As commuting rail networks expand and passenger demand grows, service delays have become a growing challenge, propagating through the network and undermining system reliability. Prevailing research, often reliant on statistical correlations or 'black-box' predictive models, fails to reveal the causal mechanisms of delay propagation. To address this gap, this study proposes a network-centric approach grounded in causal inference to explicitly map the directional pathways of delay.Focusing on the Stockholm commuter rail system, we employ the Peter and Clark Momentary Conditional Independence (PCMCI) algorithm on real-world time-series data to construct a Delay Propagation Causal Network (DPCN). Our multi-stage analysis of the DPCN reveals a highly structured network where delays propagate along stable, predictable pathways. A novel classification identifies four distinct station roles with a clear core-periphery spatial logic. To identify the most critical nodes, we introduce a composite causal delay impact index, which integrates causal strength with real-world delay probabilities and successfully identifies high-impact station clusters that align with peak-hour commuter traffic. A final comparison illustrates the advantages of a causality-based approach over correlation-based methods in distinguishing causal propagation links from spurious associations. This study presents a generalizable, causality-based framework and practical tools for transit authorities, offering a data-driven foundation for proactive network management. It enables operators to identify and mitigate systemic vulnerabilities, thereby enhancing the efficiency, reliability, and resilience of commuter rail systems.

Street-level greenery assessment through the Green View Index (GVI) is essential for transportation planning and sustainable urban development. However, current GVI estimation methods rely heavily on street-view imagery, facing significant scalability challenges including high data acquisition costs, inconsistent temporal coverage, and limited cross-regional comparability. This study proposes an explainable machine learning framework for GVI estimation using ubiquitous satellite remote sensing data from Sentinel-2A imagery. Eight complementary spectral indices were systematically evaluated: enhanced vegetation indices, spectral variation indices, and urban context indices. Multiple spatial buffer configurations (200m to 1000m) and input strategies were tested, including mean-value aggregation and raster input for CNN models. The methodology was validated using over 6,300 samples from Helsinki (Finland), with comprehensive comparisons across traditional machine learning models and deep learning architectures.The optimized XGBoost model with 1000m buffer achieved R² = 0.67 and MSE = 0.022, improved approximately 20% over baseline approaches using only NDVI and RGB bands. SHAP (SHapley Additive exPlanations) analysis revealed spatial scale-dependent feature importance patterns, with Urban Index (UI) consistently ranking highest across all scales, while NDVI showed lower importance at larger buffer sizes. Alternative spatial modeling approaches (CNN) underperformed compared to mean-value aggregation methods, suggesting that statistical aggregation may be more robust for GVI predictions. The framework eliminates dependency on street-view imagery while maintaining competitive accuracy, offering a scalable solution for large-scale green infrastructure assessment. The explainable AI approach provides interpretable insights into environmental factors influencing street-level green perception, supporting evidence-based decision-making in transportation planning and active transportation network development.

Public transportation networks are highly interconnected, where disruptions like traffic congestion propagate bus delays and impact performance. Identifying delay causes is crucial, yet most studies rely on correlation-based methods rather than causal analysis. Attribution methods like the Shapley value quantify factor contributions but often overlook causal dependencies, leading to potential bias. This study uses a causal discovery model to uncover causal relationships between bus delays and various factors (e.g., operational factors, calendar, and weather). Based on this causal graph, an explainable Artificial Intelligence (AI) method quantifies each factor's contribution to delays, focusing on how these contributions vary at different stops along a route. By integrating scheduled route data and real-time vehicle locations, we analyze factor contributions over time and space, exploring various scenarios along the route. Cross-validation is conducted by comparing the importance ranking of factors with the Seemingly Unrelated Regression Equations (SURE). Results show significant variations in factors contributing to delays along the route. Delays at upstream stops propagate downstream, indicating a cascading effect. Operational factors dominate, accounting for 50%–83% of delays. Notably, delays from the preceding two to three stops have a larger impact than just the immediately preceding one stop, and origin delays strongly affect the first half of the route.

Multimodal mobility systems provide seamless travel by integrating different types of transportation modes. Most existing studies model service operations and users’ travel choices independently or iteratively and constrained with pre-defined multimodal travel options. The paper proposes a choice-based optimization approach that optimizes service operations with explicitly embedded travelers’ choices described by the multinomial logit (MNL) model. It allows the flexible combination of travel modes and routes in multimodal mobility systems. We propose a computationally efficient linearization method for transformed MNL constraints with bounded errors to solve the choice-based optimization model. The model is validated using a mobility on demand and public transport network by comparing it with a simulation sampling-based MNL linearization method. The results show that the mixed-integer formulation provides a high-quality solution in terms of both the estimated choice probability errors and computational speed. We also conduct an error analysis and a sensitivity analysis to explore the behavior of the proposed approach. The real-world case study in Stockholm further illustrates that the analytical formulation achieves a better system operation performance than the traditional iterative supply–demand updating optimization method. The choice-based optimization model and solution formulation are highly adaptable for operations decision support integrating stochastic travel choices in multimodal mobility systems.

Understanding travelers’ route choices can help policymakers devise optimal operational and planning strategies for both normal and abnormal circumstances. However, existing choice modeling methods often rely on predefined assumptions and struggle to capture the dynamic and adaptive nature of travel behavior. Recently, Large Language Models (LLMs) have emerged as a promising alternative, demonstrating remarkable ability to replicate human-like behaviors across various fields. Despite this potential, their capacity to accurately simulate human route choice behavior in transportation contexts remains doubtful. To satisfy this curiosity, this paper investigates the potential of LLMs for route choice modeling by introducing an LLM-empowered agent, “LLMTraveler.” This agent integrates an LLM as its core, equipped with a memory system that learns from past experiences and makes decisions by balancing retrieved data and personality traits. The study systematically evaluates the LLMTraveler's ability to replicate human-like decision-making through two stages of day-to-day (DTD) congestion games: (1) analyzing its route-switching behavior in single origin–destination (OD) pair scenarios, where it demonstrates patterns that align with laboratory data but cannot be fully captured by traditional models, and (2) testing its capacity to model adaptive learning behaviors in multi-OD scenarios on the Ortuzar and Willumsen (OW) network, producing results comparable to Multinomial Logit (MNL) and Reinforcement Learning (RL) models. Additionally, the study assesses lightweight, open-source LLMs, highlighting their effectiveness in route choice simulation and their potential as cost-effective alternatives to more advanced closed-source models. These experiments demonstrate that the framework can partially replicate human-like decision-making in route choice while providing natural language explanations for its decisions. This capability offers valuable insights for transportation policymaking, such as simulating traveler responses to new policies or changes in the network. The code for this paper is open-source and available at: https://github.com/georgewanglz2019/LLMTraveler.

The surging demand for Intelligent Transportation Systems (ITS) to deliver advanced train-related Information for dispatchers and passengers has spurred the development of advanced train delay prediction models. Despite considerable efforts devoted to developing methodologies that can be used to model train operation conditions and produce anticipated train delays, the evaluation strategies for train delay prediction models remain under-researched, particularly evident when accuracy is always found to be the only determinant in model selection. The absence of a standardised evaluation procedure for assessing the effectiveness of these prediction models has hindered the practical implementation of these models. To bridge this gap, the study conducted a systematic literature review on data-driven train delay prediction models and introduced the novel AP-GRIP (Accuracy, Precision, Generalisability, Robustness, Interpretability, Practicality) evaluation framework. The framework covers six key aspects across overall, spatial, temporal, and train-specific dimensions, providing a systematic approach for the comprehensive assessment of train delay prediction models. Each aspect and dimension is thoroughly discussed and synthesised with its definitions, measuring metrics, and important considerations. A critical discussion clarifies several interactions, such as predetermined objectives, desired outputs, model type, benchmark models, and data availability, resulting in a logical framework for assessing train delay prediction models. The proposed framework uncovers inadequate prediction patterns, offering insights on when, where, and why the prediction models excel and fall short, assisting end-users in determining model suitability for specific prediction tasks.

Understanding the impact of various factors on bus deceleration behavior at intersections has important implications for bus operations control, management, and safety. This paper develops a multiple linear regression model to analyze the factors influencing bus driver deceleration (a proxy of safe driving state) at intersections using data from multiple sources, including the on-board closed-circuit television (CCTV), the advanced driver assistance system (ADAS), the bus controller area network (CAN), the bus operation, and the driver profile data. We develop a comprehensive model data extraction framework and corresponding methods to effectively estimate/calculate the bus deceleration rate (dependent variable) and its influencing factors (independent variables). We explored the factors impact on bus deceleration behavior at intersections using data from a typical bus route in China. The results highlight significant factors, including driver characteristics (age), en-route and intersection approaching driving states (trip delay, turnaround time, driving direction, and approaching speed), intersection characteristics (types, the number of lanes, zebra crossing, divider, bus lanes, right turn lanes, the stop location) and traffic conditions (surrounding vehicles). Generally, drivers with younger ages (having short reaction times) and driving with psychological anticipation of complex situations (from surrounding vehicles and pedestrians or unsignalized intersections) tend to decelerate more smoothly. The agencies may enhance safe bus driving behavior by allowing enough turnaround time in timetabling, recommending intersection approaching speed, and providing tailored ADAS system alarms (rather than flooding all alerts). Also, the planning of bus infrastructures (e.g., dedicated lanes and stop locations) should be properly evaluated considering their soft contribution to safe driving behaviors at intersections.