Wayfinding problems in complex teaching buildings can reduce destination-finding efficiency and increase users’ mental load, especially when many rooms have similar layouts and are distinguished mainly by numbering systems. Signage is commonly used to support indoor wayfinding, but its effectiveness depends on whether the sign design and placement match the spatial layout and users’ interpretation of directional information. Taking the Third Teaching Building at Beijing University of Technology as a case study, this paper investigates classroom-related wayfinding difficulties and evaluates a practical supplementary signage strategy. First, questionnaire surveys were conducted among students and teachers who had entered the building to identify common wayfinding problems and signage-related causes. Then, based on user preferences, supplementary floor signage was designed to provide directional arrows and classroom-related textual information. A field experiment was conducted on a selected route to Room 301, comparing an original signage condition with a mixed-signage condition in which the supplementary floor sign was added while the existing ceiling-mounted signs were retained. The mixed-signage condition reduced the mean wayfinding time from 163.75 s to 133.80 s, corresponding to an 18.29% reduction. Post-experiment responses also indicated fewer hesitation- or detour-related difficulties under the mixed-signage condition. These findings suggest that supplementary floor signage placed at confusing decision points can be a feasible and low-cost retrofit measure for improving classroom wayfinding in complex teaching buildings.

In the context of the “dual carbon” targets of China, effectively reducing carbon emissions in the road freight transportation sector is crucial. Carbon emission trading has emerged as a cost-effective strategy for mitigating carbon emissions, particularly in the context of road freight transport. Currently, the system design of carbon emission trading for road freight transport industry is still unknown. To fill this gap, this study presents a comprehensive policy framework tailored for China's road freight transport sector. The framework includes key components such as data collection and verification processes for carbon emissions, the establishment and distribution of carbon allowances, regulatory mechanisms for carbon emission trading, and an integrated reward and punishment system. Furthermore, the study employs case studies and scenario analyses to examine the operational dynamics of the carbon trading system in road freight transportation. The findings indicate that, to meet compliance requirements and generate additional revenue, freight vehicle operators are incentivized to reduce carbon emissions either by decreasing vehicle miles travelled or by investing in vehicles with lower energy consumption. The purpose of this paper was to provide the theoretical basis for the effective facilitation of carbon emission trading system for China's road freight transport.

This study investigates the determinants of urban residents’ low-carbon behavioral intention (BI) and travel behavior (B) to inform evidence-based sustainable transportation policies. The theory of planned behavior was extended based on low-carbon travel data collected from 1186 respondents in Zhenjiang, China. A dual-analytical approach integrating partial least squares structural equation modeling and necessary condition analysis was employed to disentangle both sufficient and necessary conditions for low-carbon travel. The results reveal that perceived behavioral control and subjective norms function as both significant predictors and necessary prerequisites for BI, whereas driving identification (DI) acts as a significant barrier. Notably, anticipated regret emerged as a necessary but insufficient condition for BI. Travel habit (TH) and attitude (ATT) positively influenced both BI and B. These findings offer actionable recommendations for policymakers and urban planners to promote sustainable transportation choices, specifically by strengthening perceived behavioral control through targeted interventions, leveraging social norms, and implementing countermeasures to mitigate resistance arising from car-dependent identities.

The future trajectories of surrounding agents are critical for the motion planning and control of autonomous vehicles. Thus, this study employs Transformer to develop a multi-agent trajectory prediction model named Multi-agent Trajectory Vector Transformer (MaTVT). MaTVT features a lightweight architecture, comprising a dual-level encoder formed by a low-level encoder and a high-level encoder, along with a multi-modal decoder. Once input enters MaTVT, the low-level encoder first constructs polar coordinate systems centered on target agents and then projects historical trajectories and map elements to each agent-centered coordinate system. Next, it utilizes attention mechanisms to encode motion features, agent-agent interactions, and agent-infrastructure constraints independently and fuses them into the agent encoding sequence. Considering the agent response delay, the low-level encoder extracts heterogeneous spatial-temporal features from agent encoding sequences as the local encodings for target agents. Afterward, the high-level encoder treats all agents as the nodes in a directed graph and utilizes a Graph Attention Network to convert inter-agent relationships into global encodings, which are fused with the local encodings of target agents. Finally, the multi-modal decoder translates these fusion encodings into multi-modal trajectory predictions for target agents. This study selects complex traffic scenarios from the Argoverse Motion Forecasting dataset to create a dedicated dataset for MaTVT training, validation, and testing. The test results demonstrate that MaTVT outperforms advanced benchmark methods in prediction performance, revealing its superb accuracy, efficiency, and robustness. In addition, ablation studies further explain the interpretability of the main functional components of MaTVT and their contributions to prediction performance.

Human lane-changing behaviour is highly variable, making accurate intention prediction and socially adaptive decision-making challenging for intelligent vehicles. This capability is essential for safe and effective human–machine collaboration. However, real-world driving data are often degraded by communication loss between vehicles and roadside units, as well as sensor noise and uncertainty. To address these challenges, this study proposes CMT-GAT, a robust self-supervised contrastive learning framework for lane-change intention prediction in urban expressway weaving areas. The model employs a dual-channel trend attention mechanism to process complete and incomplete data in parallel, while a graph attention network captures interactions among surrounding vehicles and environmental context. Experiments demonstrate that CMT-GAT consistently achieves the highest prediction accuracy across all prediction horizons and missing-data rates, particularly under severe missing-data conditions.

Participants in large-scale sports events in cold environments demand efficient and personalized travel route planning to ensure they can take part in the activities on time, safely and comfortably. While existing travel planning studies primarily address basic origin-destination requirements, they often overlook the impact of cold environments on pedestrians’ psychology and physiology and the route recommendations for personalized travel. To bridge this gap, a personalized travel path planning with integrated en-route point of interest (POI) recommendations (PTPP-POI) framework in the cold events is proposed. It takes into account the impact of cold environments on the psychology and physiology of pedestrians, and dynamically recommends personalized travel routes to meet the travel needs of pedestrians in cold events. Taking the Beijing Winter Olympics as an example, and simulation results indicate that implementing PTPP-POI increases the probability of path matching—from 77.4% to 81.6% during ingress and from 47.0% to 52.6% during egress—along with significant increases in serviced POI counts. The PTPP-POI offers a user-friendly travel guidance framework that not only enhances the probability of path matching and POI utilization but also substantially improves the overall experience of participants.

Dockless shared bicycles are a popular and eco-friendly mode of urban transport. However, their unregulated parking creates challenges for both managers and researchers. This study introduces two artificial intelligence algorithms to improve the understanding and management of shared bicycle parking. We define shared bicycle parking aggregation areas as basic units for analysis and focus on their demand patterns. Traditional clustering methods struggle with the uneven density of parking spots. To solve this, we propose the Adaptive Density Hierarchical DBSCAN (AD-HDBSCAN) algorithm, which uses a concentration index to better detect parking clusters with varying densities. Additionally, analyzing high-dimensional time series data on bicycle demand is difficult with conventional methods. To address this, we develop a deep learning-based contrastive clustering model to analyze complex demand patterns over time. A case study in Beijing shows that AD-HDBSCAN improves clustering accuracy by 13.93% compared to traditional algorithms. We identify four distinct demand patterns across parking areas. Two patterns display rapid fluctuations during weekday peak hours, reflecting strong tidal effects that require focused management. There is also a clear link between demand patterns and nearby land use; areas with strong tidal patterns are often near companies or residential zones. Based on these findings, we suggest tailored management strategies to improve shared bicycle operations. Identifying natural parking clusters helps define better parking zones, while analyzing demand patterns guides scheduling decisions that match user needs. This study offers practical insights and recommendations for more efficient shared bicycle management.

Complex confined built environments such as metro stations typically rely on pre-defined evacuation plans to guide pedestrians to exits. However, this approach relies heavily on manual calibration and is insensitive to diverse and unforeseen conditions encountered during real-world emergencies. Simulation-based methods, which can provide precise evacuation plans, are constrained by time-consuming simulation processes and computational delays, making them difficult to provide timely decisions under critical conditions. To address these challenges, this paper proposes a novel simulation-in-the-loop emergency evacuation framework based on large language models to generate evacuation plans in a real-time manner. Various evacuation scenarios and the associated optimal evacuation plans are encoded into semantic descriptions and subsequently used to fine-tune a language model for generalised evacuation decision-making. When an emergency occurs, the language model can infer an evacuation strategy from the ongoing environmental conditions without running additional simulations. Experimental results show a 94.99% match rate between the simulation-based results and those generated by the language model, with a BLEU-4 score of 96.05 indicating high textual consistency. These results demonstrate the substantial potential of the proposed method to accelerate emergency response.

Motion optimization for intelligent connected vehicles (ICVs) requires predicting the trajectories of surrounding vehicles to achieve obstacle avoidance, which can be facilitated by the Intelligent Vehicle Cyber-Physical System (IVCPS). IVCPS builds digital twins of real-time traffic scenarios in the dynamic occupancy grid map (DOGM) format to support ICV trajectory prediction. To this end, this study integrates convolutional neural networks and the transformer to develop the convolutional multitarget trajectory transformer (CMTT). The CMTT utilizes parallel convolution blocks with prime-size kernels to extract spatial features from input DOGMs and generates independent feature map sequences through progressive receptive field expansion. It then employs an encoder-decoder structure to fuse and translate them into trajectory predictions with the guidance of preaccessible information. To comply with IVCPS specifications, this study combines SUMO and CARLA to construct a joint simulation platform that utilizes OpenStreetMap and real-world trajectory data from the pNEUMA and CitySim datasets to generate DOGM-format digital twins for CMTT training and validation. Experimental results indicate that CMTT surpasses all advanced baseline models in various urban traffic scenarios, showcasing its exceptional predictive performance and adaptability. Subsequently, ablation studies demonstrate the contribution of key components of CMTT to enhance model performance. Furthermore, this study determines the most suitable parameter configuration and pruning strategy for CMTT through parameter analysis and computational complexity analysis, achieving the optimal balance between prediction accuracy and computational efficiency. This study proposes CMTT, which demonstrates high prediction accuracy and computational efficiency, indicating potential for optimizing motion planning in real-world ICV applications.

This study constructs an integrated obstacle avoidance control strategy that can account for both lateral and longitudinal dynamic obstacles, enabling vehicles to quickly and stably evade obstacles in even emergency scenarios, thereby preventing accidents in complex road environments. Firstly, based on the safety distance model, a risk area dynamic partition method that considers both lateral and longitudinal directions is established. Subsequently, within the Frenet coordinate system, quintic polynomial candidate paths are generated based on the vehicle's acceleration, speed, and position before and after lane-changing, and the optimal path is determined with consideration of driving safety and comfort, as well as the constraints associated with it. Finally, utilizing a two-degree-of-freedom vehicle dynamics model and a lateral tracking error model, a lateral tracking controller based on Linear Quadratic Regulator (LQR) and a longitudinal tracking controller based on double Proportional-Integral-Derivative (PID) approach are designed. The results of hardware-in-the-loop experiments demonstrate that, compared to the LQR algorithm, the proposed strategy more effectively ensures the stability and obstacle avoidance capability of the vehicle during operation. Furthermore, data collected under various scenarios indicate that the proposed strategy meets the requirements for tracking performance and driving comfort while successfully ensuring obstacle avoidance. Overall, the proposed method satisfies the criteria for safety, comfort, and stability in vehicle obstacle avoidance within complex environments.

Based on the statistical data of Hainan Province from 1994 to 2023, this paper systematically analyzes the influencing factors of railway freight volume and constructs multiple models to predict it. Through Pearson correlation analysis, 12 variables highly correlated with railway freight volume were selected, the CatBoost model was used to uncover the non-linear impact of each factor on freight volume, and a comparative evaluation was conducted with Stepwise Regression, Ridge Regression, the GM (1,1) Grey Model, and Random Forest models. The results show that railway freight turnover, number of employees, waterway freight volume, gross regional product, and the added value of the tertiary industry have a significant impact on railway freight volume, and the CatBoost model outperforms other models on evaluation metrics such as mean relative error and absolute error, possessing higher prediction accuracy and robustness. This research helps to deepen the understanding of the influencing mechanism of railway freight demand in the Hainan region and provides a reference for optimizing regional logistics network layout and formulating transportation policies.

In the future, during the long transition period in which regular human-driven vehicles and connected and autonomous vehicles coexist, conflicts between the two types of vehicles could hinder traffic efficiency. To address this challenge, the optimization problem of deploying dedicated lanes and toll lanes in the road network has been proposed. Due to the complexity and scale of the problem, much research relies on heuristic algorithms to identify the optimal solution. However, heuristic methods’ lack of stability and convergence results in only near-optimal solutions, which limits the capacity of connected and autonomous vehicles to boost the road network’s performance. Therefore, this study focuses on the optimization problem of deploying dedicated lanes and toll lanes, where the route decisions of diverse vehicle types are considered and modeled based on the user equilibrium principle. The objective is to reduce total travel time by optimizing the spatial arrangement of dedicated lanes and toll lanes within the road network, and by setting tolls for human-driven vehicles on toll lanes. By link-node modeling, the optimization problem is subsequently constructed into a mixed-integer nonlinear programming model, which circumvents the need for time-intensive path enumeration. The global optimization algorithm, along with the outer approximation method and other linearization methods, is employed to achieve the global optimum. Finally, numerical experiments conducted on the Nguyen-Dupuis network and the Sioux-Falls network illustrate the effectiveness of the proposed model and algorithm. The results show that the proposed algorithm achieves the globally optimal solution, which demonstrates a 7.51% reduction in total travel time compared to the initial feasible deployment scheme of dedicated lanes and toll lanes. Furthermore, sensitivity analyses based on the proportion of connected and autonomous vehicles offer valuable insights for traffic planners to make informed decisions.

With the end of the COVID-19 pandemic, the tourism industry has experienced a rapid recovery. However, traditional research often overlooks changes in tourists' travel habits and preferences during this rapid resurgence. This study takes Hainan Island as the research area and uses mobile phone signaling data, combined with complex network analysis and community detection algorithms, to analyze the dynamic evolution of tourism travel characteristics in Hainan Island from the pandemic period to the post-pandemic era. The study finds that Hainan Island's tourism industry has recovered rapidly after the pandemic, with a sharp increase in the number of visitors to scenic spots. Post-pandemic tourists have shifted their preferences from natural and cultural scenic spots to comprehensive scenic areas, and the tendency to visit multiple scenic spots has become more evident. From a network perspective, the area around Sanya has emerged as a high-weighted hub in the tourism network, while Haikou's core position in the tourism network has declined. The structure of tourism communities exhibits dynamic evolution, with an increase in the number of communities but a decrease in modularity after the pandemic. Based on these findings, suggestions are put forward to integrate regional tourism resources, develop comprehensive tourism routes, and deepen regional tourism cooperation with Sanya and Haikou as cores. This study can provide a theoretical basis for optimizing tourism operation and management in Hainan Island and offer guidance for promoting post-pandemic tourism development.

In commercial complexes, the functional interwoven areas (FIA) with multiple attributes shops accommodate diverse population of pedestrians with heterogeneous travel purposes. A well-designed layout scheme of FIA can effectively increase the revenue while reduce the pedestrian congestion. However, existing layout schemes are more developed from the operator’s perspective, focusing on maximizing rents while neglecting shop attributes and pedestrians heterogeneity, which results in excessive corridor congestion and suboptimal store visit rates. To address these issues, this article proposes a novel macro-to-micro layout optimization framework that enhances pedestrian in-store rate while mitigating internal congestion. The proposed framework consists of two stages: 1) a macrolevel optimization model which generates multiple feasible layout schemes by using shop attributes based attractiveness algorithm; and 2) a microlevel simulation model which evaluates pedestrian movement dynamics and corridor density under different layouts by integrating an attractive potential based social force model (AP-SFM). Based on parameters calibrated from real collected data, we conduct extensive numerical and simulation experiments across two distinct commercial complexes: 1) China World Mall (CWM); and 2) Shin Kong Place (SKP). The results demonstrate that the optimized layouts significantly increase pedestrian in-store rates while effectively reducing average corridor density and alleviating local congestion, without altering the total pedestrian volume. Furthermore, the successful application in two distinct commercial complexes proves the validity, rationality, and generality of the proposed framework. Our research effectively integrates shop diversity and pedestrian heterogeneity into shop layout optimization, providing a practical and scalable decision-support tool for layout design and management.

Autonomous vehicles operating at signalized intersections face fundamental challenges arising from queue dynamics, signal-phase transitions, and tightly coupled multi-vehicle interactions. Conventional motion-planning methods, which rely primarily on instantaneous perception, are inherently reactive and struggle to reason about short-term traffic evolution. This paper presents a Transformer-based predictive motion-planning framework that embeds short-term traffic state prediction directly into the structure of the planning problem. A lightweight spatial–temporal Transformer model is designed to forecast traffic occupancy, queue evolution, and interaction patterns using historical trajectories, signal-phase information, and road topology. By converting predicted traffic dynamics into explicit spatial–temporal constraints, a hierarchical motion planner jointly optimizes path geometry and speed profiles through dynamically constructed feasible corridors. The proposed framework is evaluated using a joint SUMO–CARLA simulation platform under realistic traffic conditions derived from real-world datasets, including pNEUMA and CitySim. The experimental results across straight-through, queueing, and turning scenarios show that prediction-aware planning significantly reduces high-risk driving time and intersection travel time while maintaining stable real-time computational performance. Beyond scenario-level improvements, the results indicate that transforming traffic prediction into planning constraints provides a generalizable paradigm for proactive, feasibility-aware autonomous driving at signalized intersections. From a methodological perspective, the proposed framework can be interpreted through the lens of symmetry and asymmetry in intelligent transportation systems: the conventional symmetric decoupling between prediction and planning modules is deliberately broken by embedding predicted traffic states as time-varying, directionally asymmetric constraints, while the permutation symmetry of the multi-head attention mechanism is preserved over lane-segment tokens to provide a structured inductive bias for traffic state forecasting. This symmetry-aware design highlights how controlled symmetry breaking in modeling and optimization can yield safer, more efficient, and more adaptive autonomous driving behaviors in signalized urban environments.

Driving risk assessment in tunnels is not only fundamental for evaluating traffic safety but also a critical prerequisite for analyzing and understanding the formation mechanisms of vehicle operational risks. Based on naturalistic driving trajectory data of the Xiaxiyao tunnel of Wuyu freeway from the Tongji Road Trajectory Sharing Platform (TJRD TS), this study extracts 19 driving behavior parameters and classifies drivers into four categories using the PCA-K-medoids method: conservative, moderate, aggressive-a, and aggressive-b. Considering six dimensions (safety speed difference, driving instability, time headway, acceleration, speed difference, and acceleration difference), the driving risk assessment model of the freeway tunnel is constructed by determining the weight by the entropy weight-CRITIC method. The model analyzes the risk evolution of different driver types during travel. The results indicate that conservative and moderate drivers typically maintain larger time headways and following distances. The proportions of these drivers whose peak risk values remain at no risk or low risk level are 85.95% and 76.79%, respectively. Regarding driving risks in different zones, aggressive-a drivers exhibit higher risk levels from 400 m before the tunnel exit to 300 m after the exit, whereas aggressive-b drivers demonstrate an increased risk propensity from 200 m before the entrance and 300 m after the entrance. Further behavioral analysis highlights that aggressive-a drivers exhibit higher acceleration, maximum speed, and speed differences with leading vehicles, and aggressive-b drivers demonstrate higher deceleration. During speeding, aggressive-b drivers show the highest average driving risk index (0.28), with 65.34% of cases classified as medium- or high-risk. Under low-speed conditions, aggressive-a and aggressive-b drivers exhibit medium- or high-risk proportions of 62.16% and 57.67%, respectively. Compared with moderate drivers, aggressive-a drivers experience a 22.7% increase in average driving risk index during rapid acceleration, whereas aggressive-b drivers show a 15.4% increase. The probability of aggressive-a and aggressive-drivers being at moderate and high risk levels in the car-following conditions was 58.41% and 63.74%, respectively. This study reveals the risk characteristics of different driving styles in tunnels, providing valuable insights for precise vehicle management, hazard mitigation, and proactive safety governance in tunnel environments.

Compared to other demographic groups, elderly individuals exhibit unique travel characteristics. They often face challenges associated with extended walking distances, encounter difficulties in navigating overcrowded public transportation, and have limited options for travel destinations. These factors frequently limit the elderly's choice of public transport options; consequently, it is crucial to consider these elements when assessing the dynamic accessibility of public transport in order to develop age-friendly public transport services. However, existing studies on accessibility frequently fail to capture these specificities. To address these gaps, this study integrates travel purposes and end-to-end travel into the dynamic accessibility evaluation for elderly public transport. Specifically, factors such as walking time, waiting time, in-vehicle travel time, bus load factor, and transfer time are taken into account in the context of end-to-end travel due to their significant impact on elderly individuals. Accessibility is assessed through a weighted summation of the number of points of interest within different time-based bus accessibility ranges, while considering the limited categories of points of interest that align with the specific travel needs of elderly individuals. In the case study of the core area of Beijing, the findings reveal that accessibility is significantly lower during peak hours on weekdays compared to other daytime periods. Furthermore, accessibility for elderly individuals who are picking up their grandchildren from school is notably reduced in comparison to other travel purposes. This study can provide a reference for the development and age-friendly improvement of public transportation systems.

The rapid growth of electric bicycles (e-bikes) has raised increasing safety concerns in urban environments, where e-bike users account for a substantial share of non-motorized traffic casualties. While risk perception and timely braking are critical for conflict avoidance, limited evidence exists on how braking behavior varies across riding contexts and rider groups. This study investigates e-bike braking as a multi-stage behavioral process encompassing risk perception, decision-making, and execution. Data were collected from 66 e-bike riders in Beijing through a controlled urban riding experiment. Head movements, speed, and braking behavior were recorded. A deep embedded multi-view clustering (DEMVC) approach was used to classify complete braking chains into four distinct patterns, and a partially constrained random parameters logit model was applied to examine factors influencing pattern occurrence. Results indicate that roadway characteristics and rider attributes significantly shape braking behavior. Delivery riders were more likely to adopt decisive braking responses, whereas non-delivery riders showed more cautious or delayed braking tendencies. Longer daily riding duration was associated with more anticipatory and stable braking, while slower riders exhibited delayed risk perception and abrupt responses. Intersections, physical separation, age, and gender were also found to influence braking patterns. The findings provide empirical evidence to support targeted infrastructure design and behaviororiented safety interventions aimed at improving electric micromobility safety.