Journal of Advanced Transportation最新文献

Traffic scene understanding plays a crucial role in reasoning about and predicting relationships among entities in traffic images. It focuses on analyzing behavioral interaction patterns and global semantic associations to support higher-level traffic requirements. However, few existing frameworks can achieve comprehensive scene understanding and semantic description in complex traffic environments. In particular, effective multiview semantic association modeling is still lacking. To address these challenges, we propose multiview large language model (MVLLM), which integrates YOLO-based object detection with the reasoning ability of large language models (LLMs). Through prompt engineering, MVLLM utilizes the visual information extracted by YOLO to constrain the semantic space and guide the reasoning behavior, thereby enhancing the scene parsing capability. Meanwhile, we design a Chain-of-Thought (CoT) reasoning mechanism to establish spatiotemporal associations across multiple views and to integrate their scene understanding with semantic descriptions. The framework enables intent understanding for vehicles in dynamic environments, enhancing driving safety. It also provides comprehensive semantic descriptions for traffic management agencies, supporting holistic analyses of vehicles, roads, and environmental contexts.

The study of accompanying vehicles is a hot topic in the field of intelligent transportation. Because of the multiple selectivity of the traffic path and the loss of sampling in traditional companion vehicles discovery, the method based on path similarity mining will result in the omission of the companion candidates. This paper recognizes that the upstream and downstream relevance of trajectory intentions in traffic is similar to the contextual relevance of text semantics, as inspired by the semantic similarity of texts. Simultaneously, taking into account the generalization and tolerance of semantic processing, a companion vehicle discovery method based on “trajectory semantics” similarity is proposed. First, this paper proposes a trajectory semantic vectorized representation method trajectory semantic to vector (TS2vec), which realizes the low-dimensional dense vectorization of the trajectory in the context of dynamic time slicing of the trajectory, fusion of the temporal and spatial characteristics of the trajectory, and text information. Then, based on the “trajectory pair,” this paper proposes the trajectory pair bidirectional GRU (TPBi-GRU) model. This paper constructs forward and reverse subnetworks using the trajectory pair set—which is made up of the actual trajectory and ts sampled trajectories—realizes parameter transfer and contribution during training; gains a thorough understanding of trajectory semantics; and mines the internal relationship between vehicles more effectively. Finally, given the difference in the degree of contribution of the road shape in forming the adjoint pattern, and the sensitivity of the attention mechanism to local features, the attention mechanism is used to weigh the key nodes that affect the trajectory shape in order to obtain a more accurate trajectory representation. The experimental results show that the method in this paper can discover local and overall concomitant patterns more effectively and effectively overcome the interference of multiple selectivity of traffic paths on concomitant pattern mining.

Understanding the relationship between travel behavior and modifiable built environment attributes is essential for promoting low-carbon urban mobility, particularly under emerging carbon peaking and neutrality targets. While previous studies have explored this relationship, limited attention has been paid to residents’ intentions for low-carbon travel modes. To address this gap, this study employs large-scale, anonymized map usage data from Beijing and applies a gradient boosting decision trees (GBDT) model to examine the nonlinear and interaction effects of built environment attributes on behavioral intentions at both trip origins and destinations. The results indicate that destination road density exerts the strongest influence on low-carbon mode choices, whereas factors such as scenery density and residential density display notable threshold effects. Furthermore, strong interaction effects between residential density and living service density highlight the importance of integrated urban planning to facilitate sustainable mobility. Model validation demonstrates that the GBDT approach outperforms both random forest and multinomial logit models, achieving superior predictive accuracy (85.7%) and effectively capturing complex nonlinear relationships. These findings offer actionable insights for policymakers: interventions should prioritize enhancing road network density up to 18.5 km/km², fostering medium-density residential areas (10–35 units/km²), and integrating comprehensive living services within neighborhoods. Overall, this study contributes a reliable, data-driven evidence base to inform targeted urban transport planning and land-use management for advancing low-carbon urban development.

Universities play a crucial role in alleviating students’ financial burdens to ensure that the cost of education remains manageable. Parking fines, though often overlooked, contribute to these ancillary costs. While existing literature explores the monetary effects and compliance rates of digital and physical ticketing systems, a significant gap remains in understanding how these methods specifically affect university settings. This study aims to fill that gap by assessing the effectiveness of ticketing practices in reducing parking violations on university campuses, with a focus on the role of warning tickets in promoting compliance and the financial implications of transitioning from digital to physical ticketing methods. The methodology involved comprehensively analyzing 5 years of parking violation data collected from a university campus, applying chi-square and two-sample z-tests, and developing a random forest model. The results show that warning tickets significantly reduce the incidence of repeat violations, making them an effective nonpunitive strategy. Additionally, the transition from digital to physical ticketing methods led to a reduction in multiple violations and a decrease in the average cost per violator by $25. Physical tickets were found to have a stronger deterrent effect due to their immediacy and visibility. The study provides an evidence-based decision framework to help universities calibrate enforcement design choices under budget and equity constraints.

Autonomous trucks in busy port terminals must navigate narrow aisles, tight corners, and frequent interactions with multiple vehicles while maintaining both safety and efficiency. This paper presents a hierarchical multiagent navigation framework that integrates an enhanced grid-based Theta^∗ global planner with obstacle inflation and clearance-aware costs, an artificial potential field (APF)–based local controller augmented by lightweight neural correction, and a simple coordination protocol for resolving intertruck conflicts. We evaluate the approach in a high-fidelity Unity digital twin of the Port of Oulu using two traffic scenes with three trucks executing simultaneous tasks. Experiments are repeated under identical initial conditions with independent random perturbations to capture run-to-run variability, and results are reported as the mean ± standard deviation. We compare the proposed Theta^∗-based planner with a standard grid-based A^∗ baseline and an 8-neighborhood A^∗ variant under the same occupancy grid, obstacle inflation, and curvature constraints to isolate the impact of expanded action sets within the A^∗ framework. A greedy heuristic baseline is also included in the simpler scene, where it can complete scheduling. Across trucks, Theta^∗ achieves 43.0% lower travel time and 39.4% fewer avoidance events than A^∗ in the dense-yard scene and 59.5% lower travel time and 91.4% fewer avoidance events in the gate–yard scene, while also improving a combined tracking-accuracy index by 22.1% and 12.7%, respectively. Path-tracking evaluation shows stable mean errors (average mean lateral deviation ≈ 0.40 m and mean heading error ≈ 1.69° across trucks), with transient peaks mainly occurring at high-curvature segments, narrow-clearance passages, and interaction-driven maneuvers. We further include a time-bounded scalability study by increasing the local fleet size to assess the coordination overhead under denser intertruck interactions. These results indicate that clearance-aware any-angle planning, together with neural-tuned local avoidance and lightweight coordination, can improve both efficiency and execution quality for port–yard truck autonomy.

On-ramp merging is a common expressway maneuver for connected and automated vehicles (CAVs), where trajectory planning and tracking control are central to avoiding collisions. However, existing studies rarely optimize the selection of merge start and end points and give limited attention to constraints from acceleration-lane length. This study proposes a structured trajectory planning and tracking method with multiobjective optimization under the CAV’s environment. First, by sampling the starting and ending points of the merging process, the quintic polynomial is used to plan the initial trajectory of the merging vehicles, and trajectory safety is checked with a collision-avoidance algorithm based on rectangular vehicle geometry. Then, a multiobjective optimization model selects the on-ramp trajectory by balancing merging urgency, driving safety, traffic efficiency, and comfort. Finally, an integrated tracking strategy combines lateral and longitudinal control: a feedforward LQR for lateral motion and a PID-based longitudinal controller. To further improve the tracking accuracy, the particle swarm algorithm tunes key parameters of the lateral LQR controller. The numerical result demonstrates that the planner can generate smooth and stable trajectories that could be selected as an optimal reference for the tracking controller. The simulation results show that when the initial speed of the on-ramp vehicle is 68 km/h, the maximum tracking errors of lateral and longitudinal displacements are less than 0.02 and 0.2 m, respectively.

Electric vehicles (EVs) are increasingly becoming the main mobility option, with global sales in 2023 driven predominantly by China (60%), Europe (25%) and the United States (10%). This widespread adoption has intensified demand for EV batteries, which exceeded 750 GWh in 2023, marking a 40% year-on-year increase. As EVs expand in number, battery performance and reuse emerge as pivotal challenges in the transition to sustainable mobility. This study presents a systematic literature review focussing on the performance evaluation and life prediction of EV batteries, with a primary emphasis on lithium-ion technology, which is the more common technologies. Using a bibliometric clustering approach by CiteSpace©, the review identifies key research domains and highlights critical gaps, especially the underrepresentation of real-world usage conditions in performance models. This review integrates findings across diagnostic, modelling and field-performance studies to propose a unified framework for evaluating EV battery performance under dynamic operating conditions. Findings reveal that while numerous diagnostic tools and modelling techniques exist, their applicability to dynamic operational contexts remains limited. The results aim to support future innovation in battery management systems (BMSs), second-life strategies and potentially circular economy applications, providing both a scientific foundation and practical guidance for advancing sustainable energy solutions.

The driving safety field (DSF) model, which comprehensively evaluates driving risks in complex environments by integrating human–vehicle–road factors, serves as a quantitative methodology for assessing dynamic traffic risks. However, it exhibits limitations in certain scenarios where its risk characterization deviates from actual risk variations. To address the challenges, an improved driving safety field (IDSF) model is proposed. The new framework redesigns the calculation of virtual mass, field force, and driving safety index. Parameters in the model were calibrated using accident data and driving simulator experiments. Results demonstrate that the IDSF outperforms conventional time-to-collision (TTC) inverse (TTCi) and DSF models. Specifically, in car-following scenarios, IDSF demonstrates higher correlation (r≈0.9) with the TTCi model. In complex environments with high vehicular heterogeneity, compared with the DSF model, the IDSF model exhibits greater stability (80% lower coefficient of variation) and fewer extreme deviations (38% reduction). This study provides a novel theoretical framework for automotive intelligent safety technologies and offers valuable insights for designing more reasonable driving safety algorithms.

Construction waste hauling (CWH) trucks are a significant source of air pollution and particulate emissions in urban environments, prompting strict regulatory controls and monitoring. Accurate prediction of their transportation activities, including destinations and arrival times, is critical for improving environmental management and regulatory enforcement. In this study, we present a probabilistic approach that captures the complex spatiotemporal dynamics inherent in the transportation behavior of CWH trucks using the input–output hidden Markov model (IOHMM). This model leverages contextual factors such as historical trajectories, weather conditions, and time-based patterns to make real-time predictions of transportation activities with high accuracy. The model is applied to a dataset of 1000 CWH trucks collected over a 5-month period in Chengdu, China. The model’s performance was evaluated against several baseline methods, including traditional Markov chains, long short-term memory (LSTM) networks, and DeepMove, an attention-based deep learning model. Results demonstrated that the IOHMM outperforms these models in both prediction accuracy and interpretability. Specifically, the IOHMM achieved an average destination prediction accuracy of 51.2%, compared to 47.9% for DeepMove, 43.1% for LSTM, and 39.4% for Markov chains. In terms of arrival time prediction, the IOHMM obtained an accuracy of 38.8%, outperforming all other models, with DeepMove at 36.8%, LSTM at 35.6%, and Markov chains at 27.5%. These findings highlight the IOHMM’s ability to effectively incorporate both spatial and temporal factors in predicting transportation dynamics, providing a powerful tool for regulatory agencies to improve real-time interventions and environmental management of heavy-duty vehicles.

This study investigates the temporal evolution of nonmandatory trip frequencies in Shanghai over a decade using a temporally adaptive multivariate ordered probit (MOP) model. Two large-scale travel surveys are pooled, and temporal changes are captured through year dummy interaction terms, year-specific threshold shifts, and a year-specific correlation structure. Parameters are estimated using full-information maximum likelihood estimation with an analytic approximation of multivariate normal cumulative distribution. The findings reveal substantial decade-long transformations in discretionary mobility. Gender differences narrowed or reversed across several activities; the impact of aging was apparent; occupational constraints persisted; the influence of central-area residence intensified, reflecting uneven spatial development; and weekend effects weakened, indicating increasingly blurred boundaries between weekday and weekend activity patterns. Correlation patterns across activities also shifted, suggesting changes in trip chaining and time allocation. By developing a unified, temporally adaptive MOP framework capable of jointly capturing structural stability and temporal change, this study provides new empirical evidence on how nonmandatory trip adapts to rapid sociodemographic, economic, and spatial transformations. It offers rare evidence from a major megacity of developing country where activity-based modeling applications remain limited. These insights support the design of context-sensitive transportation and land-use policies.