Transportation Research Part C-Emerging Technologies最新文献_第7页

Which type of backpressure is more stable? – Comparative analysis based on two-movement intersections 哪种类型的背压更稳定？-基于双运动交叉口的对比分析

IF 7.6 1区工程技术 Q1 TRANSPORTATION SCIENCE & TECHNOLOGY

Transportation Research Part C-Emerging Technologies

Pub Date : 2025-12-03 DOI: 10.1016/j.trc.2025.105478

Dianchao Lin , Li Li

Stability, which indicates that queues do not grow infinitely over time, is a key concept in control policies such as BackPressure (BP). However, its abstract nature and diverse definitions make its comparative analysis difficult both theoretically and experimentally. As a result, simulations in existing studies often use alternative metrics, such as average delay, to evaluate the performance of different control policies. Little research directly compares different stabilities through theory or experiments. In this paper, we compare seven common stability definitions and theoretically demonstrates that they are equivalent in simulations and applications. Furthermore, we propose a t-test method for identifying whether a queue is stable based on the sequence of queueing differences. This method allows us to classify any sampled demand as stable or unstable based on simulated queues for a given control policy. Therefore, if the network’s dimension, i.e., the number of movements, does not exceed three, we can directly draw the stability region (SR) for all policies and compare their sizes. To accurately reproduce various BP theories, ensure fair comparisons, and facilitate the visualization of SRs, we use simulation codes to simulate a two-movement intersection scenario and discuss its extension to networks. Six distinct types of BP policies are compared, along with analysis for fixed-time and actuated controls. We obtain many insights that are difficult to achieve through purely theoretical analysis and delay-based simulations, including: 1) variability in BP’s SR: the SR typically varies when the BP changes its queue status weight or efficiency weight; 2) size hierarchy of SR: BPs generally outperform actuated controls in terms of SR, and actuated controls tend to outperform fixed-time controls; 3) non-cyclic vs. cyclic BP: non-cyclic BP usually has a larger SR than cyclic BP; 4) effect of real-time supply information: using real-time supply increases the SR of BP, even under the assumption of fixed saturation headway; and 5) SR degradation phenomenon: longer cycle lengths in cyclic BP may cause its SR to degenerate into a rectangular shape typical of fixed-time control.

稳定性是BP等控制策略中的一个关键概念，它表明队列不会随时间无限增长。然而，它的抽象性和定义的多样性使得对其进行比较分析在理论上和实验上都很困难。因此，现有研究中的仿真通常使用替代度量，例如平均延迟，来评估不同控制策略的性能。很少有研究直接通过理论或实验来比较不同的稳定性。本文比较了7种常用的稳定性定义，并从理论上证明了它们在仿真和应用上是等价的。在此基础上，提出了一种基于排队差异序列来判别队列是否稳定的t检验方法。该方法允许我们根据给定控制策略的模拟队列将任何抽样需求分类为稳定或不稳定。因此，如果网络的维度，即运动次数不超过3次，我们可以直接绘制所有策略的稳定区域（SR），并比较它们的大小。为了准确再现各种BP理论，确保公平的比较，并促进SRs的可视化，我们使用仿真代码模拟了两运动交叉场景，并讨论了其在网络中的扩展。对六种不同类型的BP策略进行了比较，并对固定时间和驱动控制进行了分析。我们获得了许多通过纯理论分析和基于延迟的模拟难以获得的见解，包括：1)BP SR的可变性：当BP改变其队列状态权重或效率权重时，SR通常会发生变化；2) SR的大小层次：bp在SR方面普遍优于驱动控制，而驱动控制往往优于固定时间控制；3)非环BP与环BP：非环BP的SR通常大于环BP；4)实时供给信息效应：即使在饱和车头距固定的情况下，使用实时供给也会增加BP的SR；5) SR退化现象：周期BP中较长的周期长度可能导致其SR退化为典型的固定时间控制的矩形。

{"title":"Which type of backpressure is more stable? – Comparative analysis based on two-movement intersections","authors":"Dianchao Lin , Li Li","doi":"10.1016/j.trc.2025.105478","DOIUrl":"10.1016/j.trc.2025.105478","url":null,"abstract":"<div><div>Stability, which indicates that queues do not grow infinitely over time, is a key concept in control policies such as BackPressure (BP). However, its abstract nature and diverse definitions make its comparative analysis difficult both theoretically and experimentally. As a result, simulations in existing studies often use alternative metrics, such as average delay, to evaluate the performance of different control policies. Little research directly compares different stabilities through theory or experiments. In this paper, we compare seven common stability definitions and theoretically demonstrates that they are equivalent in simulations and applications. Furthermore, we propose a <em>t</em>-test method for identifying whether a queue is stable based on the sequence of queueing differences. This method allows us to classify any sampled demand as stable or unstable based on simulated queues for a given control policy. Therefore, if the network’s dimension, i.e., the number of movements, does not exceed three, we can directly draw the stability region (SR) for all policies and compare their sizes. To accurately reproduce various BP theories, ensure fair comparisons, and facilitate the visualization of SRs, we use simulation codes to simulate a two-movement intersection scenario and discuss its extension to networks. Six distinct types of BP policies are compared, along with analysis for fixed-time and actuated controls. We obtain many insights that are difficult to achieve through purely theoretical analysis and delay-based simulations, including: 1) variability in BP’s SR: the SR typically varies when the BP changes its queue status weight or efficiency weight; 2) size hierarchy of SR: BPs generally outperform actuated controls in terms of SR, and actuated controls tend to outperform fixed-time controls; 3) non-cyclic vs. cyclic BP: non-cyclic BP usually has a larger SR than cyclic BP; 4) effect of real-time supply information: using real-time supply increases the SR of BP, even under the assumption of fixed saturation headway; and 5) SR degradation phenomenon: longer cycle lengths in cyclic BP may cause its SR to degenerate into a rectangular shape typical of fixed-time control.</div></div>","PeriodicalId":54417,"journal":{"name":"Transportation Research Part C-Emerging Technologies","volume":"183 ","pages":"Article 105478"},"PeriodicalIF":7.6,"publicationDate":"2025-12-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145657758","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

V2X-VLM: End-to-End V2X cooperative autonomous driving through large vision-Language models V2X- vlm：基于大型视觉语言模型的端到端V2X协同自动驾驶

IF 7.6 1区工程技术 Q1 TRANSPORTATION SCIENCE & TECHNOLOGY

Transportation Research Part C-Emerging Technologies

Pub Date : 2025-11-30 DOI: 10.1016/j.trc.2025.105457

Junwei You , Zhuoyu Jiang , Zilin Huang , Haotian Shi , Rui Gan , Keshu Wu , Xi Cheng , Xiaopeng Li , Bin Ran

Vehicle-to-everything (V2X) cooperation has emerged as a promising paradigm to overcome the perception limitations of classical autonomous driving by leveraging information from both ego-vehicle and infrastructure sensors. However, effectively fusing heterogeneous visual and semantic information while ensuring robust trajectory planning remains a significant challenge. This paper introduces V2X-VLM, a novel end-to-end (E2E) cooperative autonomous driving framework based on vision-language models (VLMs). V2X-VLM integrates multiperspective camera views from vehicles and infrastructure with text-based scene descriptions to enable a more comprehensive understanding of driving environments. Specifically, we propose a contrastive learning-based mechanism to reinforce the alignment of heterogeneous visual and textual characteristics, which enhances the semantic understanding of complex driving scenarios, and employ a knowledge distillation strategy to stabilize training. Experiments on a large real-world dataset demonstrate that V2X-VLM achieves state-of-the-art trajectory planning accuracy, significantly reducing L2 error and collision rate compared to existing cooperative autonomous driving baselines. Ablation studies validate the contributions of each component. Moreover, the evaluation of robustness and efficiency highlights the practicality of V2X-VLM for real-world deployment to enhance overall autonomous driving safety and decision-making.

车联网（V2X）合作已经成为一种很有前途的模式，通过利用车辆和基础设施传感器的信息来克服传统自动驾驶的感知限制。然而，在保证鲁棒轨迹规划的同时，如何有效地融合异构视觉和语义信息仍然是一个重大挑战。介绍了一种基于视觉语言模型（VLMs）的端到端协作式自动驾驶框架V2X-VLM。V2X-VLM将车辆和基础设施的多视角摄像头视图与基于文本的场景描述集成在一起，从而能够更全面地了解驾驶环境。具体而言，我们提出了一种基于对比学习的机制来加强异构视觉和文本特征的一致性，从而增强对复杂驾驶场景的语义理解，并采用知识蒸馏策略来稳定训练。在大型真实数据集上的实验表明，与现有的协作式自动驾驶基线相比，V2X-VLM实现了最先进的轨迹规划精度，显著降低了L2误差和碰撞率。消融研究证实了每个组成部分的贡献。此外，对鲁棒性和效率的评估凸显了V2X-VLM在实际部署中的实用性，以提高整体自动驾驶安全性和决策能力。

{"title":"V2X-VLM: End-to-End V2X cooperative autonomous driving through large vision-Language models","authors":"Junwei You , Zhuoyu Jiang , Zilin Huang , Haotian Shi , Rui Gan , Keshu Wu , Xi Cheng , Xiaopeng Li , Bin Ran","doi":"10.1016/j.trc.2025.105457","DOIUrl":"10.1016/j.trc.2025.105457","url":null,"abstract":"<div><div>Vehicle-to-everything (V2X) cooperation has emerged as a promising paradigm to overcome the perception limitations of classical autonomous driving by leveraging information from both ego-vehicle and infrastructure sensors. However, effectively fusing heterogeneous visual and semantic information while ensuring robust trajectory planning remains a significant challenge. This paper introduces V2X-VLM, a novel end-to-end (E2E) cooperative autonomous driving framework based on vision-language models (VLMs). V2X-VLM integrates multiperspective camera views from vehicles and infrastructure with text-based scene descriptions to enable a more comprehensive understanding of driving environments. Specifically, we propose a contrastive learning-based mechanism to reinforce the alignment of heterogeneous visual and textual characteristics, which enhances the semantic understanding of complex driving scenarios, and employ a knowledge distillation strategy to stabilize training. Experiments on a large real-world dataset demonstrate that V2X-VLM achieves state-of-the-art trajectory planning accuracy, significantly reducing L2 error and collision rate compared to existing cooperative autonomous driving baselines. Ablation studies validate the contributions of each component. Moreover, the evaluation of robustness and efficiency highlights the practicality of V2X-VLM for real-world deployment to enhance overall autonomous driving safety and decision-making.</div></div>","PeriodicalId":54417,"journal":{"name":"Transportation Research Part C-Emerging Technologies","volume":"183 ","pages":"Article 105457"},"PeriodicalIF":7.6,"publicationDate":"2025-11-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145651056","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Improving traffic signal data quality for the Waymo open motion dataset 改善Waymo开放运动数据集的交通信号数据质量

IF 7.6 1区工程技术 Q1 TRANSPORTATION SCIENCE & TECHNOLOGY

Transportation Research Part C-Emerging Technologies

Pub Date : 2025-11-30 DOI: 10.1016/j.trc.2025.105476

Xintao Yan , Erdao Liang , Jiawei Wang , Haojie Zhu , Henry X. Liu

Datasets pertaining to autonomous vehicles (AVs) hold significant promise for a range of research fields, including artificial intelligence (AI), autonomous driving, and transportation engineering. Nonetheless, these datasets often encounter challenges related to the states of traffic signals, such as missing or inaccurate data. Such issues can compromise the reliability of the datasets and adversely affect the performance of models developed using them. This research introduces a fully automated approach designed to tackle these issues by utilizing available vehicle trajectory data alongside knowledge from the transportation domain to effectively impute and rectify traffic signal information within the Waymo Open Motion Dataset (WOMD). The proposed method is robust and flexible, capable of handling diverse intersection geometries and traffic signal configurations in real-world scenarios. Comprehensive validations have been conducted on the entire WOMD, focusing on over 360,000 relevant scenarios involving traffic signals, out of a total of 530,000 real-world driving scenarios. In the original dataset, 71.7 % traffic signal states are either missing or unknown, all of which were successfully imputed by our proposed method. Furthermore, in the absence of ground-truth signal states, the accuracy of our approach is evaluated based on the rate of red-light violations among vehicle trajectories. Results show that our method reduces the estimated red-light running rate from 15.7 % in the original data to 2.9 %, thereby demonstrating its efficacy in rectifying data inaccuracies. This paper significantly enhances the quality of AV datasets, contributing to the wider AI and AV research communities and benefiting various downstream applications. The code and improved traffic signal data are open-sourced at: https://github.com/michigan-traffic-lab/WOMD-Traffic-Signal-Data-Improvement.

与自动驾驶汽车（AVs）相关的数据集在人工智能（AI）、自动驾驶和交通工程等一系列研究领域具有重大前景。然而，这些数据集经常遇到与交通信号状态相关的挑战，例如数据缺失或不准确。这些问题可能会损害数据集的可靠性，并对使用它们开发的模型的性能产生不利影响。该研究引入了一种完全自动化的方法，旨在通过利用可用的车辆轨迹数据以及来自交通领域的知识来有效地在Waymo开放运动数据集（WOMD）中输入和校正交通信号信息，从而解决这些问题。该方法具有鲁棒性和灵活性，能够处理现实场景中不同的交叉口几何形状和交通信号配置。在53万个真实驾驶场景中，重点对36万个涉及交通信号的相关场景进行了全面验证。在原始数据集中，71.7%的交通信号状态缺失或未知，我们的方法都成功地估算了这些状态。此外，在没有真实信号状态的情况下，我们的方法的准确性是基于车辆轨迹之间的红灯违例率来评估的。结果表明，该方法将估计的红灯运行率从原始数据的15.7%降低到2.9%，从而证明了该方法在纠正数据不准确性方面的有效性。本文显著提高了自动驾驶汽车数据集的质量，为更广泛的人工智能和自动驾驶研究社区做出了贡献，并使各种下游应用受益。代码和改进的交通信号数据在https://github.com/michigan-traffic-lab/WOMD-Traffic-Signal-Data-Improvement上开源。

{"title":"Improving traffic signal data quality for the Waymo open motion dataset","authors":"Xintao Yan , Erdao Liang , Jiawei Wang , Haojie Zhu , Henry X. Liu","doi":"10.1016/j.trc.2025.105476","DOIUrl":"10.1016/j.trc.2025.105476","url":null,"abstract":"<div><div>Datasets pertaining to autonomous vehicles (AVs) hold significant promise for a range of research fields, including artificial intelligence (AI), autonomous driving, and transportation engineering. Nonetheless, these datasets often encounter challenges related to the states of traffic signals, such as missing or inaccurate data. Such issues can compromise the reliability of the datasets and adversely affect the performance of models developed using them. This research introduces a fully automated approach designed to tackle these issues by utilizing available vehicle trajectory data alongside knowledge from the transportation domain to effectively impute and rectify traffic signal information within the <span>Waymo Open Motion Dataset</span> (WOMD). The proposed method is robust and flexible, capable of handling diverse intersection geometries and traffic signal configurations in real-world scenarios. Comprehensive validations have been conducted on the entire WOMD, focusing on over 360,000 relevant scenarios involving traffic signals, out of a total of 530,000 real-world driving scenarios. In the original dataset, 71.7 % traffic signal states are either missing or unknown, all of which were successfully imputed by our proposed method. Furthermore, in the absence of ground-truth signal states, the accuracy of our approach is evaluated based on the rate of red-light violations among vehicle trajectories. Results show that our method reduces the estimated red-light running rate from 15.7 % in the original data to 2.9 %, thereby demonstrating its efficacy in rectifying data inaccuracies. This paper significantly enhances the quality of AV datasets, contributing to the wider AI and AV research communities and benefiting various downstream applications. The code and improved traffic signal data are open-sourced at: <span><span>https://github.com/michigan-traffic-lab/WOMD-Traffic-Signal-Data-Improvement</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":54417,"journal":{"name":"Transportation Research Part C-Emerging Technologies","volume":"183 ","pages":"Article 105476"},"PeriodicalIF":7.6,"publicationDate":"2025-11-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145651055","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Integrated optimization of ride-pooling and shared micro-mobility services with meeting points 带集合点的拼车、共享微出行服务集成优化

IF 7.6 1区工程技术 Q1 TRANSPORTATION SCIENCE & TECHNOLOGY

Transportation Research Part C-Emerging Technologies

Pub Date : 2025-11-29 DOI: 10.1016/j.trc.2025.105452

Tingting Dong , Yulong Hu , Zemin Wang , Sen Li

This paper studies the integrated optimization of ride-pooling and micro-mobility services within a multimodal transportation network to enhance the likelihood of ride-pooling, reduce the waiting time for passengers, and improve the platform’s profit. We consider an integrated platform that simultaneously provides ride-pooling and micro-mobility services. Instead of door-to-door ride-pooling services, we consider inter-modal transfer at meeting points, where riders can access or egress nearby pickup or dropoff (PUDO) points using micro-mobility vehicles and then get picked up and dropped off at these meeting points by ride-pooling vehicles. We investigate the real-time operational strategies of such a multimodal system taking into account the availability of micro-mobility vehicles at meeting points. In view that riders’ PUDO choices are interdependent in the integrated services, we devise a new dual-graph-based method that enables the decomposition of ride-pooling vehicle routing decisions and assignment decisions. We develop several interconnected subproblems that consider the mutual impacts between ride-pooling and micro-mobility services while we jointly determine rider-vehicles assignment, PUDO selection, ride-pooling vehicle routing decisions, micro-mobility vehicles repositioning decisions, and the transportation of micro-mobility vehicles on ride-pooling vehicles along with riders. We devise efficient algorithms for constructing graphs, identifying feasible trips, and making routing decisions for ride-pooling vehicles based on dynamic programming. The proposed models and algorithms are validated using real-world ride-hailing and bike-sharing data from Manhattan, New York City. The tests demonstrate that our algorithms can efficiently compute optimal matching. The results suggest that jointly operating ride-pooling and micro-mobility services can enhance rider shareability and reduce the repositioning costs of micro-mobility vehicles, thus benefiting both systems. Compared to door-to-door services, the integrated services can increase the number of served riders by more than 10 % while reducing repositioning costs by more than 50 % in the morning peak hour.

本文研究了多式联运网络中拼车和微出行服务的集成优化，以提高拼车的可能性，减少乘客的等待时间，提高平台的利润。我们考虑一个综合平台，同时提供拼车和微移动服务。与门到门的拼车服务不同，我们考虑在集合点进行多式联运，乘客可以使用微型移动车辆进入或离开附近的接送点，然后由拼车车辆在这些集合点上下车。我们研究了这样一个多模式系统的实时操作策略，考虑到在会议点的微移动车辆的可用性。考虑到综合服务中乘客的PUDO选择是相互依赖的，我们设计了一种新的基于双图的方法，实现了拼车车辆路线决策和分配决策的分解。我们开发了几个相互关联的子问题，这些子问题考虑了拼车和微出行服务之间的相互影响，同时我们共同确定了乘客-车辆分配、PUDO选择、拼车车辆路线决策、微出行车辆重新定位决策以及微出行车辆与乘客一起在拼车车辆上的运输。我们设计了高效的算法来构建图，识别可行的行程，并根据动态规划为拼车车辆制定路线决策。所提出的模型和算法使用来自纽约市曼哈顿的真实乘车和共享单车数据进行了验证。实验表明，我们的算法可以有效地计算最优匹配。结果表明，联合运营拼车和微出行服务可以增强乘客共享性，降低微出行车辆的重新定位成本，从而使两个系统都受益。与上门服务相比，综合服务可使服务乘客数量增加10%以上，同时在早高峰时段减少50%以上的重新定位成本。

{"title":"Integrated optimization of ride-pooling and shared micro-mobility services with meeting points","authors":"Tingting Dong , Yulong Hu , Zemin Wang , Sen Li","doi":"10.1016/j.trc.2025.105452","DOIUrl":"10.1016/j.trc.2025.105452","url":null,"abstract":"<div><div>This paper studies the integrated optimization of ride-pooling and micro-mobility services within a multimodal transportation network to enhance the likelihood of ride-pooling, reduce the waiting time for passengers, and improve the platform’s profit. We consider an integrated platform that simultaneously provides ride-pooling and micro-mobility services. Instead of door-to-door ride-pooling services, we consider inter-modal transfer at meeting points, where riders can access or egress nearby pickup or dropoff (PUDO) points using micro-mobility vehicles and then get picked up and dropped off at these meeting points by ride-pooling vehicles. We investigate the real-time operational strategies of such a multimodal system taking into account the availability of micro-mobility vehicles at meeting points. In view that riders’ PUDO choices are interdependent in the integrated services, we devise a new dual-graph-based method that enables the decomposition of ride-pooling vehicle routing decisions and assignment decisions. We develop several interconnected subproblems that consider the mutual impacts between ride-pooling and micro-mobility services while we jointly determine rider-vehicles assignment, PUDO selection, ride-pooling vehicle routing decisions, micro-mobility vehicles repositioning decisions, and the transportation of micro-mobility vehicles on ride-pooling vehicles along with riders. We devise efficient algorithms for constructing graphs, identifying feasible trips, and making routing decisions for ride-pooling vehicles based on dynamic programming. The proposed models and algorithms are validated using real-world ride-hailing and bike-sharing data from Manhattan, New York City. The tests demonstrate that our algorithms can efficiently compute optimal matching. The results suggest that jointly operating ride-pooling and micro-mobility services can enhance rider shareability and reduce the repositioning costs of micro-mobility vehicles, thus benefiting both systems. Compared to door-to-door services, the integrated services can increase the number of served riders by more than 10 % while reducing repositioning costs by more than 50 % in the morning peak hour.</div></div>","PeriodicalId":54417,"journal":{"name":"Transportation Research Part C-Emerging Technologies","volume":"183 ","pages":"Article 105452"},"PeriodicalIF":7.6,"publicationDate":"2025-11-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145613579","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Assessing the benefits of collaborative ridesharing across transportation network companies 评估跨交通网络公司合作拼车的好处

IF 7.6 1区工程技术 Q1 TRANSPORTATION SCIENCE & TECHNOLOGY

Transportation Research Part C-Emerging Technologies

Pub Date : 2025-11-29 DOI: 10.1016/j.trc.2025.105475

Xiaohan Wang , Zhan Zhao , Hongmou Zhang , Xiaotong Guo , Jinhua Zhao

By allowing two or more passengers to dynamically share parts of their trips in a single vehicle, ridesharing can lead to significant vehicle mileage traveled (VMT) savings, but its potential is often limited by market fragmentation, due to the co-existence of multiple Transportation Network Companies (TNCs). Although a collaborative ridesharing market that allows sharing across TNCs can produce additional VMT savings, to what extent such benefits vary based on market characteristics and behaviors of individual TNCs remains insufficiently understood. This study presents a framework to assess the maximum potential benefits of collaborative ridesharing and the contrasting equilibrium benefits resulting from varying strategic behaviors of TNCs. Specifically, we adopt a multi-TNC shareability network approach to estimate the maximum benefits under various market conditions, assuming that all TNCs are fully collaborative. In reality, TNCs’ willingness to collaborate is primarily motivated by their own profit gains, and thus we further propose a game-theoretic model to capture the collaboration dynamics among TNCs as a Nash game, allowing us to estimate the equilibrium benefits of collaborative ridesharing and evaluate the effectiveness of various profit-sharing schemes. Using the real-world TNC market in Manhattan, New York City as a case study, we find that a fully collaborative ridesharing market can generate additional VMT savings of up to 10.3 % over the existing fragmented market, and this upper bound is jointly determined by demand density, market division, competition intensity, and trip length. The Nash game results also reveal that the TNCs’ willingness to collaborate largely depends on the profit-sharing scheme; the Shapley value scheme tends to favor smaller, higher-priced TNCs, while the Equal Profit Method benefits dominant TNCs and more effectively facilitates collaboration among TNCs with greater pricing disparities. The proposed framework provides valuable insights for market regulators and business alliances, enabling them to evaluate collaboration outcomes and design appropriate profit-sharing schemes to promote and sustain collaborative ridesharing.

通过允许两名或多名乘客动态地共享一辆车的部分行程，拼车可以节省大量的车辆行驶里程（VMT），但由于多家交通网络公司（TNCs）并存，其潜力往往受到市场分散的限制。尽管允许跨国公司之间共享的协作式拼车市场可以节省额外的VMT，但这种效益在多大程度上取决于市场特征和各个跨国公司的行为，目前还没有得到充分的了解。本研究提出了一个框架来评估协同拼车的最大潜在效益，以及跨国公司不同战略行为所产生的对比均衡效益。具体而言，我们采用多跨国公司共享性网络方法来估算不同市场条件下的最大效益，假设所有跨国公司都是充分合作的。在现实中，跨国公司的合作意愿主要是由其自身的利润收益驱动的，因此我们进一步提出了一个博弈论模型，将跨国公司之间的合作动态作为纳什博弈来捕捉，使我们能够估计合作拼车的均衡效益，并评估各种利润分享方案的有效性。以纽约曼哈顿的跨国公司市场为例，我们发现，一个完全协作的拼车市场可以在现有的分散市场上节省高达10.3%的额外VMT，而这个上限是由需求密度、市场划分、竞争强度和行程长度共同决定的。纳什博弈结果还表明，跨国公司的合作意愿在很大程度上取决于利润分享方案；Shapley价值方案倾向于支持规模较小、价格较高的跨国公司，而等利润法则有利于占主导地位的跨国公司，并更有效地促进定价差异较大的跨国公司之间的合作。拟议的框架为市场监管机构和商业联盟提供了宝贵的见解，使他们能够评估合作成果并设计适当的利润分享计划，以促进和维持合作拼车。

{"title":"Assessing the benefits of collaborative ridesharing across transportation network companies","authors":"Xiaohan Wang , Zhan Zhao , Hongmou Zhang , Xiaotong Guo , Jinhua Zhao","doi":"10.1016/j.trc.2025.105475","DOIUrl":"10.1016/j.trc.2025.105475","url":null,"abstract":"<div><div>By allowing two or more passengers to dynamically share parts of their trips in a single vehicle, ridesharing can lead to significant vehicle mileage traveled (VMT) savings, but its potential is often limited by market fragmentation, due to the co-existence of multiple Transportation Network Companies (TNCs). Although a collaborative ridesharing market that allows sharing across TNCs can produce additional VMT savings, to what extent such benefits vary based on market characteristics and behaviors of individual TNCs remains insufficiently understood. This study presents a framework to assess the maximum potential benefits of collaborative ridesharing and the contrasting equilibrium benefits resulting from varying strategic behaviors of TNCs. Specifically, we adopt a multi-TNC shareability network approach to estimate the maximum benefits under various market conditions, assuming that all TNCs are fully collaborative. In reality, TNCs’ willingness to collaborate is primarily motivated by their own profit gains, and thus we further propose a game-theoretic model to capture the collaboration dynamics among TNCs as a Nash game, allowing us to estimate the equilibrium benefits of collaborative ridesharing and evaluate the effectiveness of various profit-sharing schemes. Using the real-world TNC market in Manhattan, New York City as a case study, we find that a fully collaborative ridesharing market can generate additional VMT savings of up to 10.3 % over the existing fragmented market, and this upper bound is jointly determined by demand density, market division, competition intensity, and trip length. The Nash game results also reveal that the TNCs’ willingness to collaborate largely depends on the profit-sharing scheme; the Shapley value scheme tends to favor smaller, higher-priced TNCs, while the Equal Profit Method benefits dominant TNCs and more effectively facilitates collaboration among TNCs with greater pricing disparities. The proposed framework provides valuable insights for market regulators and business alliances, enabling them to evaluate collaboration outcomes and design appropriate profit-sharing schemes to promote and sustain collaborative ridesharing.</div></div>","PeriodicalId":54417,"journal":{"name":"Transportation Research Part C-Emerging Technologies","volume":"183 ","pages":"Article 105475"},"PeriodicalIF":7.6,"publicationDate":"2025-11-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145619526","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

From prediction to explanation: A machine learning and causal mediation framework for roadway crash risk with connected vehicle data 从预测到解释：连接车辆数据的道路碰撞风险的机器学习和因果中介框架

IF 7.6 1区工程技术 Q1 TRANSPORTATION SCIENCE & TECHNOLOGY

Transportation Research Part C-Emerging Technologies

Pub Date : 2025-11-29 DOI: 10.1016/j.trc.2025.105479

Chenzhu Wang, Mohamed Abdel-Aty, Shaoyan Zhai, Abu Saif Md Nasim Uddin, Zubayer Islam

Traditional safety analysis often relies on static roadway attributes and crash frequencies, lacking insight into the behavioral mechanisms through which the roadway environment influence crash occurrence. To address this gap, this study proposes a novel machine learning–guided causal framework that integrates crash risk classification with causal mediation analysis to uncover the behavioral pathways linking road environment factors to crash risk. First, an ensemble learning framework is developed to classify high-risk road segments using connected vehicle (CV) data, where a double-layer meta-soft voting model achieves the best performance (AUC = 0.879; F1 = 0.655). Second, a Double Machine Learning (DML)–based mediation analysis is employed to quantify the indirect effects of operating speed and speed deviation in transmitting the impact of contextual features (e.g., speed limit, sidewalk coverage, building density) on crash risk. The findings demonstrate that operating speed and speed deviation significantly mediate the effects of posted speed limits, whereas abrupt driving maneuvers (e.g., hard braking) do not serve as structural mediators. Notably, the strength and structure of these mediation pathways vary across functional road classes, highlighting spatial heterogeneity in behavioral responses. The results support a shift from reactive, regulation-centric strategies to behavior-aware safety interventions informed by CV data. Practical recommendations include incorporating CV-derived metrics into real-time safety monitoring and prioritizing adaptive speed controls on collector and local roads. By bridging predictive analytics and causal inference, this study enhances the methodological toolkit for traffic safety research and contributes to a greater understanding of how roadway design, driver behavior, and crash risk interact through quantifiable mechanisms.

传统的安全分析往往依赖于静态道路属性和碰撞频率，缺乏对道路环境影响碰撞发生的行为机制的洞察。为了解决这一差距，本研究提出了一个新的机器学习引导的因果框架，该框架将碰撞风险分类与因果中介分析相结合，以揭示将道路环境因素与碰撞风险联系起来的行为途径。首先，利用车联网数据构建集成学习框架对高风险路段进行分类，其中双层元软投票模型的分类效果最佳（AUC = 0.879, F1 = 0.655）。其次，采用基于双机器学习（DML）的中介分析，量化运行速度和速度偏差在传递上下文特征（如限速、人行道覆盖、建筑密度）对碰撞风险的影响时的间接影响。研究结果表明，车速和车速偏差显著调节限速的影响，而突发性驾驶动作（如急刹车）不具有结构性调节作用。值得注意的是，这些中介路径的强度和结构在不同的功能道路类别中有所不同，突出了行为反应的空间异质性。研究结果支持从被动的、以监管为中心的策略转向由CV数据提供信息的行为意识安全干预措施。实际建议包括将cv衍生指标纳入实时安全监控，并优先考虑收集器和当地道路的自适应速度控制。通过连接预测分析和因果推理，本研究增强了交通安全研究的方法论工具包，并有助于更好地理解道路设计、驾驶员行为和碰撞风险如何通过可量化的机制相互作用。

{"title":"From prediction to explanation: A machine learning and causal mediation framework for roadway crash risk with connected vehicle data","authors":"Chenzhu Wang, Mohamed Abdel-Aty, Shaoyan Zhai, Abu Saif Md Nasim Uddin, Zubayer Islam","doi":"10.1016/j.trc.2025.105479","DOIUrl":"10.1016/j.trc.2025.105479","url":null,"abstract":"<div><div>Traditional safety analysis often relies on static roadway attributes and crash frequencies, lacking insight into the behavioral mechanisms through which the roadway environment influence crash occurrence. To address this gap, this study proposes a novel machine learning–guided causal framework that integrates crash risk classification with causal mediation analysis to uncover the behavioral pathways linking road environment factors to crash risk. First, an ensemble learning framework is developed to classify high-risk road segments using connected vehicle (CV) data, where a double-layer <em>meta</em>-soft voting model achieves the best performance (AUC = 0.879; F1 = 0.655). Second, a Double Machine Learning (DML)–based mediation analysis is employed to quantify the indirect effects of operating speed and speed deviation in transmitting the impact of contextual features (e.g., speed limit, sidewalk coverage, building density) on crash risk. The findings demonstrate that operating speed and speed deviation significantly mediate the effects of posted speed limits, whereas abrupt driving maneuvers (e.g., hard braking) do not serve as structural mediators. Notably, the strength and structure of these mediation pathways vary across functional road classes, highlighting spatial heterogeneity in behavioral responses. The results support a shift from reactive, regulation-centric strategies to behavior-aware safety interventions informed by CV data. Practical recommendations include incorporating CV-derived metrics into real-time safety monitoring and prioritizing adaptive speed controls on collector and local roads. By bridging predictive analytics and causal inference, this study enhances the methodological toolkit for traffic safety research and contributes to a greater understanding of how roadway design, driver behavior, and crash risk interact through quantifiable mechanisms.</div></div>","PeriodicalId":54417,"journal":{"name":"Transportation Research Part C-Emerging Technologies","volume":"183 ","pages":"Article 105479"},"PeriodicalIF":7.6,"publicationDate":"2025-11-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145613580","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Multiperiod alternative service optimization responding to joint disruptions in multimodal transit systems 响应联合中断的多式联运系统多时段替代服务优化

IF 7.6 1区工程技术 Q1 TRANSPORTATION SCIENCE & TECHNOLOGY

Transportation Research Part C-Emerging Technologies

Pub Date : 2025-11-29 DOI: 10.1016/j.trc.2025.105419

Hankun Zheng , Huijun Sun , Jianjun Wu , Dario Pacciarelli , Marcella Samà , Liujiang Kang

Besides disruptions within individual bus or metro transit systems, joint disruptions can occur in multimodal transit systems, which are typically more prolonged and extensive. This paper focuses on joint metro station closures and road disruptions over an entire day, divided into several time periods with various passenger demands and bus travel times. In response, alternative metro services are scheduled to skip the closed stations, while affected bus routes are adapted to provide alternative bus services. We simultaneously optimize adaptive bus routes, vehicle frequencies, and passenger demand assignment to develop multiperiod alternative metro and bus services. Notably, alternative service optimization within each time period is integrated to maintain bus link consistency, enhancing service reliability and passenger travel satisfaction. A novel discount-based strategy is introduced to balance bus link consistency with alternative service effectiveness. For the studied problem, we develop an integer non-linear programming model based on the set of candidate passenger paths, aiming to minimize total passenger and operation costs with discounts. Afterwards, to efficiently generate passenger paths and solve the model, we propose an innovative link-sequence-based column generation with station clustering. In our column generation, we iteratively solve an aggregated restricted master problem and simpler pricing subproblems to first generate passengers’ link sequences, which are subsequently expanded to passenger paths by solving a series of tailored expansion models. Additionally, a priority-based rule is incorporated into column generation to avoid generating duplicate link sequences. A station clustering procedure is developed to reduce the problem size of column generation, further improving computational efficiency. Finally, we validate our methodology using mid- and large-scale instances in Beijing, as well as performing comparative and sensitivity analyses.

除了个别公共汽车或地铁交通系统内的中断外，多式联运系统中也可能出现联合中断，这种中断通常时间更长，范围更广。本文关注的是地铁车站联合关闭和道路中断的整个一天，分为几个时间段，不同的乘客需求和公共汽车旅行时间。因应情况，我们安排替代地铁服务跳过关闭的车站，同时调整受影响的巴士路线，以提供替代巴士服务。我们同时优化自适应公交路线、车辆频率和乘客需求分配，以开发多时段的地铁和公交替代服务。值得注意的是，在每个时间段内整合了备选服务优化，以保持公交线路的一致性，提高服务可靠性和乘客出行满意度。提出了一种新的基于折扣的策略来平衡总线链路一致性和备选服务有效性。针对所研究的问题，我们建立了一个基于候选乘客路径集的整数非线性规划模型，以最小化总乘客和具有折扣的运营成本为目标。然后，为了有效地生成乘客路径并求解模型，我们提出了一种创新的基于链路序列的车站聚类列生成方法。在我们的列生成中，我们迭代地解决一个聚合的受限主问题和更简单的定价子问题，首先生成乘客的链路序列，然后通过求解一系列定制的扩展模型将其扩展到乘客路径。此外，在列生成中加入了基于优先级的规则，以避免生成重复的链接序列。为了减少列生成问题的规模，进一步提高了计算效率，提出了一种站聚类方法。最后，我们使用北京的大中型实例验证了我们的方法，并进行了比较和敏感性分析。

{"title":"Multiperiod alternative service optimization responding to joint disruptions in multimodal transit systems","authors":"Hankun Zheng , Huijun Sun , Jianjun Wu , Dario Pacciarelli , Marcella Samà , Liujiang Kang","doi":"10.1016/j.trc.2025.105419","DOIUrl":"10.1016/j.trc.2025.105419","url":null,"abstract":"<div><div>Besides disruptions within individual bus or metro transit systems, joint disruptions can occur in multimodal transit systems, which are typically more prolonged and extensive. This paper focuses on joint metro station closures and road disruptions over an entire day, divided into several time periods with various passenger demands and bus travel times. In response, alternative metro services are scheduled to skip the closed stations, while affected bus routes are adapted to provide alternative bus services. We simultaneously optimize adaptive bus routes, vehicle frequencies, and passenger demand assignment to develop multiperiod alternative metro and bus services. Notably, alternative service optimization within each time period is integrated to maintain bus link consistency, enhancing service reliability and passenger travel satisfaction. A novel discount-based strategy is introduced to balance bus link consistency with alternative service effectiveness. For the studied problem, we develop an integer non-linear programming model based on the set of candidate passenger paths, aiming to minimize total passenger and operation costs with discounts. Afterwards, to efficiently generate passenger paths and solve the model, we propose an innovative link-sequence-based column generation with station clustering. In our column generation, we iteratively solve an aggregated restricted master problem and simpler pricing subproblems to first generate passengers’ link sequences, which are subsequently expanded to passenger paths by solving a series of tailored expansion models. Additionally, a priority-based rule is incorporated into column generation to avoid generating duplicate link sequences. A station clustering procedure is developed to reduce the problem size of column generation, further improving computational efficiency. Finally, we validate our methodology using mid- and large-scale instances in Beijing, as well as performing comparative and sensitivity analyses.</div></div>","PeriodicalId":54417,"journal":{"name":"Transportation Research Part C-Emerging Technologies","volume":"183 ","pages":"Article 105419"},"PeriodicalIF":7.6,"publicationDate":"2025-11-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145614118","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Self-supervised graph learning for OD flow semantic awareness: harnessing high-order relationships via trajectory chains and POI contexts OD流语义感知的自监督图学习：通过轨迹链和POI上下文利用高阶关系

IF 7.6 1区工程技术 Q1 TRANSPORTATION SCIENCE & TECHNOLOGY

Transportation Research Part C-Emerging Technologies

Pub Date : 2025-11-28 DOI: 10.1016/j.trc.2025.105433

Tao Wu , Xinyi Lin , Jianxin Qin , Qianqian Deng

Origin-Destination (OD) flow semantic awareness enables the identification of trip patterns across diverse groups and spatiotemporal contexts, revealing spatial relationships and interactions. However, the lack of annotated semantic information in OD data poses a significant challenge, as it hinders the analysis of travel purposes. To address this challenge, we propose a novel self-supervised graph learning framework that leverages bicycle trajectory chains and Points of Interest (POIs) within a 15-minute walking radius. By integrating Graph Attention Networks (GAT) and Hypergraph Convolutional Networks (HGCN), our framework extracts spatial and high-order semantic features from mobility data without requiring labeled training data. A Transformer encoder further enriches node contextual features, enabling the inference of trip purposes and the identification of diverse spatiotemporal travel patterns. Empirical validation in Xiamen, China, demonstrates the framework’s effectiveness in uncovering meaningful OD flow semantics, providing new insights into urban mobility dynamics. The identified flow semantics successfully reveal the underlying mobility patterns and emphasize the synergistic potential of shared bikes within the public transportation network.

起点-目的地（OD）流语义感知能够识别不同群体和时空背景下的出行模式，揭示空间关系和相互作用。然而，OD数据中缺乏注释语义信息构成了一个重大挑战，因为它阻碍了对旅行目的的分析。为了应对这一挑战，我们提出了一种新的自监督图学习框架，该框架利用自行车轨迹链和15分钟步行半径内的兴趣点（poi）。通过集成图注意网络（GAT）和超图卷积网络（HGCN），我们的框架在不需要标记训练数据的情况下从移动数据中提取空间和高阶语义特征。Transformer编码器进一步丰富了节点上下文特征，从而能够推断旅行目的和识别不同的时空旅行模式。在中国厦门的实证验证表明，该框架在揭示有意义的OD流量语义方面是有效的，为城市交通动态提供了新的见解。确定的流量语义成功地揭示了潜在的移动模式，并强调了共享单车在公共交通网络中的协同潜力。

{"title":"Self-supervised graph learning for OD flow semantic awareness: harnessing high-order relationships via trajectory chains and POI contexts","authors":"Tao Wu , Xinyi Lin , Jianxin Qin , Qianqian Deng","doi":"10.1016/j.trc.2025.105433","DOIUrl":"10.1016/j.trc.2025.105433","url":null,"abstract":"<div><div>Origin-Destination (OD) flow semantic awareness enables the identification of trip patterns across diverse groups and spatiotemporal contexts, revealing spatial relationships and interactions. However, the lack of annotated semantic information in OD data poses a significant challenge, as it hinders the analysis of travel purposes. To address this challenge, we propose a novel self-supervised graph learning framework that leverages bicycle trajectory chains and Points of Interest (POIs) within a 15-minute walking radius. By integrating Graph Attention Networks (GAT) and Hypergraph Convolutional Networks (HGCN), our framework extracts spatial and high-order semantic features from mobility data without requiring labeled training data. A Transformer encoder further enriches node contextual features, enabling the inference of trip purposes and the identification of diverse spatiotemporal travel patterns. Empirical validation in Xiamen, China, demonstrates the framework’s effectiveness in uncovering meaningful OD flow semantics, providing new insights into urban mobility dynamics. The identified flow semantics successfully reveal the underlying mobility patterns and emphasize the synergistic potential of shared bikes within the public transportation network.</div></div>","PeriodicalId":54417,"journal":{"name":"Transportation Research Part C-Emerging Technologies","volume":"183 ","pages":"Article 105433"},"PeriodicalIF":7.6,"publicationDate":"2025-11-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145613581","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Chance-constrained eco-driving control of connected autonomous vehicles in mixed traffic environment at signalized intersections with uncertain signal timings 不确定信号配时交叉口混合交通环境下联网自动驾驶汽车的机会约束生态驾驶控制

IF 7.6 1区工程技术 Q1 TRANSPORTATION SCIENCE & TECHNOLOGY

Transportation Research Part C-Emerging Technologies

Pub Date : 2025-11-27 DOI: 10.1016/j.trc.2025.105460

Yongjie Xue , Dongxuan Bai , Yu Zhou , Chuan Ding , Hai L. Vu , Bin Yu

At signalized intersections, Connected Autonomous Vehicles (CAV) technology allows vehicle trajectory optimization using signal phasing and timing (SPaT) information, thereby reducing energy consumption caused by unnecessary deceleration and acceleration both upstream and downstream of signalized intersections. However, in widely deployed actuated traffic signal controllers, accurate and timely SPaT information is often unavailable for CAVs due to continuous adjustments in phase durations based on real-time traffic flow. Instead, only predicted SPaT information with inherent uncertainties can be provided within a specified confidence interval. Moreover, CAVs will coexist with human-driven vehicles (HVs) in mixed traffic environment, where the uncertainties of start-up lost time of HVs also present significant challenges to the trajectory optimization of CAVs. Hence, this paper proposes a chance-constrained eco-driving control for CAVs in mixed traffic environment with consideration of uncertainties in SPaT information and HVs. The method introduces a risk coefficient to represent the confidence level of CAVs regarding the probability distribution of uncertainties. The stochastic model predictive control (SMPC) with chance constraints is constructed to ensure that CAVs do not rear-end the preceding vehicle or exceed maximum acceleration/deceleration. The variable step size for SMPC is designed to enhance the performance of the proposed method while reducing the computation time and ensuring the real-time control. Safety analysis is conducted to demonstrate that the proposed method prevents CAVs from running red phases even under high risk coefficients. Numerical simulation under unsaturated, saturated and oversaturated traffic flow indicate that the proposed method effectively smooths the trajectories of CAVs and benefits the traffic efficiency. Comparing results at CAV penetration rates of 20 % and 80 % with 0 % (i.e., fully HVs), the proposed method reduces energy consumption by an average of 5.77 % and 10.24 % in the unsaturated traffic flow, 17.50 % and 31.32 % in the saturated traffic flow, and 22.99 % and 20.30 % in the oversaturated traffic flow, respectively.

在有信号的十字路口，联网自动驾驶汽车（CAV）技术可以利用信号相位和定时（SPaT）信息优化车辆轨迹，从而减少在有信号的十字路口上下行不必要的减速和加速造成的能源消耗。然而，在广泛部署的驱动式交通信号控制器中，由于基于实时交通流的相位持续时间不断调整，自动驾驶汽车往往无法获得准确及时的交通信号信息。相反，只能在指定的置信区间内提供具有固有不确定性的预测的痰信息。此外，在混合交通环境下，自动驾驶汽车将与人类驾驶汽车共存，人类驾驶汽车启动损失时间的不确定性也对自动驾驶汽车的轨迹优化提出了重大挑战。因此，本文针对混合交通环境下的自动驾驶汽车，提出了一种考虑车辆信息和HVs不确定性的机会约束生态驾驶控制方法。该方法引入风险系数来表示cav对不确定性概率分布的置信度。为了保证自动驾驶汽车不会追尾或超过最大加减速速度，构造了带有机会约束的随机模型预测控制（SMPC）。设计了SMPC的变步长以提高该方法的性能，同时减少了计算时间，保证了控制的实时性。安全性分析表明，即使在高风险系数较高的情况下，该方法也能防止自动驾驶汽车出现红相。非饱和、饱和和过饱和交通流下的数值模拟结果表明，该方法能有效地平滑自动驾驶汽车的行驶轨迹，提高交通效率。对比CAV渗透率为20%和80%与0%（即全HVs）时的结果，该方法在不饱和交通流中平均降低能耗5.77%和10.24%，在饱和交通流中平均降低能耗17.50%和31.32%，在过饱和交通流中平均降低能耗22.99%和20.30%。

{"title":"Chance-constrained eco-driving control of connected autonomous vehicles in mixed traffic environment at signalized intersections with uncertain signal timings","authors":"Yongjie Xue , Dongxuan Bai , Yu Zhou , Chuan Ding , Hai L. Vu , Bin Yu","doi":"10.1016/j.trc.2025.105460","DOIUrl":"10.1016/j.trc.2025.105460","url":null,"abstract":"<div><div>At signalized intersections, Connected Autonomous Vehicles (CAV) technology allows vehicle trajectory optimization using signal phasing and timing (SPaT) information, thereby reducing energy consumption caused by unnecessary deceleration and acceleration both upstream and downstream of signalized intersections. However, in widely deployed actuated traffic signal controllers, accurate and timely SPaT information is often unavailable for CAVs due to continuous adjustments in phase durations based on real-time traffic flow. Instead, only predicted SPaT information with inherent uncertainties can be provided within a specified confidence interval. Moreover, CAVs will coexist with human-driven vehicles (HVs) in mixed traffic environment, where the uncertainties of start-up lost time of HVs also present significant challenges to the trajectory optimization of CAVs. Hence, this paper proposes a chance-constrained eco-driving control for CAVs in mixed traffic environment with consideration of uncertainties in SPaT information and HVs. The method introduces a risk coefficient to represent the confidence level of CAVs regarding the probability distribution of uncertainties. The stochastic model predictive control (SMPC) with chance constraints is constructed to ensure that CAVs do not rear-end the preceding vehicle or exceed maximum acceleration/deceleration. The variable step size for SMPC is designed to enhance the performance of the proposed method while reducing the computation time and ensuring the real-time control. Safety analysis is conducted to demonstrate that the proposed method prevents CAVs from running red phases even under high risk coefficients. Numerical simulation under unsaturated, saturated and oversaturated traffic flow indicate that the proposed method effectively smooths the trajectories of CAVs and benefits the traffic efficiency. Comparing results at CAV penetration rates of 20 % and 80 % with 0 % (i.e., fully HVs), the proposed method reduces energy consumption by an average of 5.77 % and 10.24 % in the unsaturated traffic flow, 17.50 % and 31.32 % in the saturated traffic flow, and 22.99 % and 20.30 % in the oversaturated traffic flow, respectively.</div></div>","PeriodicalId":54417,"journal":{"name":"Transportation Research Part C-Emerging Technologies","volume":"183 ","pages":"Article 105460"},"PeriodicalIF":7.6,"publicationDate":"2025-11-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145611672","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Addressing corner cases in autonomous driving: A world model-based approach with mixture of experts and LLMs 解决自动驾驶中的边缘案例：专家和法学硕士混合的基于世界模型的方法

IF 7.6 1区工程技术 Q1 TRANSPORTATION SCIENCE & TECHNOLOGY

Transportation Research Part C-Emerging Technologies

Pub Date : 2025-11-26 DOI: 10.1016/j.trc.2025.105456

Haicheng Liao , Bonan Wang , Junxian Yang , Chengyue Wang , Zhengbing He , Guohui Zhang , Chengzhong Xu , Zhenning Li

Accurate and reliable motion forecasting is essential for the safe deployment of autonomous vehicles (AVs), particularly in rare but safety-critical scenarios known as corner cases. Existing models often underperform in these situations due to an over-representation of common scenes in training data and limited generalization capabilities. To address this limitation, we present WM-MoE, the first world model-based motion forecasting framework that unifies perception, temporal memory, and decision making to address the challenges of high-risk corner-case scenarios. The model constructs a compact scene representation that explains current observations, anticipates future dynamics, and evaluates the outcomes of potential actions. To enhance long-horizon reasoning, we leverage large language models (LLMs) and introduce a lightweight temporal tokenizer that maps agent trajectories and contextual cues into the LLM’s feature space without additional training, enriching temporal context and commonsense priors. Furthermore, a mixture-of-experts (MoE) is introduced to decompose complex corner cases into subproblems and allocate capacity across scenario types, and a router assigns scenes to specialized experts that infer agent intent and perform counterfactual rollouts. In addition, we introduce nuScenes-corner, a new benchmark that comprises four real-world corner-case scenarios for rigorous evaluation. Extensive experiments on four benchmark datasets (nuScenes, NGSIM, HighD, and MoCAD) showcase that WM-MoE consistently outperforms state-of-the-art (SOTA) baselines and remains robust under corner-case and data-missing conditions, indicating the promise of world model-based architectures for robust and generalizable motion forecasting in fully AVs.

准确可靠的运动预测对于自动驾驶汽车的安全部署至关重要，尤其是在罕见但对安全至关重要的情况下。由于训练数据中常见场景的过度表示和有限的泛化能力，现有模型在这些情况下往往表现不佳。为了解决这一限制，我们提出了WM-MoE，这是世界上第一个基于模型的运动预测框架，它统一了感知、时间记忆和决策制定，以解决高风险边缘情况的挑战。该模型构建了一个紧凑的场景表示，解释了当前的观察结果，预测了未来的动态，并评估了潜在行动的结果。为了增强长视界推理，我们利用大型语言模型（LLM）并引入轻量级的时间标记器，该标记器将代理轨迹和上下文线索映射到LLM的特征空间中，而无需额外的训练，从而丰富了时间上下文和常识先验。此外，引入混合专家（MoE）将复杂的角落案例分解为子问题并跨场景类型分配容量，路由器将场景分配给推断代理意图并执行反事实部署的专业专家。此外，我们还介绍了nuScenes-corner，这是一个新的基准测试，它包含四个真实世界的角落案例场景，用于严格的评估。在四个基准数据集（nuScenes， NGSIM， HighD和MoCAD）上进行的大量实验表明，WM-MoE始终优于最先进的（SOTA）基线，并且在角落情况和数据缺失条件下保持稳健，这表明基于世界模型的架构有望在全自动驾驶汽车中实现稳健和通用的运动预测。

{"title":"Addressing corner cases in autonomous driving: A world model-based approach with mixture of experts and LLMs","authors":"Haicheng Liao , Bonan Wang , Junxian Yang , Chengyue Wang , Zhengbing He , Guohui Zhang , Chengzhong Xu , Zhenning Li","doi":"10.1016/j.trc.2025.105456","DOIUrl":"10.1016/j.trc.2025.105456","url":null,"abstract":"<div><div>Accurate and reliable motion forecasting is essential for the safe deployment of autonomous vehicles (AVs), particularly in rare but safety-critical scenarios known as corner cases. Existing models often underperform in these situations due to an over-representation of common scenes in training data and limited generalization capabilities. To address this limitation, we present WM-MoE, the first world model-based motion forecasting framework that unifies perception, temporal memory, and decision making to address the challenges of high-risk corner-case scenarios. The model constructs a compact scene representation that explains current observations, anticipates future dynamics, and evaluates the outcomes of potential actions. To enhance long-horizon reasoning, we leverage large language models (LLMs) and introduce a lightweight temporal tokenizer that maps agent trajectories and contextual cues into the LLM’s feature space without additional training, enriching temporal context and commonsense priors. Furthermore, a mixture-of-experts (MoE) is introduced to decompose complex corner cases into subproblems and allocate capacity across scenario types, and a router assigns scenes to specialized experts that infer agent intent and perform counterfactual rollouts. In addition, we introduce nuScenes-corner, a new benchmark that comprises four real-world corner-case scenarios for rigorous evaluation. Extensive experiments on four benchmark datasets (nuScenes, NGSIM, HighD, and MoCAD) showcase that WM-MoE consistently outperforms state-of-the-art (SOTA) baselines and remains robust under corner-case and data-missing conditions, indicating the promise of world model-based architectures for robust and generalizable motion forecasting in fully AVs.</div></div>","PeriodicalId":54417,"journal":{"name":"Transportation Research Part C-Emerging Technologies","volume":"183 ","pages":"Article 105456"},"PeriodicalIF":7.6,"publicationDate":"2025-11-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145598640","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0