Archive

25 October 2026, Volume 26 Issue 5 Previous Issue   

---

For Selected:

Download Citations EndNote Ris BibTeX

Toggle Thumbnails

Select

Review of EEG-Based Driving Research: State Perception, Cognitive Decoding, and Human-Machine Interaction and Shared Driving QI Geqi, DONG Zhentao, LUAN Zhengkai, GUAN Wei 2026, 26(5): 1-21.  DOI: 10.16097/j.cnki.1009-6744.2026.05.001 Abstract ( )   PDF (4779KB) ( )   Electroencephalography (EEG) provides a non-invasive means of characterizing drivers' attention, fatigue, workload, emotion, risk perception, and action preparation during continuous traffic tasks, providing information for analyzing the relationship between observable driving behavior and internal cognitive processes. This study adopted systematic review method to search the English-and Chinese-language literature from the Web of Science Core Collection and China National Knowledge Infrastructure up to April 20, 2026. Ninety-seven articles were included. This paper adopts an overarching framework of perceiving the human, understanding the human, and collaborating with the human to systematically review research on driver state monitoring, behavior and intention decoding, autonomous driving takeover, interaction trust, and brain-computer collaborative control. It further provides a comparative analysis of the experimental paradigms, labeling schemes, EEG representations, validation protocols, and application contexts across these research domains. In specific tasks such as emergency braking and steering, EEG studies have detected the neural activity related to action preparation and intention before pedal or steering-wheel movements occur. Related methodologies have evolved from handcrafted feature analysis to spatiotemporal modeling, brain network analysis, cross-subject generalization, and multimodal fusion. However, the substantial differences persist across studies with respect to task definitions, sample sizes, channel configurations, and validation protocols limit direct comparison. The majority of existing work remains confined to the offline-recognition stage, and EEG-derived information has not yet been reliably integrated into takeover decision-making, control authority allocation, and interaction strategy adaptation. Future research may focus on trustworthy perception and unified representations, the closed loops of cognition-decision-control, multi-agent and mixed traffic collaboration, as well as closed-loop validation in naturalistic driving and engineering deployment. This review summarizes the available evidence, comparability boundaries, and evaluation priorities pertaining to the evolution of EEG-based driving research from state recognition toward human-machine collaboration. References | Related Articles | Metrics

Select

Review on Emergency Evacuation Behavior of Minors YAO Yishu, ZHANG Xinhai, LU Wei, CAI Jun 2026, 26(5): 22-33.  DOI: 10.16097/j.cnki.1009-6744.2026.05.002 Abstract ( )   PDF (2486KB) ( )   The physiological development and cognitive abilities of minors show pronounced age-related characteristics. To systematically review the research progress, behavioral mechanisms, and future directions which concern the evacuation behavior among minors in different age groups, this study retrieved the literature from core databases including CNKI, Web of Science, and Scopus. Following the PRISMA procedure, 272 publications were selected, and a review was conducted from three perspectives: research evolution, research methods, and age-specific behavioral characteristics. The results show that the research on the evacuation of minors has gradually shifted from early macro-level assessments, such as total evacuation time and exit capacity, toward the mechanisms of pre-movement behavior, microscopic pedestrian-flow parameters, route decision-making, and the effects of adult accompaniment. There is a clear difference between Chinese and international research orientations that the Chinese studies tend to emphasize the empirical measurements and practical applications in places where minors congregate, whereas the international studies place greater emphasis on the accumulation of basic data and the refined calibration of model parameters. The existing studies have formed a methodological system with functionally complementary: field investigations and questionnaire surveys support the analysis of cognitive patterns; evacuation drills enable the behavioral observation in realistic scenarios; controlled experiments extract key movement parameters; and computer simulations support multi-scenario evacuation assessment. Age-specific studies reveal the behavioral heterogeneity among minors. The evacuation of infants and preschool children（(0, 6) years old）is highly dependent on adult care and assisted transfer, and modeling approaches need to shift from treating them as independent low-speed individuals to representing care-dependent coupled movement units. School-aged children（[6, 12) years old）have basic rule comprehension and autonomous movement ability, but their route judgment and queue coordination remain constrained by multiple external factors, making them limited autonomous actors under rule-based constraints. Adolescents （[12, 18) years old）have movement abilities close to those of adults, but their walking speed, bottleneck flow rate, and spatial behavior are still regulated by crowd density and other factors; therefore, adult parameter systems cannot be directly applied. Overall, the existing studies have preliminarily established empirical parameter datasets for child evacuation and heterogeneous behavior modeling frameworks. However, research gaps remain in the precision of age stratification, adult-child coupled movement units, and scenario expansion. Future research should broaden the research scope and develop a continuous evacuation analysis framework integrating evacuation, assembly, transportation, and guardian handover. References | Related Articles | Metrics

Select

Influencing Mechanism of Spatial Heterogeneity in Usage Intensity of Urban Public Fast-Charging Stations Considering Built Environment SHAN Xiaonian, WANG Zikang, GONG Chengming, LI Zhibin 2026, 26(5): 34-45.  DOI: 10.16097/j.cnki.1009-6744.2026.05.003 Abstract ( )   PDF (2668KB) ( )   With the continued surge in electric vehicle ownership and the increasing demand for fast charging, unraveling the spatial heterogeneity in the usage intensity of public fast-charging station is of great significance for optimizing the layout of urban charging infrastructure. This paper aims to investigate the spatial heterogeneity and underlying driving mechanisms in the usage intensity of public fast-charging station under weekday and non-working day scenarios. Using the actual operation data from 157 public fast-charging stations in Nanjing, this study constructs a built environment indicator system from six dimensions: density, diversity, design, public transit proximity, destination accessibility, and socioeconomic characteristics. The multiscale geographically weighted regression model is employed to identify the key influencing factors of public fast-charging station and to decipher their spatially heterogeneous mechanisms. Furthermore, a sensitivity analysis on the buffer radius is conducted to compare the model fitting performance and the effects of variables under different radius scales. The findings reveal that the proximity to subway stations and railway stations exerts a positive effect on the usage intensity of public fast-charging station, while office areas show a negative effect. Road network density is predominantly negatively associated with the usage intensity, with stronger significance on non-working days. The number of bus stops exhibits a significant positive association only on weekdays. The sensitivity analysis results indicate that 1.0 kilometer buffer radius effectively captures the relationship between the utilization intensity of public fast-charging stations and the surrounding built environment. The results uncover the built environment driving mechanisms behind the spatial heterogeneity in the usage intensity of public fast-charging station, which can provide scientific support for differentiated layout and refined the facility planning of urban public fast-charging stations. References | Related Articles | Metrics

Select

Vulnerability and Uncertainty of Urban Metro Networks Under Multiple Risks DUAN Jinxiao, LIU Yanan, WU Ke'er, WU Yidan, ZENG Guanwen, WANG Yun, YAN Xuedong 2026, 26(5): 46-55.  DOI: 10.16097/j.cnki.1009-6744.2026.05.004 Abstract ( )   PDF (2531KB) ( )   Comprehensively analyzing the vulnerability and its uncertainty of metro networks under different disturbances, as well as identifying their influencing factors is the key to improve the operational reliability. Considering the typical risks of metro systems, this paper proposes a percolation theory-based method to quantify the vulnerability and uncertainty. This method defines the risk-resistance ability and performance-loss degree of metro systems, as well as the uncertainty of these two vulnerability indicators. Taking the metro networks of 40 cities in China as research object, this study simulates the network percolation process under random attacks, degree attacks, and betweenness attacks, and analyzes the influence of network layout, structural redundancy, and structural bottlenecks on metro network vulnerability. The results show that the average clustering coefficient and the bridge count have consistent effects on risk-resistance ability and performance-loss degree, while the influence of connection density on them is opposite. The influencing factors for the uncertainty of risk-resistance ability are diverse, while those for uncertainty of performance-loss degree are simple. The comprehensive analysis of the nonlinear factors on network vulnerability and its uncertainty, reveals that connection density, network bottleneck cuts, and the number of lines are the most key factors. This study provides a theoretical support and practical guidance for the design of resilient metro networks. References | Related Articles | Metrics

Select

Model and Algorithm for Vehicle Spatiotemporal Routing Optimization in Smart Recycling Bin Collection Under Time-Varying Disposal Demand YAO Yu, ZHU Changhao, WANG Yiran, GUO Menghuan 2026, 26(5): 56-67.  DOI: 10.16097/j.cnki.1009-6744.2026.05.005 Abstract ( )   PDF (2832KB) ( )   Focusing on the problem of smart recycling bin collection, this paper proposes a vehicle spatiotemporal routing optimization method for the collection problem under time-varying disposal demand. The problem is characterized by the dynamic evolution of bin occupancy, the non-predetermined number of services for each bin, and the strong coupling among service timing, actual collected amount, and bin occupancy state. To address these features, each recycling bin is expanded into multiple candidate emptying service events, and a fine-grained discrete-time mixed-integer linear programming model is developed to minimize the sum of operating cost, in-process rejection penalty, and end-of-horizon residual occupancy penalty. Based on the problem structure, an adaptive large neighborhood search algorithm embedded with an alternating iterative strategy is designed, forming an iterative solution framework consisting of candidate service event generation and updating, route search, and spatiotemporal route generation and adjustment. The effectiveness of the model and algorithm is verified through small-scale comparative instances, sensitivity analysis, and a real-world case study. The results show that the proposed method can generate the high-quality smart recycling bin collection plans, with an average relative gap of only 0.59% from the exact solutions. The penalty coefficient determines the trade-off between service level and operating cost, while the spatiotemporal characteristics of demand affect the relative effectiveness of two resource allocation strategies, namely increasing the number and capacity of vehicles. The real-world case study further verifies the applicability of the proposed method in an actual smart recycling network. References | Related Articles | Metrics

Select

Dynamic Joint Pickup Scheduling of Demand-Responsive Freight Transit and Crowdsourced Riders JIANG Xiaohong, HUA Jingwen, WANG Duo, LI Jiawei, LI Zhongkai 2026, 26(5): 68-78.  DOI: 10.16097/j.cnki.1009-6744.2026.05.006 Abstract ( )   PDF (2047KB) ( )   To extend the service scope of instant retail in rural areas and improve timeliness, this paper proposes a collaborative pickup strategy which integrates demand-responsive freight buses and crowdsourced couriers. Orders are uniformly delivered to the urban-rural bus passenger-freight-mail integration terminals, where the urban-rural buses perform the last-mile delivery to achieve half-day delivery. First, an offset distance threshold is set based on the spatial distribution of orders to assign scattered orders to couriers, while the freight buses focus on high-density areas to reduce detours. Second, an order transshipment consolidation strategy is introduced, using fixed stops as transfer nodes to consolidate multi-vehicle tasks into a single vehicle, thereby avoiding empty backhauls. Third, a two-stage scheduling model is formulated with the objective of minimizing total pickup cost, and is solved by a genetic algorithm with variable neighborhood search (GA-VNS). Finally, a case study is conducted in the Dongshan area of Jiangning District, Nanjing, with small, medium, and large order scales. The results show that under the large-scale orders, the collaborative pickup mode reduces the total cost by 7.6% and travel distance by 12.2% compared with the single pickup mode, demonstrating significant effectiveness. The transshipment consolidation strategy is particularly suitable for the early stages of rural instant retail development: under the small-scale orders, it increases the load factor by 32% and reduces empty backhaul trips. The adaptive adjustment of delivery time windows and batching intervals can effectively balance the timeliness and operational costs. References | Related Articles | Metrics

Select

Cooperative Vehicle Group Control for Mitigating Multi-cluster Traffic Waves in Connected and Automated Vehicles Environments YANG Jingwen, QI Xudong, WANG Ping, XIA Xinhai 2026, 26(5): 79-88.  DOI: 10.16097/j.cnki.1009-6744.2026.05.007 Abstract ( )   PDF (2746KB) ( )   To address the difficulties of cooperatively regulating multi-cluster traffic waves in a freeway corridor, as well as the limitations of existing methods in cross-region coordination and fine-grained control, this paper proposes a distributed cooperative vehicle-group control method based on deep reinforcement learning. First, the corridor is divided into some units and groups according to the merge-area structure and control objects. A scalable modular architecture is then constructed to enable unified modeling of the corridor traffic system, which transforms the dissipation of multi-cluster traffic waves into a multi-agent cooperative decision-making problem. Second, considering the dynamic influence of traffic-flow interactions on bottleneck capacity, the state space and reward function is designed, and a lane-level speed control model for vehicle groups is developed. Finally, a distributed control policy is trained based on the multi-agent proximal policy optimization algorithm. Lane-level cooperative regulation is achieved through joint decision-making of multi-lane speed limits, which improves the overall operational performance of the multi-bottleneck traffic system. Simulation results show that the proposed method reduces the number of stops, average waiting time, and potential collisions by 49.88%, 80.65% and 26.84% respectively under a control interval of 30 seconds. By refining the control granularity from the road-segment level to the lane level and extending the control scope from a single bottleneck to multiple bottlenecks, the proposed method realizes the cross-level collaborative optimization of traffic system. References | Related Articles | Metrics

Select

Research Progress of Unmanned Aerial Vehicle in Road Traffic Operation and Management FAN Bosong, ZHAO Dan, NIU Xuejun, SHAO Chunfu 2026, 26(5): 89-111.  DOI: 10.16097/j.cnki.1009-6744.2026.05.008 Abstract ( )   PDF (1950KB) ( )   With the rapid development of the low-altitude economy and artificial intelligence technologies, Unmanned Aerial Vehicle (UAV) have gradually evolved from auxiliary sensing devices into a key technological pillar of intelligent traffic management systems, demonstrating systemic value in enhancing traffic order, ensuring traffic safety, and improving traffic operation smoothness. This paper proposes an analytical framework for three typical application scenarios—"orderly, safe, and smooth"—and provides a systematic review of recent research on UAV-based road traffic operation and management. The results indicate that research in this field is evolving from "single-modal perception and localized applications" toward "multi-modal collaboration and system integration." In traffic order scenarios, the research focus has shifted from early rule-based approaches and manual interpretation to methods centered on deep learning, such as object detection, integrated detection-tracking, and trajectory analysis. UAVs demonstrate significant advantages in the coverage and accuracy of traffic element identification; however, their reliance on high-quality data and ideal observation conditions limits their cross-scenario application. In traffic safety scenarios, multi-sensor fusion, abnormal behavior recognition, multi-UAV collaborative response, and 3D scene reconstruction have become research hotspots. The unique value of UAVs in accident identification, emergency support, and risk assessment has been validated; however, existing research lacks an in-depth evaluation of system-level collaborative effectiveness within the complete emergency management chain. In traffic smooth management scenarios, UAV applications have gradually expanded from localized congestion monitoring to macro-level traffic situation awareness and three-dimensional (3D) reconstruction of accident sites. Multi-UAV coordination and air-ground integration have emerged as key directions for enhancing congestion management capabilities; however, this places higher demands on communication infrastructure, airspace management, and operational mechanisms. Building on this, this paper summarizes a common evaluation framework and data foundation for UAV applications across three types of road traffic operation and management scenarios, and outlines future development trends on technological generalization capabilities, multi-agent coordination mechanisms, and institutionalized operational systems. References | Related Articles | Metrics

Select

Green Shipping Route Design Considering Multiple Climate Policies and Stochastic Demand ZHANG Yan, LING Shiqi, ZHENG Jianfeng 2026, 26(5): 112-121.  DOI: 10.16097/j.cnki.1009-6744.2026.05.009 Abstract ( )   PDF (1409KB) ( )   As the global carbon reduction targets continue to advance, the multiple climate policies such as emission control areas, carbon taxes, and green corridors are being rolled out rapidly. How to reduce environmental compliance costs while maintaining route profitability becomes a pressing challenge for liner shipping operators. This paper addresses this challenge by proposing a green route design problem. It integrates multiple climate policies and demand uncertainties, employing multiple vessel types for transshipment operations. A two-stage stochastic programming model is developed to jointly optimize the vessel deployment, port calling, sailing speed, and capacity allocation across shipping segments. Nonlinearities are neutralized via segment-specific speed optimization, and demand uncertainty is mitigated through a sample average approximation approach. To ensure solution tractability, a specialized Benders decomposition algorithm is engineered. Numerical experiments validate the effectiveness of the proposed model and algorithm. A case study based on the China-Europe route is conducted to derive optimal operational strategies under various climate policy scenarios. The results indicate that under the joint implementation of emission control areas (ECAs) and carbon taxes, the transshipment mode shifting from conventional fuel ships to dual-fuel vessels reduces carbon emissions by approximately 22% and boosts voyage profits by nearly 4%, compared to the current single-vessel direct routing mode. Furthermore, doubling the carbon tax rate makes the deployment of purely green-powered vessels more viable, achieving a carbon emission reduction of around 72%.These findings provide a scientific decision support for shipping companies in designing green transition strategies aligned with evolving climate policies. References | Related Articles | Metrics

Select

Traffic Emergency Event Extraction Method Based on Social Media Data CHEN Jian, HU Gongming, JIAN Hongbo, FU Zhiyan 2026, 26(5): 122-131.  DOI: 10.16097/j.cnki.1009-6744.2026.05.010 Abstract ( )   PDF (1915KB) ( )   With the continuous growth of motor vehicle ownership in China, urban road traffic incidents are exhibited the characteristics of high frequency and strong disturbance, which presents severe challenges to the operation monitoring and emergency management of road traffic. Traditional traffic incident monitoring systems have limitations in terms of real-time performance, spatial coverage, and information dimensionality, whereas social media texts offer fine-grained incident information from a public perception perspective. To address the issues such as high noise levels, non-standard semantic expressions, and vague spatiotemporal descriptions in social media texts on traffic incidents, this paper proposes a traffic incident extraction method that integrates deep semantic modeling and traffic geographic knowledge. First, a "Scrapy+Selenium" data collection framework is established, and acquisition rules for traffic incident texts are formulated to retrieve incident text data, followed by data preprocessing. Second, accurate identification of traffic incident text content is achieved based on the DeBERTa model with the disentangled attention mechanism and enhanced mask decoder. Finally, a geographic-knowledge-infused DeBERTa-BiLSTM-CRF named entity recognition (NER) model is constructed in combination with an established external entity database, which realizes the precise extraction of temporal and location information for traffic incidents. Experimental results demonstrate that the F1 score of the DeBERTa classification model reaches 0.926 5. For the DeBERTa-BiLSTM-CRF model, the F1 scores for locations and temporal entities are 0.859 5 and 0.869 3, respectively, both outperforming the baseline models. References | Related Articles | Metrics

Select

Coordinated Optimization of Mixed Passenger-Cargo Train Operation Plans for Urban Rail Transit Under Cross-Line Operation HUANG Junsheng, LUO Qin, WANGYi, YE Tian 2026, 26(5): 132-141.  DOI: 10.16097/j.cnki.1009-6744.2026.05.011 Abstract ( )   PDF (2068KB) ( )   The idle capacity in urban rail transit during off-peak hours offers the potential for passenger-freight mixed transportation. To address the mixed transportation under cross-line operation, this study takes the Y-shaped network as the research object and achieves a cross-line freight transport through trunk-branch line coordination. Building on this, an optimization framework is constructed that integrates the differences in time sensitivity among three cargo types: express, general cargo, and heavy cargo. With minimizing passenger and freight transportation times and operating costs as its two objectives, the model differentiates the departure frequencies for pure passenger trains and three types of mixed trains: passenger-express, passenger general cargo, and passenger-heavy cargo, and uses the ideal point method to convert the bi-objectives into a single objective. To address the nonlinear characteristics of the model, a PCCO-VNS (Passenger-Cargo Coordination Optimization based Variable Neighborhood Search) algorithm is designed, which employs hierarchical passenger-cargo optimization, adaptive neighborhood adjustment, and cargo-priority-oriented repair to improve the search efficiency and solution quality of the algorithm. Case results show that the optimized scheme reduces the comprehensive objective function by 10.00%, passenger waiting time by 11.44%, transfer time by 7.19%, and the platform dwell times for the three cargo types by 10.59%, 16.35%, and 17.79% respectively. Sensitivity analysis indicates that setting the passenger weight to 0.6 achieves a balance between cost control and service level. These findings provide a decision support for the planning and operation of passenger-freight mixed transportation systems under cross-line operation. References | Related Articles | Metrics

Select

General Optimization Method of Cross-Line Train Service Route Design for Urban Rail Transit Network with Construction Budget SONG Qifeng, QI Jianguo, ZHOU Housheng, HE Jianian, WANG Zhimei 2026, 26(5): 142-151.  DOI: 10.16097/j.cnki.1009-6744.2026.05.012 Abstract ( )   PDF (3291KB) ( )   To reduce the transfer time and provide high-quality transportation service, this paper proposes a general method for the cross-line train service design problem on the urban rail transit network. Based on the candidate cross-line service route pool, a mixed-integer linear programming model is formulated with the objectives to maximize the service quality for passengers and minimize the costs of reconstruction and operation. The model enables to optimize the specific reconstructed transfer stations, operated cross-line service routes and the corresponding train frequencies simultaneously, while comprehensively considering practical constraints such as the renovation conditions and construction costs of transfer stations, and the number of available rolling stock. Numerical experiments based on the Beijing metro network are implemented to validate the effectiveness of the proposed model. The results indicate that the proposed method can balance the relationship between the reconstruction costs and service quality, and can find a cross-line route service design scheme within a short time. Compared with the baseline plan, the cross-line operation strategy requires additional reconstruction costs and operating costs, and it increases the direct passenger volume by 10.06% and reduces the number of stranded passengers by 52.85%. In summary, operating cross-line service routes can effectively improve the direct service level and connectivity of the network, thereby significantly improving passenger service level. References | Related Articles | Metrics

Select

Spatio-Temporal Heterogeneity of Impact of Built Environment on Mobility Potential of Bike-Sharing YANG Hu, ZHANG Zhenzhao, ZHENG Guangming 2026, 26(5): 152-161.  DOI: 10.16097/j.cnki.1009-6744.2026.05.013 Abstract ( )   PDF (2510KB) ( )   Existing studies on the impact of built environment on bike-sharing commonly use the trip volume as the main indicator of travel demand, which may fail to capture the area-level attractiveness for bike-sharing trips and its spatial organization. Based on the bike-sharing order data from Shenzhen, this study constructs a bike-sharing mobility potential field and characterizes the built environment attributes from five dimensions: density, diversity, design, destination accessibility, and public transit proximity. A multiscale geographically weighted regression model is employed to analyze the spatiotemporal heterogeneity of the impacts of built environment on bike-sharing mobility potential. The results show that, compared with the model using the trip demand as the dependent variable, the mobility potential reveals more localized built environment effects. The bandwidths of non-motorized road density, distance to the business district center, and metro accessibility decrease from global or medium-to-large scales to local scales, which indicate that the formation of bike-sharing attractiveness exhibits stronger local spatial dependence. The distance to the business district center has an overall negative effect on the mobility potential, suggesting that commercial agglomeration strengthens the tendency of bike-sharing trips to cluster toward urban core areas. Population density has a stable positive effect on the mobility potential, with a stronger influence during the evening peak. The effect of metro accessibility on the mobility potential exhibits the spatial differentiation characterized by zonal clustering and node-oriented patterns, while the positive effect of non motorized road density on the bike-sharing attractiveness varies across local contexts. References | Related Articles | Metrics

Select

Optimization for Regional Time-Varying Cold Chain Distribution Path Considering Traffic Congestion Avoidance Strategy ZHANG Xiaohua, ZHONG Tingting, ZENG Tao, HUANG Zhaoxuan, ZHANG Wenfeng, WU Zhuokui, ZHANG Quan 2026, 26(5): 162-172.  DOI: 10.16097/j.cnki.1009-6744.2026.05.014 Abstract ( )   PDF (1872KB) ( )   Aiming at the problem that traffic congestion during peak traffic flow periods impairs the efficiency of cold chain distribution, this paper investigates the regional routing problem of time-varying cold chain vehicles with time windows and proposes a two-stage collaborative decision-making strategy for traffic congestion avoidance tailored to this issue. Synthesizing the considerations of vehicle fixed costs, transportation costs, cargo damage costs, refrigeration costs, carbon emission costs, and time window penalty costs, a cold chain vehicle routing optimization model is constructed with the objective of total cost minimization. Subsequently, an improved Adaptive Genetic Algorithm integrated with Traffic Congestion Avoidance (AGATCA) mechanisms is designed to solve multi-type test instances. Experimental results demonstrate that the proposed model exhibits relatively low computational complexity, and improves the convergence speed of the algorithm. It can enable efficient and rapid adaptation to customer distributions of varying scales and types. For small-scale problems, the AGATCA algorithm can achieve higher-quality solutions with faster convergence compared to the CPLEX solver. In terms of total cost and carbon emissions, the model achieves reductions of 8.4% and 2.3% compared to the Adaptive Genetic Algorithm (AGA); it reduces by 15.7% and 19.9% compared to the traditional Genetic Algorithm (GA); and decreases by 15.0% and 4.9% compared to the Brain Storm Optimization algorithm (BSO). This model can quickly generate distribution routes for decision-makers to effectively avoid traffic congestion, reduce vehicle carbon emissions, and lower total distribution costs, thereby promoting the harmonious development of cold chain logistics distribution and the ecological environment. References | Related Articles | Metrics

Select

Harness Engineering-Based Granular-Flow Cooperative Control Method for Work-Zone Traffic with Temporary Exclusive Container-Truck Lanes WENG Jinxian, GUO Haofei, DING Haifeng 2026, 26(5): 173-183.  DOI: 10.16097/j.cnki.1009-6744.2026.05.015 Abstract ( )   PDF (2857KB) ( )   This study proposes a Harness Engineering-based large language model approach for granular-flow cooperative control. Connected and automated passenger cars (CAV-Cs) are treated as controllable particles to guide the traffic flow upstream of a temporary exclusive container truck lane (ECTL) in a road work zone, thereby creating the stable operating conditions for human driven trucks and connected and automated truck platoons (CAV-TPs) to merge into the ECTL. First, to address the heterogeneity of traffic information and context loss in the continuous decision-making process of large language models, a structured representation and encapsulation module, including vehicle position, speed, merging demand, and collision risk, is developed for key information to support the generation of continuous control decision. Second, to address the potential infeasibility and insufficient safety constraints of actions generated by the large language model, a Harness Engineering safety-checking module is constructed to verify lane legality, speed boundaries, safe gaps, and collision risk constraints for high-level cooperative control intentions. Finally, the verified high-level actions are transformed into executable commands, including lane changing, acceleration, deceleration, and maintaining. Taking the port collector-distributor road work zone of the Donghai Bridge in Shanghai as the study site, a mixed-traffic simulation scenario was developed by using the open-source traffic simulation platform SUMO to evaluate the proposed method. The simulation results show that, compared with the uncontrolled scenario, the proposed method reduces the total delay cost in the work zone by 11.4%~21.82%, which is lower than that from the dynamic speed control in most scenarios. The numbers of controlled vehicles and control actions, as well as memory usage, are 14.0%~19.8%, 6.5%~ 9.6%, and 14.7%~23.0% of those required by dynamic speed control, respectively, which indicate that the proposed method requires a lower intervention scale and fewer computational resources. Sensitivity analysis shows that when the penetration rates of CAV-TP and CAV-C are 75% and 40%, respectively, a relatively favorable balance between delay cost and collision risk can be achieved. References | Related Articles | Metrics

Select

Highway Ramp Merging Control Method for Autonomous Vehicles Integrating Vehicle Trajectory Prediction DU Kai, LIU Huiyang, HE Qihang, YU Yanbo, CHEN Dan, SONG Jingni 2026, 26(5): 184-196.  DOI: 10.16097/j.cnki.1009-6744.2026.05.016 Abstract ( )   PDF (2436KB) ( )   The ramp merging area of expressways has become a key bottleneck for traffic safety and efficiency due to the large speed difference and complex weaving of vehicle flows, especially against the backdrop of continuous traffic growth. A lane changing control method for autonomous vehicles at highway on-ramp merging areas is proposed in this paper. It aims to address the critical safety and efficiency bottlenecks caused by large speed differences and complex weaving under increasing traffic demand. The lack of predictive capability of traditional ramp merging control methods makes it hard to coordinate merging timing and safety, so the proposed method embeds merging decisions and predicted vehicle trajectories into the longitudinal control module. By combining the macro-level merging intentions with micro-level future position information, it generates an optimized acceleration profile that balances safety and efficiency. To reflect the real driving behavior, this paper builds a realistic merging scenario dataset with the public exiD dataset and field data from the Xi'an Ring Expressway. This dataset is used to train the Autoformer-based trajectory prediction model and the GWO-XGBoost lateral merging decision model. On this basis, an integrated merging control framework is developed by combining Autoformer, GWO-XGBoost, and DDQN. Simulation results show that the average merging success rate of proposed method reaches 98.21%, which is markedly higher than those of XGBoost-DDQN, LC2013-DDQN, and DDQN. The proposed method also improves merging efficiency and safety. SHAP-based interpretability analysis further shows that the current and predicted positions of the merging vehicle, together with the positions of mainline vehicles, are the dominant factors to shape the control outputs. This confirms that the model can effectively use temporal information to support more rational merging decisions. References | Related Articles | Metrics

Select

Spatiotemporal Propagation Characteristics of Traffic Phase Transitions at Consecutive Merge Bottlenecks CHEN Qiucheng, ZHU Shunying, ZHANG Cen, LI Xiaowei 2026, 26(5): 197-209.  DOI: 10.16097/j.cnki.1009-6744.2026.05.017 Abstract ( )   PDF (4088KB) ( )   To reveal the spatiotemporal propagation characteristics of traffic phase transitions and their key influencing factors at consecutive merge bottlenecks, this paper selected the Zouhuang Interchange in Wuhan as the study site, and conducted research by the high-precision trajectory data under congestion intensification and congestion mitigation scenarios collected via unmanned aerial vehicle. The stages of congested traffic flow were first identified by three-phase traffic theory combined with changepoint detection, after which the congestion propagation speed was used to determine the slope of spatiotemporal units. Microscopic trajectories were then aggregated into explanatory variables; the wide moving jam proportion was constructed as the dependent variable, and a spatiotemporal weight matrix incorporating bottleneck constraints was designed; a spatiotemporal autoregressive model was established to estimate the spatial spillover effects of phase transitions under both scenarios and to conduct scenario simulations. Results show that neglecting bottleneck constraints causes the spatial spillover coefficient to exceed the reasonable range. After incorporating the constraints, the spatial spillover effects are opposite in the direction between two scenarios: significantly negative during congestion intensification (−0.655) and significantly positive during congestion mitigation (0.947). Scenario simulations indicate that the priority deployment zone for variable speed limit measures is the 300~540 meters downstream section from bottlenecks, and the key location for lane-changing control is the upstream merge bottleneck. Early intervention at the upstream on-ramp is recommended, while downstream on-ramp control becomes more effective after congestion has sufficiently propagated upstream. References | Related Articles | Metrics

Select

Transmission Mechanism of Customized Public Transportation Travel Intention Based on Multi-dimensional Service Perception and Planned Behavior CHEN Deqi, YANG Xinxia, ZHANG Wenhui, HUO Xinran, PAN Fuquan 2026, 26(5): 210-226.  DOI: 10.16097/j.cnki.1009-6744.2026.05.018 Abstract ( )   PDF (3375KB) ( )   To deeply study the formation mechanism of passengers' travel intentions for customized public transportation, this paper proposes a Bayesian structural equation model (BSEM) based on the Theory of Planned Behavior (TPB) framework, multi dimensional service perception latent states (SP), and multi-dimensional service perception and planned behavior. The formation mechanism of passengers' travel intentions is depicted from three levels: observation operator, state mapping operator, and hierarchical transmission operator. The research results show that in the mapping layer from psychological cognition to behavioral intention, the mapping intensity of perceived behavioral control is the greatest, followed by subjective norms, and attitude is relatively weak. There is significant heterogeneity in the mapping channel from service perception latent state to psychological cognition latent state, and different service attributes exert their effects through differentiated psychological paths. In the hierarchical transmission analysis, safety has the greatest cross-level transmission intensity, economy has a dual-channel transmission structure, reliability is influenced through social norms diffusion, convenience belongs to the capacity constraint factor, and comfort is manifested as a weak signal factor. Further multi-dimensional heterogeneity analysis reveals significant differences in the intention formation mechanisms among different groups: commuters are more likely to be affected by various constraints, while non-commuters lean towards evaluation-driven mechanisms. The low-to-middle-income groups pay more attention to cost constraints and use convenience, whereas the medium-to-high-income groups value safety guarantees and service reliability more. At the same time, by comparing the proposed model with the SP-BSEM, TPB-BSEM, and TAM-BSEM models, the results show that the explanatory power R2 of the proposed model relative to the traditional TPB model has increased by 39.1%. In terms of prediction performance, its WAIC-elpd and LOO-elpd indicators are the best, confirming the necessity of incorporating service perception factors into the psychological decision-making framework. The research indicates that the formation of travel intentions for customized public transportation shows significant hierarchical transmission structure characteristics. Different service perception latent states exert structural influences through differentiated psychological channels, providing systematic theoretical basis and practical paths for the optimization of customized public transportation services. References | Related Articles | Metrics

Select

Expressway Travel Time Estimation Method Considering Trajectory Semantics and Spatiotemporal Correlations ZOU Fumin, JIANG Linwen, CAI Qiqin, LIAO Lvchao, LAI Shukun 2026, 26(5): 227-237.  DOI: 10.16097/j.cnki.1009-6744.2026.05.019 Abstract ( )   PDF (2829KB) ( )   To address the insufficient mining of semantic correlations among road segments in trajectory data and the inadequate modeling of rich spatiotemporal information in existing travel time estimation studies, this paper proposes a travel time estimation method that considers trajectory semantics and spatiotemporal correlations. First, a semantic-aware trajectory representation learning module is designed, in which the road segment sequence of a trip is regarded as a trajectory semantic unit. A multi-layer Transformer encoder is used to learn the contextual correlations among the road segments, and a masked prediction self-supervised task is introduced to optimize the quality of road segment representations, thereby obtaining the trajectory representations that integrate the road segment semantics and topological correlations. Second, an attribute embedding module is constructed to map multidimensional features, such as vehicle type, departure time, and road segment length, into a unified feature space, which enhances the ability of model to represent external influencing factors. Third, an adaptive gated attention network is designed to jointly model the global path dependencies and local correlations among the adjacent road segments, while an attention output module dynamically assigns the contribution of each road segment to travel time. Experimental results on the Electronic Toll Collection (ETC) and Manual Toll Collection (MTC) data from three expressways in Fujian Province show that the proposed method outperforms comparison methods in terms of Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE) and Root Mean Square Error (RMSE). The errors are reduced by 1.44%, 1.46%, and 1.22%, respectively, and the prediction efficiency is improved by 19.30% compared with the best baseline method. Ablation experiments further verify the effectiveness of trajectory representation learning, adaptive gated attention network, and attention output module in improving prediction performance. In conclusion, the proposed method enhances the accuracy and efficiency of expressway travel time estimation through precise trajectory representation modeling and the extraction of underlying spatiotemporal patterns. References | Related Articles | Metrics

Select

Hierarchical Cooperative Control of Connected Vehicle Ramp Merging Under Spatiotemporal Constraints ZHANG Jianhua, CHEN Yunshuai 2026, 26(5): 238-247.  DOI: 10.16097/j.cnki.1009-6744.2026.05.020 Abstract ( )   PDF (2377KB) ( )   To address the issues such as low traffic efficiency and frequent vehicle conflicts in highway ramp merging areas caused by the coupling of spatial and temporal constraints for intelligent connected vehicles, this paper proposes a hierarchical cooperative control framework using a "global optimization-local tracking" structure. The upper layer generates a globally optimal target merging time sequence through a mixed-integer linear programming (MILP) scheduler, while the lower layer uses a virtual platoon based distributed controller to achieve precise vehicle trajectory tracking. Based on a co-simulation platform combining SUMO and Python, the study constructs ramp merging scenarios and analyzes indicators such as average vehicle speed, average delay, and throughput under different traffic volumes. The results show that the proposed method can effectively mitigate the "bottleneck effect" and disturbance the propagation caused by vehicle waiting under uncontrolled conditions, keeping the per-vehicle delay within 5 seconds; compared with the uncontrolled natural merging, the cooperative merging control method can increase the speed of average vehicle by 23.4% ~46.7% and reduce the average delay by 37.2%~74.9% across different traffic demand levels; compared with the first-come-first-served merging control, it can increase the speed of average vehicle by 2.1%~15.9% and reduce the average delay by 4.7%~33.3% under high ramp traffic demand. Furthermore, when the combined density of mainline and ramp traffic reaches 2 800 vehicles per hour, the system throughput tends to saturate, with the total number of vehicles passing stabilizing at around 119, which reflects good traffic stability. References | Related Articles | Metrics

Select

Optimization of Explosives Container Stowage Plan Based on Improved Artificial Bee Colony Algorithm REN Hongxiang, ZHONG Xin, LI Ye 2026, 26(5): 248-259.  DOI: 10.16097/j.cnki.1009-6744.2026.05.021 Abstract ( )   PDF (2477KB) ( )   A scientifically sound and rational stowage plan is the prerequisite and foundation for ensuring the safety of maritime dangerous goods container transport. Taking the draft difference, initial stability height, and number of container shifts as optimization objectives, this study constructs a multi-objective optimization model for mixed stowage of explosives within compatible groups in accordance with the stowage and segregation requirements for Class 1 explosives specified in the International Maritime Dangerous Goods Code. A Particle Bee Colony algorithm is then proposed based on a fitness-segmentation search strategy. The algorithm integrates the global search capability of the Artificial Bee Colony algorithm with the local optimization capability of Particle Swarm Optimization, and improves feasible-solution screening efficiency and convergence accuracy through fitness-segmentation search, particle-bee updating, and infeasible-solution handling mechanisms. Multi-port simulation experiments show that, under the premise of satisfying explosive stowage, segregation, and ship safety constraints, the proposed algorithm achieves higher fitness than the comparison algorithms. It reduces computational time by 8%~16%, and decreases the number of container shifts by 20.0%~38.5%. The results demonstrate the efficiency and applicability of the proposed algorithm, and provide a reference for safe and efficient explosives container stowage. References | Related Articles | Metrics

Select

Integrated Optimization of Consolidation and Demand Assignment for High Speed Railway Express Delivery with Stochastic Demand and Capacity SHUAI Bin, XU Haoran, XIE Anhao, SUN Chaoyuan 2026, 26(5): 260-271.  DOI: 10.16097/j.cnki.1009-6744.2026.05.022 Abstract ( )   PDF (2262KB) ( )   High-speed railway express delivery (HSReD) faces a trade-off between timeliness and cost due to dual uncertainties in freight demand and available transport capacity. To address this issue, this paper investigates the integrated optimization of front end freight consolidation process at hub stations and the train loading scheme. A two-stage stochastic programming model is formulated to minimize the expected total cost. In the first stage, the decision variables are the consolidation dispatch thresholds for freight with different timeliness grades and the number of reserved carriages. The second-stage variables include the proportions of direct-shipping and consolidation, as well as the train loading scheme. This model incorporates some constraints such as inventory balance, threshold triggering, train capacity, and station unloading capacity. To characterize the uncertainties of market demand and available piggyback capacity, the model generates a scenario subset by Monte Carlo sampling and the backward reduction method. Then it is solved by using Gurobi via MATLAB. A case study based on the selected sections of the Chengdu-Chongqing-Guiyang railway loop is conducted to verify the model. The results demonstrate that the consolidation threshold for next-day delivery freight increases monotonically with destination distance. The threshold for next-morning delivery freight remains generally low, except at Chengdu East Station. Meanwhile, the threshold for same-day delivery freight is lower for nearby stations and higher for distant stations. These three thresholds stabilize under scenario scales of 10, 40, and 50, respectively. Reserved carriages are concentrated on key train services experiencing order surges and backlog releases, accounting for approximately 23.5%. The value of the stochastic solution for the two-stage model, relative to the expected value model, stabilizes between 0.69% and 1.38%. Compared to a pure direct-shipping mode, the cumulative average cost savings of the consolidation mechanism converge to approximately 11 000 yuan. Finally, by adjusting the fixed consolidation cost and delay penalty coefficient, this paper demonstrates that the consolidation mechanism can simultaneously optimize the train load factor and the freight delay rate. References | Related Articles | Metrics

Select

Travel Patterns of Elderly Passengers in Urban Rail Transit Based on Automatic Fare Collection Data LIU Chenhui, CHEN Sheng 2026, 26(5): 272-282.  DOI: 10.16097/j.cnki.1009-6744.2026.05.023 Abstract ( )   PDF (3188KB) ( )   To improve the aging-friendly service level of urban rail transit precisely and effectively, this study analyzes the travel characteristics of elderly passengers using the Automatic Fare Collection (AFC) data from Changsha metro. First, the natural breaks method was applied to classify the elderly passengers into high-, medium-, and low-frequency metro user groups based on their number of trips. The high-frequency group accounted for only 6.13% of elderly passengers but contributed 22.83% of total elderly trips; their average weekly metro trips reached 14.28, which is 3.72 times of the overall average. Then, focusing on the high frequency elderly passengers, a Gaussian Mixture Model (GMM) was used for clustering based on the Gini coefficient of travel days, hourly entropy, OD entropy, and weekend-to-weekday travel ratio. The results show that the high-frequency elderly passengers can be divided into three types: (1) the full-week active type accounts for 59.88% of the total elderly passenger flow, with stable travel intensity throughout the week and high spatiotemporal diversity; (2) the weekday-active type, characterized by travel peaks on weekdays and a dramatic decline in trips during weekends, accounts for 30.74% of the total elderly passenger flow; and (3) the point-to-point type characterized by highly rigid spatiotemporal patterns, accounts for only 9.37%, with a Gini coefficient of 0.072 and an OD entropy of 1.10. Compared with commuters, the elderly passengers have shorter travel distances but longer travel times. Spatially, their trips are concentrated in the central urban area. Temporally, their evening peak is about one hour earlier, and the morning peak on weekends is about one hour later. This two-stage framework effectively reveals the heterogeneity of elderly passengers and provides a quantitative basis for formulating differentiated aging-friendly policies for metro systems. References | Related Articles | Metrics

Select

Urban Electric Vehicle Charging Demand Prediction for Capacity Pressure Identification YAO Shengyong, ZHAO Chen, SUN Weiping 2026, 26(5): 283-296.  DOI: 10.16097/j.cnki.1009-6744.2026.05.024 Abstract ( )   PDF (2584KB) ( )   To address the problems that the charging demand in urban electric vehicle features the strong short-term fluctuations and significant regional differences, as well as the continuous demand forecasts cannot directly characterize the operational pressure of public charging facilities, this paper develops a short-term charging demand forecasting and pressure transformation framework for capacity-pressure identification. First, the proposed framework integrates multi-source features, including regional charging volume, facility occupancy rate, charging duration, electricity price, service price, and weather factors, and employs a TCN-BiLSTM-Attention model for short-term charging demand forecasting. Specifically, the TCN extracts local temporal features, the BiLSTM captures temporal dependencies, and the attention mechanism emphasizes key historical time steps for next-period demand prediction. Second, taking the 95th percentile of historical regional charging demand during the training period as a relative capacity benchmark, a capacity-pressure index is constructed to transform continuous demand forecasts into capacity pressure states at different levels. Finally, key high-pressure regions and periods are identified according to the spatiotemporal distribution characteristics of the pressure levels. Empirical analysis based on a public dataset shows that the developed model achieves a goodness-of-fit value of 0.968 2, with a root mean square error, mean absolute error, and weighted mean absolute percentage error of 152.719 7 kWh, 41.248 9 kWh, and 16.633 6%, respectively. Under the 95 th-percentile capacity benchmark, the proportion of samples classified as high pressure or above account for 22.638 0% of all samples, among which the proportion of overload-risk samples account for 8.751 4%. The proposed framework enables the transformation from short-term charging demand forecasting to regional capacity-pressure state identification and provides a quantitative basis for the operational monitoring of public charging facilities, the screening of key regions, and differentiated resource allocation. References | Related Articles | Metrics

Select

Individual Workload Tolerance Differences Among Air Traffic Controllers Through Multimodal Data Fusion GU Qiuli, WANG Lili 2026, 26(5): 297-308.  DOI: 10.16097/j.cnki.1009-6744.2026.05.025 Abstract ( )   PDF (2580KB) ( )   The individual workload tolerance of air traffic controllers is the decisive determinant of air traffic control (ATC) system resilience. Existing studies on air traffic controller workload are largely based on group-level mean models, which are inadequate for capturing individual variability. To address this limitation, this paper recruited 24 frontline controllers and conducted simulated ATC experiments under contrasting pre-duty (alert) and post-duty (fatigued) conditions, while multimodal physiological data was collected simultaneously to quantify individual workload tolerance across a complete work cycle. First, a weighted Partial Least Squares Path Model (PLS-PM) was employed to fuse the multidimensional physiological and oculomotor indicators into a Latent Variable (LV) which representes comprehensive workload, thereby establishing its coupling relationship with traffic flow characteristics. Subsequently, guided by the Bayesian Information Criterion (BIC), a Piecewise Linear Regression Model (PWLF) was fitted to calibrate the personalized initial thresholds within the LV space for each controller. Building on this, a dynamic update mechanism-defined as 1.0 standard deviation (σ) above the mean of standardized latent scores-was introduced to determine adaptive thresholds, enabling precise labeling of high-workload states and identification of the onset point of sustained high workload. Finally, by integrating the initial thresholds, adaptive thresholds, and the onset point of sustained high workload, and mapping these back to traffic flow coordinates, we established personalized aircraft count thresholds for both pre-duty and post duty phases. Longitudinal within-subject comparisons facilitated the quantitative assessment and classification of workload tolerance profiles. Results revealed three distinct types: Fatigue-Resistant(20.8%), Steady-State Maintaining (66.7%), and Fatigue Sensitive(12.5%). Operational practice should give priority attention to the fatigue-sensitive group. These findings provide a valuable guidance for ATC units to optimize person-job matching strategies. References | Related Articles | Metrics

Select

Short-Term Passenger Flow Prediction for Urban Rail Transit with Multi-Head Cross Attention Fusion GONG Liang, ZHU Xinyu, XU Dejie, XU Lei, CHEN Zihao 2026, 26(5): 309-319.  DOI: 10.16097/j.cnki.1009-6744.2026.05.026 Abstract ( )   PDF (2059KB) ( )   Given the strong coupling between temporal dependencies and spatial correlations in short-term passenger flow prediction for urban rail transit, the existing models have difficulties in fully achieving the deep interactions among heterogeneous spatiotemporal features. To address this issue, this paper proposed a short-term passenger flow forecasting model which integrates Long Short-Term Memory (LSTM), Graph Convolutional Network (GCN), and Multi-Head Cross Attention (MHCA). The model employs the LSTM to extract the temporal evolution patterns of station-level passenger flow sequences and uses the GCN to learn spatial dependencies under the metro network topology. Based on these representations, a bidirectional multi-head cross-attention mechanism is constructed to achieve the deep interaction and information fusion between the temporal encoding features and spatially enhanced representations. The mechanism further incorporates asymmetric projection matrices and a spatial masking mechanism to enhance heterogeneous feature representation, and reduce the interference from weakly correlated distant stations. Experiments are conducted by using the Automatic Fare Collection (AFC) data collected from 80 stations of the Hangzhou Metro. Compared with the LSTM, LSTM+GCN, STGCN (Spatio-Temporal Graph Convolutional Network), ASTGCN (Attention Based Spatial-Temporal Graph Convolutional Network), DCRNN (Diffusion Convolutional Recurrent Neural Network), and Graph WaveNet (Graph WaveNet for Deep Spatial-Temporal Graph Modeling), the proposed model achieves the best overall performance, with RMSE (Root Mean Square Error), MAE (Mean Absolute Error), and MAPE (Mean Absolute Percentage Error) of 37.85, 30.94, and 22.68%, respectively, in the network-wide 15-minute forecasting task. A superior performance is also observed under different forecasting horizons, passenger flow types, and peak-hour scenarios. The ablation studies further confirm the effectiveness of the proposed architecture and demonstrate the positive contributions of the spatial masking strategy, asymmetric projection mechanism, and the synergistic interaction between GCN and MHCA. Research findings indicate that the proposed model can effectively enhance the capability to model complex spatiotemporal dependencies in urban rail transit, and provide a reference for passenger flow prediction and intelligent operation management of urban rail transit. References | Related Articles | Metrics

Select

Two-Level Takeoff and Landing Facility Siting for Urban Low-Altitude Logistics Under Complex Spatial Constraints LI Nan, WANG Zishi, ZHENG Zhigang, ZHANG Yujiao 2026, 26(5): 320-330.  DOI: 10.16097/j.cnki.1009-6744.2026.05.027 Abstract ( )   PDF (2306KB) ( )   A fixed-size two-level facility siting method is developed for urban low-altitude logistics under concentrated demand and complex spatial constraints. Population and functional activities are integrated to represent the potential demand in planning stage and estimate the daily freight demand. The areas unsuitable for construction, such as water bodies and airport-restricted spaces, are removed as hard constraints, while sensitive facilities are accounted for proximity penalties. The candidate secondary sites for terminal coverage and primary hubs for regional consolidation are generated separately. A two-level siting model considering two-level facility capacity and freight-flow balance is formulated based on demand coverage and secondary-site-to hub connections. The COIN-OR Branch-and-Cut (CBC) solver provides a time-limited feasible solution as the mixed-integer linear programming (MILP) benchmark, and a coverage-balance dual-population heuristic algorithm generates alternative layouts. Site matching links the siting results to rooftops, platforms, or continuous open spaces. In the ten-district Tianjin case, differentiated candidate generation for the two facility levels forms the corresponding candidate spaces that reflect demand distribution and facility roles. The proposed heuristic algorithm improves the mean objective value by 2.44% and 3.62% over variable neighborhood search and adaptive large neighborhood search, respectively, and its best solution reaches 97.90% of the 3 600 s time limited mathematical programming solution. The sensitivity analysis shows that facility scale, service range, and capacity jointly affect coverage, distance, and facility loads. The secondary sites in both layouts can be matched to rooftops or platforms satisfying the area requirement, whereas no primary hub can be matched to a continuous open space satisfying both area and shape requirements, indicating the different site conditions for two facility levels. References | Related Articles | Metrics

Select

Demand-Responsive Modular Bus Scheduling Considering Stochastic Travel Times DONG Xianlong, WANG Li, YUE Hao, YANG Shanglin 2026, 26(5): 331-341.  DOI: 10.16097/j.cnki.1009-6744.2026.05.028 Abstract ( )   PDF (1874KB) ( )   This paper addresses the static scheduling optimization problem of demand-responsive modular bus systems under stochastic travel times. A chance-constrained programming model is developed, integrating route planning, passenger assignment, platooning, and in-motion transfer decisions. The stochastic model is equivalently reformulated as a mixed-integer second-order cone programming (MISOCP) model via deterministic transformation. To enable an efficient solution, an adaptive large neighborhood search (ALNS) algorithm is designed with embedded platooning and transfer search strategies, augmented by periodic local search to enhance global optimization performance. A case study is conducted with the real travel demand data from Shijingshan District, Beijing. The results demonstrate that the proposed scheduling strategy can effectively adapt to stochastic travel time environments and exhibit strong collaborative platooning capabilities. Compared with the conventional demand responsive transit operations, the proposed strategy reduces the total system cost by approximately 4.47% and the vehicle operating cost by approximately 5.85%. Sensitivity analyses are further performed to systematically examine the impacts of bus capacity, confidence level, and the travel time variability on scheduling decisions. The findings reveal that the bus capacity should be configured to strike a balance between reducing operating costs and maintaining passenger service quality. Increasing the confidence level of the chance constraints enhances on-time arrival reliability but incurs higher operating costs, while ignoring travel-time stochasticity systematically underestimates operational risks. In contrast, accounting for stochasticity enables an effective trade-off between service reliability and economic efficiency. This study provides theoretical foundations and practical decision-making support for the robust scheduling design of modular bus systems under uncertainty. References | Related Articles | Metrics

Select

Cascade Prediction Model for Flight Ground Support Milestone Nodes Based on Evidential Reasoning XU Tao, ZHANG Qian, SUN Xiaojing, XIE Jiwen, FENG Xia 2026, 26(5): 342-351.  DOI: 10.16097/j.cnki.1009-6744.2026.05.029 Abstract ( )   PDF (2076KB) ( )   To address the issues of seasonal concept drift in the time distribution of milestone nodes and the error propagation in multi-node cascading prediction during flight ground support processes, this paper proposes a Cascade Evidential Reasoning-Light Gradient Boosting Machine (CER-LGBM) prediction model. The model consists of two sub-modules: evidential reasoning prediction and cascading rolling update. In the evidential reasoning prediction module, concept intervals are partitioned based on backtracking error sequences, and fixed expert models are trained within each interval. During the testing phase, drift detection is dynamically performed based on the recent error levels of experts in each interval; subsequently, evidential credibility is adjusted adaptively, and the outputs of all experts are weighted and fused. The cascading rolling update module constructs process features by using the real values of newly observed nodes and the prediction errors from the previous round. It controls the update time by temporal consistency constraints, and implements the rolling update prediction for incomplete nodes. Experimental results demonstrate that, in scenarios with drifted data distributions, the accuracy of ±5-minute prediction for 66.67% of nodes reaches 70% or above in the initial prediction phase. As the support process progresses, the prediction accuracy of ±5-minute for all nodes reaches 95% or higher. References | Related Articles | Metrics

Select

Scenario-Adaptive Gated Fusion Forecasting of Short-Term Urban Rail Transit Passenger Flow Under Environmental Disturbances LI Xianwang, ZHANG Kaixin, HU Ke, YANG Zixin 2026, 26(5): 352-361.  DOI: 10.16097/j.cnki.1009-6744.2026.05.030 Abstract ( )   PDF (2227KB) ( )   Existing models struggle to adapt to the dynamic changes in the contributions of multi-source information under environmental disturbances, which leads to the degraded forecasting performance under abnormal weather conditions. To address this problem, the study proposes a scenario-adaptive gated fusion method for metro passenger flow forecasting. The proposed method integrates four types of sources: passenger inflow, passenger outflow, graph topology, and environmental information. A scaled dot-product graph attention module is developed to capture the dynamic spatial dependencies beyond static topology, and an environment-driven gating mechanism is introduced to adaptively adjust the fusion weights of the four information sources. Experiments are conducted by using the automatic fare collection data from 276 Beijing subway stations under 15 minutes, 30 minutes, and 60 minutes forecasting intervals. The proposed method reduces mean absolute error (MAE) by 28.5% compared with a bidirectional long short-term memory and graph neural network hybrid model at the 60 minutes interval. At the 15 minutes interval, the proposed method achieves a MAE of 22.34, a root mean square error (RMSE) of 41.41, and weighted mean absolute percentage error (WMAPE) of 8.79%, outperforming all comparison methods. Under abnormal weather conditions, the branch weight of graphs decreases from 0.599 6 to 0.530 8, while the branch weight of inflows increases from 0.181 5 to 0.232 1, and the weather branch weight rises correspondingly. Ablation experiments demonstrate that removing the gated fusion mechanism alone increases MAE to 34.10, while removing the graph attention module alone increases MAE to 36.47. The results show that the proposed scenario-adaptive gating mechanism effectively adjusts the contributions of multi-source information according to the intensity of environmental disturbances, and improves the forecasting robustness under complex weather conditions. References | Related Articles | Metrics

Select

Drivers' Visual Load of Desert Expressways Under Influence of Long Straight Alignments and Monotonous Landscape ZHU Xinglin, LIU Sijia, XU Jin, LIU Hongjun, LIANG Guangze, GUO Rui, SUN Xiaodong 2026, 26(5): 362-373.  DOI: 10.16097/j.cnki.1009-6744.2026.05.031 Abstract ( )   PDF (3800KB) ( )   To investigate the variation patterns of drivers' visual characteristics and visual workload under the special environment of long straight sections and monotonous landscapes on desert highways, a real-vehicle experiment was carried out on S21 Altay Urumqi Desert Highway in Xinjiang with 32 drivers. k-medoids dynamic clustering divided the visual field into six fixation areas of interest; visual search patterns and transition rules were analyzed with the Markov chain. Taking fixation duration, horizontal saccade amplitude and pupil diameter as core indicators, an entropy weight-fuzzy comprehensive evaluation model was built, and the Gaussian mixture model was applied to fit the dynamic visual load variation. Results show that the fixation points of drivers are highly concentrated in the forward lane area under the monotonous desert highway conditions. With increasing mileage, the average fixation duration rises, and the pupil diameter shrinks. The saccade amplitude narrows, and the visual load accumulates significantly. Driving speed is linearly correlated with visual indicators; higher visual load reduces speed perception and increases speeding risk. Visual load presents periodic cumulative characteristics with peaks at 12.97 km, 34.05 km, and 57.8 km. Its baseline keeps rising after relief, and drivers can hardly recover to the initial alert state via self- regulation. The findings provide a theoretical basis and data support for alignment optimization and safety facility layout of desert highways. References | Related Articles | Metrics

Select

Evaluation of Drivers' Physiological-Behavioral Characteristics and Stable Driving Compensation Efficiency in Urban Underground Ring Roads SHANG Ting, LI Huan, ZENG Jiajia, GAO Caihua, ZHANG Yu 2026, 26(5): 374-385.  DOI: 10.16097/j.cnki.1009-6744.2026.05.032 Abstract ( )   PDF (3352KB) ( )   To clarify the drivers' physiological-behavioral response characteristics and stable-driving compensation efficiency in curved scenarios of urban underground ring roads, an on-road driving experiment was conducted in the Jiefangbei Underground Ring Road in Chongqing. Vehicle operation and electrocardiographic data were collected from 38 drivers in the entrance section, single-curve sections, continuous-curve sections, and exit section. Factor analysis, difference testing, and an input-oriented DEA (Data Envelopment Analysis) model were used to identify the driving behavior, electrocardiographic responses, and the matching efficiency between physiological compensation input and behavioral stability output across different scenarios. The results show that driving behavior and electrocardiographic responses differed significantly across scenarios, and the scenario-related variations of behavioral indicators show more distinct. The entrance transition section was characterized by speed recovery, increased mean heart rate, and enhanced heart rate fluctuation, whereas the exit transition section shows anticipatory deceleration control. Single curve scenarios combined with access points were more likely to induce speed correction and electrocardiographic fluctuation, while driving states were relatively stable in continuous-curve scenarios. At the driver-level DEA evaluation, the average efficiency across scenarios ranged from 0.898 to 0.920; however, the numbers of DEA strongly efficient drivers were relatively small in the single-curve L2 scenario and the continuous-curve L5 scenario, with 10 and 12 drivers, respectively. At the scenario-level DEA evaluation, continuous- curve scenarios and some access-point scenarios reached DEA strong efficiency, whereas the entrance transition section and the single- curve scenario combined with a garage access point were not achieved DEA efficient. The entrance transition section had the lowest comprehensive efficiency, at 0.446, and the comprehensive efficiency of single-curve scenario combined with a garage access point was 0.896. These two scenarios were mainly characterized by electrocardiographic input redundancy and insufficient speed-stability output. The results indicate that stable-driving compensation efficiency in underground ring roads is jointly affected by curve type, curve radius, and access-point conditions. The entrance transition section and single-curve scenario combined with a garage access point are the main areas of efficiency loss. References | Related Articles | Metrics

Select

Coordinated Lane-Changing Methods for Connected and Automated Vehicles in Upstream of Urban Expressway Work Zones CHENG Guozhu, LUO Tianjian, WANG Wenzhi, LIANG Jiaqi, FENG Tianjun 2026, 26(5): 386-402.  DOI: 10.16097/j.cnki.1009-6744.2026.05.033 Abstract ( )   PDF (3007KB) ( )   To address traffic bottlenecks and safety risks induced by work zones on urban expressways, this study proposes a coordinated lane-changing method for connected and automated vehicles (CAV) based on the Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm. Considering the spatiotemporally deterministic characteristics of work zones, a staged control strategy is developed, consisting of "advance preparation and platoon splitting in the upstream warning zone-guided lane changing in the transition zone-steady mixed traffic maintenance in the work zone". Incorporating the concept of "particle-flow integration", a two-layer Markov Decision Process (MDP) model for CAV coordinated lane-changing in work zones is established. An extended reward function is designed to accommodate platoon reconfiguration and merging tasks. The problem is solved using an improved Predictive Recurrent Multi-Agent Deep Deterministic Policy Gradient (PR-MADDPG) algorithm, which encodes work zone partitioning information into the state space, enabling agents to make anticipatory decisions based on future traffic conflict types. Simulation experiments are conducted on the SUMO-Python co-simulation platform. The results demonstrate that the proposed method can reliably trigger control behaviors such as platoon splitting, gap creation, cooperative yielding, and downstream reformation under various work zone scenarios. In three types of lane-closure scenarios, the standard deviation of lane changing speed remains below 3.5, indicating small speed fluctuations and smooth lane- changing maneuvers. Furthermore, compared with the improved Deep Q-Network (DQN) algorithm, within a CAV penetration rate range of 31%~60%, the average traffic flow increases by 7.6% and 7.1% in the inner dedicated lane and outer mixed lane scenarios, respectively, while the peak flow increases by 7.9% and 8.5%. The most significant improvement is observed in the middle mixed-lane closure scenario, where, at densities of 20~40 veh·km-1, the traffic flow is increased by 12.6% compared to the improved DQN algorithm. Moreover, the PR MADDPG strategy reduces the traffic conflict probability (Time-To-Collision, TTC<4 s) by an average of 79.8% and the occurrence of extreme deceleration maneuvers by approximately 80%. Furthermore, horizontal comparisons with mainstream multi-agent algorithms such as Q-Value Mixing (QMIX) and Multi-Agent Proximal Policy Optimization (MAPPO) demonstrate that, taking the middle lane closure scenario as an example, the peak flow of the proposed method is improved by 6.8% and 3.9%, respectively, and the extreme deceleration maneuvers are reduced by 75.4% and 63.7%, exhibiting superior cooperative control capabilities in complex and restricted spaces. Ablation studies validate the independent performance gains of the zone encoding, dual- layer MDP architecture, and policy clipping mechanism in enhancing decision predictability, ensuring macroscopic coordination, and maintaining training stability. These results indicate that the proposed strategy can effectively mitigate upstream traffic disturbances in work zones and enhance traffic efficiency and operational stability under mixed traffic conditions. References | Related Articles | Metrics

Select

Identification and Stability Control Method for Dynamic Traffic Bottleneck Sections in Long Tunnels LIU Wei, YUE Jia 2026, 26(5): 403-412.  DOI: 10.16097/j.cnki.1009-6744.2026.05.034 Abstract ( )   PDF (2087KB) ( )   Under the conditions of synchronized flow when the traffic in long tunnels approaches saturation, local traffic bottleneck sections are easy to form due to the changes in driving environment and random traffic disturbances, resulting in congestion that propagates backward and turns into a wide moving jam. To address the potential random congestion problems in the long tunnels near saturation, a dynamic identification method for bottleneck sections and a stability-constrained control method based on real time detection data from characteristic cross-sections are proposed. According to the characteristics of the driving environment inside the long tunnels, the tunnel is divided into multiple sections, and the traffic flow and speed parameters of each section are detected in real time. By introducing the change in section traffic density, a macroscopic traffic flow model for dynamic traffic state prediction, bottleneck section identification, and stability control in the long tunnels is established. Near quasi-steady state points, the model is linearized, and a discrete state matrix is constructed. Based on the eigenvalue criteria, the bottleneck section is identified where the maximum eigenvalue modulus of traffic flow is closest to 1, and the corresponding critical density and critical speed parameters are extracted. A model predictive control method centered on critical density is constructed, and the traffic flow in the bottleneck section is actively regulated via upstream variable speed guidance. Taking the cross-section detection data from a specific long tunnel as an example, a dynamic identification method for section bottlenecks and a model predictive stability control strategy were developed. Simulation experiments were compared with no control and fixed-speed control strategies. The simulation results show that the model predictive stability control method can suppress flow and speed fluctuations in bottleneck sections, maintain the section density stably near the critical density within the efficient operating range, increase lane capacity utilization to 98.1%, and achieve effective intervention before traffic flow density reaches the critical state. This enables the traffic flow at bottleneck sections in long tunnel to shift from ' intervenes after the congestion' to 'predictive and proactive control.' References | Related Articles | Metrics