peak and off-peak periods. Stability criteria were theoretically established, and the influence of platoon size on stability was systematically analyzed. Finally, numerical simulations validated traffic flow stability under perturbations, with comparative analysis of the effect of velocity disturbance amplitude through traffic speeds and CAV penetration rates of interactions in both states. The results indicate that the separated-state reduces the critical CAV penetration rate by 12.7% on average compared to the mixed-state at identical speeds, exhibiting nonlinear stability growth with platoon expansion and enhancing peak-hour capacity by 14.7%. Mixed-state demonstrates superior disturbance resistance when CAV penetration rate falls below 0.2. CAV penetrations rate of 0.86, 0.71, and 0.47 at 15 m·s^-1, 25 m·s^-1, and 35 m·s^-1 respectively, with both states maintaining stability under perturbations. At high-speed conditions, the required CAV penetration rate difference between the two states narrows to 0.0236.

Adapting to stochastic traffic demand and optimizing the spatiotemporal allocation of intersection resources coordinately is a crucial strategy for enhancing traffic efficiency. This study addresses the temporal misalignment characteristics of motorized and non-motorized traffic and proposes a bi-level stochastic programming-based lane management approach to achieve dynamic coordination among lane type, lane marking, and signal timing. First, this study moves beyond the traditional paradigm of space optimization centered on motorized lanes by incorporating non-motorized lanes into a unified lane allocation system. Second, considering the fluctuating spatiotemporal demands of mixed traffic flows, a lane-reuse strategy is introduced, allowing non- motorized vehicles to temporarily utilize certain motorized lane spaces during certain periods. Furthermore, an integrated optimization model for lane markings and signal timing is established to achieve the coordinated optimization of spatiotemporal resources. Numerical simulation results indicate that, compared to traditional static lane management models, the proposed method enhances intersection capacity by 9.90%. Moreover, when compared to non-reuse lane management strategies, capacity gains reach 3.27%. Sensitivity analysis reveals a significant positive correlation between the proportion of electric bicycle traffic, the speed heterogeneity of non-motorized vehicles, and the efficiency of spatiotemporal resource optimization. Every 0.5 m expansion in road width contributes to a 6% marginal improvement in intersection capacity. The theoretical framework proposed in this study, which integrates flexible resource allocation and dynamic marking adaptation, effectively accommodates the heterogeneous distribution and stochastic fluctuations of mixed traffic demand, significantly improving the utilization efficiency of spatiotemporal resources at urban intersections.

Pedestrian crossings at unsignalized midblock crosswalks are traffic facilities where conflicts between pedestrians and vehicles directly occur. As pedestrians are vulnerable in urban traffic, and their right of way must be protected. Consequently, regulatory authorities have implemented policies and laws to enforce driver compliance at such crosswalks, with violations subject to penalties. However, on certain unmonitored roads, driver adherence to yielding regulations remains unsatisfactory. To investigate driver yielding behavior, this study collected vehicle, pedestrian, and traffic environmental data that may influence drivers' yielding behavior from three unsignalized midblock crosswalks in Changchun, Jilin Province. The factors were analyzed for the influence on driver yielding decisions. Binary Logistic regression models were developed to predict yielding probabilities, followed by sensitivity analyses. The findings indicate that vehicles queuing at upstream stop lines have a negative impact on yielding behavior in most segments, whereas a red light at downstream signals significantly promotes drivers' yielding behavior (p <0.05) . The regression analysis demonstrate that multiple factors significantly correlate with yielding behavior, including time headway, number of crossing pedestrians, pedestrian kinematic cues, pedestrian position, presence of electronic police monitoring, number of lanes, upstream queuing conditions, downstream signal light color (p<0.01) , and pedestrian direction (p<0.05) . Additionally, chi-square tests revealed behavioral differences: commercial vehicle drivers at crosswalks exhibited varying yielding rates depending on e-police presence, and drivers at monitored crosswalks responded differently to pedestrians crossing from the middle versus the curb. Based on these findings, the study also proposes suggestions for enhancing pedestrian safety at unsignalized midblock crosswalks.

The narrow aisle area is a critical bottleneck for occupant evacuation from transportation scenarios such as buses in case of emergencies. However, in such a scenario, the single-file movement characteristics of the crowd under different motivation levels are still unclear yet, with limited research available in literature. To address this gap, well-controlled single-file pedestrian evacuation movement experiments in the narrow aisle area under low and high levels of motivation are designed and conducted. Macroscopic and microscopic metrics such as movement trajectories, instantaneous velocities and evacuation times are collected, and the velocity-density and headway-velocity relationships are constructed. Then, systematic and quantitative analyses are performed to explore the effect of the motivation level on the microscopic movement behaviour and the macroscopic evacuation efficiency, as well as the underlying mechanisms. The experimental results show that the single-file pedestrian evacuation movement conforms to a three-stage pattern, i.e., free, congested and recovery stages. Higher level of motivation increases the individual movement velocity, but has a two-sided effect on the evacuation efficiency with a critical density of 1.05 m^-1, and its effect vanishes when the density exceeds 1.65 m^-1. Three regimes, i.e., strongly constrained, weakly constrained and free regimes, are distinguished in the headway-velocity relation. Moreover, it is found that under high motivation, a lower headway threshold exists between the strongly and weakly constrained regimes, and the movement velocity becomes more sensitive to the change of headway in the strongly constrained regime. These findings are helpful to enhance the understanding of the crowd evacuation movement characteristics under different degrees of motivation, and provide valuable references for the development of related regulations and evacuation management strategies, and safety design of relevant facilities.

This study aims to investigate the dynamic evolution law of driver's driving risk in underground roads under different traffic sign information densities. Using the driver's visual load as a representative metric, eye movement data of drivers were collected through natural driving tests. A driving risk evaluation model was constructed based on dynamic Bayesian network theory. The traffic sign information density model was constructed based on Shannon entropy to quantify the information volume of traffic signs, considering the presentation rate of traffic sign information. Four underground roads with different traffic sign information densities in Liangma Square were selected for the in-vehicle experiment, and the driver's visual characteristics indicators were extracted and analyzed. The driver's driving risk was dynamically predicted and inferred by the dynamic Bayesian network theory, and the key risk factors affecting the driver's driving risk were obtained through diagnosis inference, sensitivity analysis and impact chain analysis. The results show that the driver's fixation duration, horizontal/vertical saccadic amplitude, horizontal/vertical saccadic speed, and pupil area change rate are positively correlated with traffic sign information density, while the blinking frequency is negatively correlated; the driver's driving risk occurrence probability changes dynamically over time, first increasing and then becoming stable, and the risk probability of the four sections stabilizes at 22.6%, 35.7%, 40.1%, and 43.8% respectively as the traffic sign information density increases; the driver's driving risk is greatly affected by the link risk factors in the linkage process, including the key risk factors of fixation duration, pupil area change rate, and blinking frequency.

To investigate the correlation between road traffic violations and traffic accidents, this paper proposes a spatiotemporal constraint-based approach to analyze the contributory characteristics of various violations in triggering traffic accidents. Utilizing traffic accident data from Beijing city in 2023 and 2024, in conjunction with violation records from the electronic enforcement system, this study analyzed 2338 traffic violations associated with accidents. This data-driven approach eliminates the subjective biases inherent in traditional reports. Furthermore, by applying the Frequent Pattern Growth (FP-growth) algorithm, the study identified 18 strong association rules linking five types of traffic accidents to four categories of traffic violations. The results indicate that the spatial distribution of correlated data between traffic accidents and violations is relatively uniform. The incidents are primarily concentrated between 7:30 AM and 10:30 PM, peaking during the morning and evening rush hours. Collisions between motor vehicles and non-motor vehicles primarily occur in educational and residential areas, with illegal parking behaviors significantly influencing the occurrence of the accidents. These collisions exhibit a confidence level of 0.495 and a lift value of 2.578. Single-vehicle accidents are predominantly associated with illegal parking behaviors, exhibiting a lift value as high as 8.696. This highlights the substantial threat that illegal parking poses to traffic order and safety. The derived association rules offer decision support for precision law enforcement, intelligent signal control, and road network optimization. The study results might also provide reference for urban traffic management in other cities, to improve traffic safety.

In urban rail transit systems, traction power supply faults often lead to the unidirectional power supply shortage in a local area, which seriously limits the total traction power and line capacity. Therefore, this paper proposes a train regulation approach based on dynamic power. Considering the instantaneous train power under different driving regimes, a 0-1 integer linear programming model is constructed to maximize the line capacity under the constraints of power supply capacity. Then, an adaptive large neighborhood search algorithm, based on the principle of "peak shaving and valley filling", is designed to improve the efficiency of traction power and to reduce the fault impacts. Finally, numerical examples based on the real data of Beijing Metro Line 7 are constructed to verify the effectiveness of the proposed approach. According to the simulation results, the computational time in all scenarios is controlled within 37.3 seconds, which satisfies the real-time requirement. After considering the dynamic power, the utilization rate of traction power becomes higher, which leads to the balance between power supply and demand, and increases the maintained line capacity by 59.3%~97.8% compared with the traditional approach based on maximum traction. For the performance of the designed algorithm, the gap from the optimal solution is less than 2.8%. Meanwhile, the solution quality is superior to the traditional neighborhood search algorithm.

To clarify the tracking operation performance of urban rail train groups with virtual coupling effectively, a multi-state operation simulation model for urban rail train groups with virtual coupling was proposed. By analyzing the operational control principles of virtual coupling trains, a method for calculating the safety protection distance was presented, which integrated communication and control delays. A multi-state operation simulation model for urban rail train groups with virtual coupling was constructed based on a train dynamics model and incorporated the operational state transition process of trains with virtual coupling. Using Jinan Metro Line 3 as a case study, simulation experiments were conducted from various perspectives, including different grouping modes, formation types, and delay times. The results show that in a train group composed of two trains, the average distance interval of trains with virtual coupling was only 21.93% of that with fixed coupling. In a train group composed of four trains, the average formation length and operational speed difference of fixed formations were 1.26 times and 1.94 times of that in mixed formations, and 1.45 times and 4.17 times of that in virtual formations, respectively. Additionally, the arrival time deviation of train groups with a virtual formation type were only 46.29% and 74.56% of that with fixed formation and mixed formation train groups, respectively. Therefore, the trains with virtual coupling effectively reduced the operational distance intervals, enhanced the accuracy of speed tracking, and decreased the arrival time deviation. Furthermore, when the control delay and the communication delay time of the trains with virtual coupling were both 0.5 second, the maximum distance deviation between trains can reach 10.83 meters and 0.83 meter, respectively. The negative impact of the control delay on the tracking operation of the train groups was greater than that of the communication delay. The research results provided a theoretical support for the operational decision-making and scheme formulation of urban rail train groups with virtual coupling.

To organize the operation of newly added high-speed express trains and meet the growing demand in the express goods transportation market safely and efficiently, this study investigates a method of adding high-speed express trains to the existing train timetable. Express trains must not disrupt existing passenger train schedules. Compared to passenger trains, express trains have more flexibility in selecting their routes. Given these traits, this study proposes an express train addition model in the network using the K-shortest path algorithm to generate alternative routes for each express train, without changing the passenger train schedules. The model aims to minimize the total travel time of all express trains by considering constraints related to train routing and scheduling. After being transformed into an integer linear programming form, the model is solved by Gurobi. The study verifies the proposed model using data from selected railway lines among Chengdu, Chongqing, and Guiyang. The result shows that compared to the model in a fixed district, the proposed model can increase the number of added express trains and reduce the travel time of all express trains by up to 35.50%, which indicates the effectiveness of the proposed method in improving the efficiency of express train operation. Moreover, as the scale of alternative routes for each express train expands, the number of successfully added express trains further increases, and the improvement in transport efficiency becomes more pronounced. The study also adopts a "warm start" acceleration strategy, which can enhance solution efficiency by accelerating the convergence rate of the upper and lower bounds in the model.

In the context of railway lines with long and steep gradients, the train operation process involves traction and braking stages that last for a relatively long time. By fully considering the overlap of train working conditions between the traction stage and the braking stage within the same power supply section, favorable conditions can be provided for the energy-saving operation of trains. Firstly, based on the characteristics of long and steep gradient lines, the impact of train arrival and departure times on the overlap of working conditions is analyzed. Secondly, to avoid power supply conflicts, assuming that the start and end times of maintenance windows in the same power supply section are synchronized, a mixed-integer programming model is constructed, incorporating the constraints of train operation and maintenance window opening comprehensively. The objective of model is to maximize the train working condition overlap time while minimizing the total travel time. Based on the characteristics of the model, the ε-constraint method and GUROBI are used for solution. The effectiveness of the model and algorithm is verified through a case study. The results show that the algorithm can effectively solve the problem, increasing the total overlap time of train working conditions by 24.39% compared to pre-optimization. The research results can provide a reference for the transportation organization decision-making on long and steep gradient lines.

To investigate intelligent methods for determining the liability of both parties involved in road traffic accidents, the liability outcomes of the involved parties were categorized into five types, full liability-no liability, primary liability-secondary liability, equal liability-equal liability, secondary liability-primary liability, and no liability-full liability. Based on feature analysis and feature engineering to screen training sample characteristics, accident features were thoroughly discretized and one-hot encoded. A deep learning ensemble model for intelligent liability determination in traffic accident was trained by integrating linear models and deep neural networks. The intelligent liability determination model under complex network architectures was established through network structure and hyperparameters. The feasibility of the proposed model was cross-validated using historical traffic accident data. The results show that the deep learning ensemble method significantly elevates the accuracy of traffic accident liability determination. For the full liability-no liability and no liability-full liability accident types, the accuracy of results from the proposed model reaches 0.91, while the accuracies of liability determination results on other accident types are 0.78. The ensemble approach elevates accuracy by over 30% and more than 8% respectively compared to traditional machine learning methods and a single fully connected neural network. The deep learning ensemble method demonstrates notable enhancements in defining liability boundaries for intelligent traffic accident adjudication.

Passenger boarding is a key part of the airplane turn-around process, and an efficient boarding process can help reducing operation costs. Based on the traditional boarding strategy, this paper introduces a Slow First boarding (SF) principle, for example, giving priority to those large luggage carriers, and proposes some improved or new strategies. Then, slow first cellular automaton boarding model is developed to depict the conflicts among passengers in the cabin and simulate the boarding process of passengers in a narrow-body aircraft. The simulation results show that the introduction of the SF in most scenarios can reduce the boarding time and improve the boarding efficiency, but the priority setting when combining SF with the traditional boarding strategy to divide the boarding order of passengers will affect the boarding efficiency of passengers in varying degrees. The BF3-SF (Back-to Front-Slow First) strategy and SF-WA3 (Slow First-Window-to-Aisle) strategy cannot improve the boarding efficiency. Different strategies have different sensitivities to the passenger release interval and the proportion of large luggage. This paper also analyzes the influence of flight load factor on the stability of boarding time of each strategy. Among the Slow First boarding strategies, BF3 SF, SF-Random, and WA3-SF strategies can reduce the boarding time, and the WA3-SF strategy requires the shortest boarding time. In general, the improved WA3-SF strategy is the optimal strategy.

This paper focuses on optimizing the secondary allocation of arrival and departure slots in the context of airlines autonomously canceling flights. A bi-objective optimization model is developed to minimize both the total delay costs of airlines and the total delay time of passengers, effectively addressing the needs of both stakeholders. A parameter λ is introduced to balance the discrepancies in the delay cost functions for arrival and departure flights. Additionally, constraints are implemented to differentiate the turnaround time for flights based on the types of aircraft. According to the model's characteristics, real number coding is utilized, and the gene position corresponding to the canceled flight is represented by-1. Furthermore, a learning mechanism is integrated into the genetic operations. Distinct crossover and mutation probabilities are established for both dominated and non-dominated individuals, and a Q-learning-driven general variable neighborhood search strategy is implemented for elite individuals. The experimental results from three sets of examples indicate that the solving times of the improved algorithm were 29.34 s, 58.61 s, and 125.21 s, resulting in 9, 8, and 8 optimal schedules, respectively. In comparison to the first come first served (FCFS) method, the total delay costs of airlines were reduced by 14.85%, 8.47%, and 9.18%. Additionally, the total delay times of passengers decreased by 1.03%, 5.31%, and 4.68%. Compared with the non-cancellation strategy, the total delay costs of airlines under the cancellation strategy decreased by 7.04%, 9.38% and 11.96%, and the total delay time of passengers increased by 0.95%, 1.21% and 1.70%, respectively. The bi-objective optimization model developed in this paper, along with the proposed enhanced algorithm, effectively reduces airlines' delay costs while considering passenger interests. This approach offers valuable decision-making support for the slot secondary allocation of arrival and departure flights under autonomous cancellations.

Flood disaster is characterized by their sudden onset, and emergency response agencies often face a shortage of relief supplies in the early stages. The severity of the flood and the affected populations vary across different disaster sites, leading to diverse demands for relief materials and differing levels of urgency. Initially, the entropy weight method is employed to evaluate the risk levels of each disaster point. A relief allocation model is formulated to minimize the total suffering perception across all disaster points. A collaborative distribution routes planning model for trucks and speedboats is developed, with the dual goals of minimizing the total delivery time and reducing the sum of suffering perception costs at all disaster points. To solve these models, ant colony optimization and an improved non-dominated sorting genetic algorithm-II are designed, and tailored to the specific characteristics of each model. Based on the data of the flood in Pingjiang County, Hunan Province in 2024, the relief allocation distribution scheme was solved under the different disaster point scenarios. Compared with the equal-proportional allocation strategy, the allocation method considering the difference of disaster sites can reduce the total perceived pain of all disaster sites by 72.24%, which is more reasonable. Compared with the multi-objective artificial bee colony algorithm, the improved non dominated sorting genetic algorithm-II planned the distribution route of emergency materials according to the risk level of disaster points, resulting in an average reduction of 3.96% in total delivery time and 21.78% in overall perceived suffering cost, thereby effectively enhancing flood rescue effectiveness.

To explore the heterogeneity and transferability of factors affecting the severity of traffic accidents under different lighting conditions, this paper gave a transferability analysis based on a random parameters Logit model with heterogeneity in means and variances. The severity of accidents was classified into three categories, and the study integrated the characteristics of drivers, road, environment, and vehicles, including 25 potential explanatory variables. By capturing the mean and variance changes of random parameters, the model fully explored the heterogeneity characteristics in accident data, and used log likelihood ratio to verify the temporal instability and transferability of influencing factors under different lighting conditions. Based on the marginal effect estimation results, a quantitative analysis was conducted on the differentiation of statistically significant explanatory variables in the severity of accident injuries. The research results indicate that the factors affecting the severity of accidents exhibit significant heterogeneity under different lighting conditions and show significant non transferability. The turning behavior of vehicles and the occurrence of accidents outside the road were only significant in the traffic accident model during daytime in 2017, while the accidents at four-way intersections was only significant during nighttime in 2019. The accidents of male drivers were significantly higher during daytime in 2018 and 2019, with opposite effects on the extent of property damage and minor injury accidents. The accident due to the negligence of drivers in operation were significant during both daytime and nighttime unlit accident models without lightning in 2018, and its impact on minor injury accidents showed the opposite effect.

To accurately identify and assess the traffic risk of urban road sections connected to freeway exits, this paper proposes a lane-changing spatial-temporal risk index TRCI (Transitive Risk Coupling Index) that integrates risk proximity level (RNL) and risk severity level (RSL). First, vehicle trajectory data are used to analyze the trajectory characteristics of lane-changing vehicles on road segments, lane-changing positional properties and lane-changing influencing factors. The results show that the relative position of vehicle lane-changing is concentrated in the intervals of 0.1 to 0.2 and 0.4 to 0.6. The lane-changing collision time (LCTTC) is introduced and integrated to the MTC(Margin to Collision) to get the RNL and RSL to characterize the spatial temporal risk characteristics of lane-changing. The Deng's grey correlation method is used to obtain the spatial-temporal risk index of lane-changing (TRCI). The cumulative frequency method is used to determine the thresholds for serious, serious, general, and minor conflicts, respectively. The cumulative frequency method is used to determine the thresholds of serious, more serious, general and minor conflicts as 0.30, 0.51 and 0.67 respectively, and comparing TRCI with TTC and GAP for the identification of the risk of lane-changing conflicts. The freeway exits connecting with the urban road sections are divided into 32 segments and 128 blocks, and lane-changing conflicts are analyzed in terms of their distributional characteristics and risk level assessment. The results show that the effective identification rate of the risk of vehicle lane-changing spatio-temporal risk index TRCI is improved by 85.10% on average compared with that of TTC, and 49.75% on average compared with that of GAP. The performance of the vehicle lane change conflict severity prediction model constructed based on the XGBoost algorithm for freeway exits bridging urban roadway segments performs better than traditional methods. The F1 scores are improved by respectively 11.28% and 1.40% compared with the GBDT(Gradient Boosting Decision Tree) and RF(Random Forest) models. The conflict points in the study section are most intensive in the sections of 3 to 5, 14 to 18 and 21 to 23. The sections of 4 to 5 and 15 to 18 are in high-risk status, which indicates that the exit ramps, guided lanes and surface roadway merging points are the areas with the highest risk of lane change conflicts. The study provides theoretical support for the analysis and management of traffic safety in the urban road section connected with the freeway exit.

To clarify the benefits and risks after opening of hard shoulders on expressway, using drones to collect videos of traffic flow on the Lanhai Expressways in Chongqing, through AI.Datafromsky platform to obtain vehicle speed, lateral acceleration, and longitudinal acceleration data. Then study the speed and acceleration distribution characteristics of each lane under the open and closed states of the hard shoulders, as well as the speed acceleration correlation. It clarifies the real operating status and driving risk characteristics of each lane under different working conditions of the hard shoulders. The research results indicate that after the opening of hard shoulders, the operating speed of each lane increases, and the speed of hard shoulders is significantly higher than that of fast lane and slow lane, resulting in an overall increase in expressway traffic capacity. Low speed vehicles tend to choose the inner lane for driving, while high-speed vehicles tend to choose the outer lane for driving. The speed difference between vehicles in each lane is reduced, which enhanced the smoothness of traffic flow on the expressway. The degree of left swing of vehicles in the hard shoulder is significantly stronger than that of right swing, and the risk of lateral collision between the hard shoulders and slow lane is higher than other lanes. The deceleration behavior of hard shoulder traffic is stronger than that of fast lane and slow lane, and the sensitivity of vehicles with different speeds in the hard shoulders to longitudinal acceleration changes is consistent. There are significant differences in the selection of the mean longitudinal deceleration of each lane. The longitudinal smoothness of vehicles in slow lane is weakened, and the longitudinal smoothness of vehicles in the hard shoulder is weaker than that in fast lane and slow lane.

The turning behavior of pedestrians in a traffic environment is an important component of pedestrian traffic flow research. This paper takes the pedestrian flow in an L-shaped one-way passage as the research object and investigates the movement mechanism of turning pedestrians based on field experiments and simulation studies. First, according to the movement characteristics of pedestrians in different areas during turning, the L-shaped passage is divided into four zones. Considering the movement characteristics of pedestrians under different widths and service levels, the self-propulsion direction and movement rules of the existing model are improved, and an improved right-angle turning model for the L-shaped passage is established. Then, the pedestrian turning trajectories are collected in the right-angle turning areas of subway stations and simulated venues. The path selection, lane-changing positions, and turning positions are analyzed, and the parameters of the improved right-angle turning model are calibrated by the analysis on which a pedestrian right-angle turning simulation platform is built. Finally, by comparing the simulation results with experimental data from 100 right-turning pedestrian trajectories in L-shaped right-angle passages, the similarity between the experimental trajectory space and the simulated trajectory space is 89.4%, which verifies the accuracy and effectiveness of the improved model.