ABSTRACT. This study explores fare evasion in public transport by analysing the mode and route choices of users in Santiago, Chile. Using survey and smartcard data, latent class choice models were developed to distinguish fare payers and evaders, uncovering heterogeneity in their preferences. Results indicate that fare evaders prioritize fewer transfers, shorter wait times, and sometimes longer travel times, reflecting distinct behavioural patterns. These findings enhance understanding of travel behaviour and can inform targeted policies to reduce evasion, optimize public transport operations, and improve the overall efficiency of the transit system.

ABSTRACT. Rail and long-distance bus integration has been widely advocated to enhance connectivity. This study develops a rail-bus network to quantify integration through indicators of travel time-weighted population, network centrality, and service availability. These indicators are further incorporated into an expanded node-place model to evaluate the balance between transport provision and potential demand. Based on the proposed model, the study examines 102 rail stations across Scotland’s rural areas, identifying potential improvements, including relocating bus terminus closer to rail stations and addressing imbalances in network accessibility. The findings offer data-driven insights to support transport planners in optimising rail and bus network integration.

ABSTRACT. Decarbonizing the urban passenger transport sector faces significant barriers, highlighting the potential of innovative strategies like incentive schemes. Such a scheme could play a pivotal role in incentivizing emission reductions and fostering a shift toward sustainable mobility options. This study introduces a dual incentive scheme designed to promote sustainable shifts in travel behaviours. The framework integrates a departure time incentive, encouraging shifts from peak-hour road traffic to off-peak road traffic, and a travel mode incentive, promoting transitions from off-peak road travel to off-peak metro usage. The models capture the interactions between five distinct passenger flow types and account for incentive allocation and equilibrium conditions. Numerical simulations using standard networks show that by providing a 10 CNY incentive to each traveller, the proposed scheme reduces carbon emissions by 15.96%, lowers travel costs by 7.09%, and increases metro usage by 14.38%. This study highlights the effectiveness of incentive schemes in optimising multimodal traffic systems and offers insights into designing sustainable transport policies.

ABSTRACT. Redesigning multimodal transit network can utilize shared autonomous vehicles (SAVs) as feeders to enhance service efficiency and coverage. This paper presents an optimization framework for the joint multimodal transit frequency and SAV fleet size problem, a variant of the transit network frequency setting problem. The objective is to maximize total transit ridership (including SAV-fed trips and subtracting boarding rejections) across multiple time periods under budget constraints, considering endogenous mode choice (transit, point-to-point SAVs, driving) and route selection, while allowing for strategic route removal by setting frequencies to zero. Due to the problem's non-linear, non-convex nature and the computational challenges of large-scale networks, we develop a hybrid solution approach that combines a metaheuristic approach (particle swarm optimization) with nonlinear programming for local solution refinement. To ensure computational tractability, the framework integrates analytical approximation models for SAV waiting times based on fleet utilization, multimodal network assignment for route choice, and multinomial logit mode choice behavior, bypassing the need for computationally intensive simulations within the main optimization loop. Applied to the Chicago metropolitan area’s multimodal network, our method illustrates a 33.3% increase in transit ridership through optimized transit route frequencies and SAV integration, particularly enhancing off-peak service accessibility and strategically reallocating resources.

ABSTRACT. There is a growing interest in understanding the capacity of cities to reduce global energy use and thus mitigate climate change. The energy requirement of the transport sector remains central to this theme as it accounts for roughly one-third of global energy demand. Using unique multi-city datasets, this study provides new insights into the causal impact of density on the energy use of major urban travel modes. Our analysis shows that the densification of rail transit operations substantially reduces its energy requirement per unit of passenger kilometres. Increasing private vehicular operations on road networks, on the other hand, offers no energy benefits. These results underscore that a compact urban form, primarily driven by rail-based travel, can lead to a more sustainable future.

ABSTRACT. This study employs multiscale geographically weighted regression (MGWR) to examine how various urban forms and socio-demographic variables influence passenger boarding at transit stops in the city of Indore, India. The research analyzes the public transport network operated by Atal Indore City Transport Service Limited (AICTSL) across 39 routes. Using diverse data, including the number of transit lines passing a stop, road length, socio-demographic factors, and land use characteristics, the study reveals key patterns. The findings indicate that the number of transit lines at stops, household industries population density and transport infrastructure consistently show positive relationships with passenger boarding, while other factors such as the ratio of literate by ill-literate population, road length in stop catchment areas, and public management and service land use show mixed or negative effects on passenger boarding at a transit stop.

ABSTRACT. Cross-traffic turns (left turns in right-hand traffic) significantly delay buses at intersections, reducing transit reliability. This study introduces the Bus Cross-Traffic Turn Priority Box, a design that allows buses to bypass turn queues using through lanes and enables the pre-acceleration of turning traffic when no bus is present. Microscopic simulations conducted at two intersections demonstrate substantial delay reductions for buses and minor improvements for general traffic. Unlike traditional bus priority measures, this approach requires minimal infrastructure changes and enhances overall intersection efficiency, supporting its adoption in space-constrained urban settings to improve public transport performance without adversely affecting other road users.

ABSTRACT. We use bus GPS data across 500 routes to estimate the impact of priority infrastructure on buses' speed and ridership in Chile. Almost 100 million bus trips allow us to leverage within-route variation in the proportion of the route in which buses travel along bus lanes or Bus Rapid Transit (BRT) corridors. Corridors increase bus speeds by 20% at peak hours. Bus lanes, often seen as an equally effective but cheaper alternative to a BRT corridor, are, on average, ineffective. However, bus lanes achieve the same travel time savings as BRT corridors only when fully isolated from private vehicles, coupled with monitoring cameras and enforcement.

ABSTRACT. This study investigates the optimal deployment planning of a modular autonomous vehicle (MAV)-based shuttle transit line that connects an autonomous rail rapid transit (ART) station and a high-speed railway station. The proposed methodology establishes MAV dispatching plans and platoon formations while ensuring seamless transfer coordination with ART arrival times. A mathematical programming model and a graphical scheduling model are employed to determine the required MAV fleet size. A case study conducted in Yibin, China, validates the feasibility and practical applicability of the methodology, showcasing its potential to enhance seamless integrated mobility under the concept of Mobility-as-a-Service.

ABSTRACT. Unplanned metro disruptions can severely reduce transit capacity, triggering passenger spillovers that require rapid and adaptive shuttle bus deployment. While existing strategies often rely on static planning or scenario-based optimization, they typically neglect the need for real-time adaptation to evolving system conditions. This study develops a rolling-horizon model predictive control (MPC) framework for dynamically adjusting shuttle bus frequencies in response to uncertain metro capacity recovery. Metro capacity is modelled as a stochastic process with unknown recovery dynamics, which are estimated online through sequential data assimilation. The proposed framework generates forward-looking frequency decisions that adapt to newly observed capacity realizations over time. Numerical experiments on a synthetic transit network demonstrate that the MPC policy consistently reduces total system cost compared to benchmark strategies based on static capacity assumptions. Moreover, the adaptive policy achieves performance that does not deviate substantially from a theoretical lower bound defined by an oracle with perfect foresight. These results highlight the potential of rolling-horizon, learning-based control for improving transit service resilience under disruption uncertainty.

ABSTRACT. This study quantitatively investigates the impact of adverse weather on metro ridership dynamics during weekday peak hours, focusing on station-level variations and associated safety risks. Quantitative indicators, including peak values, directional imbalance, and flow concentration and deviation, are developed and analysed using a quasi-experimental approach. The analytics and impact evaluations are based on Automatic Fare Collection (AFC) data and weather records from Shanghai’s flood season. Results reveal that extreme weather increases crowding risks with elevating peak ridership and directional imbalance at specific stations, while time-indexed characteristics remain largely unaffected. The findings offer implications for crowd safety enhancement under extreme weather.

ABSTRACT. This study aims to solve the bus network design problem with the integration of limited stop service and consideration of autonomous bus operation to optimize both the total passenger travel time, the number of transfers, and the operator cost. This problem is modelled as a mixed integer nonlinear program and is NP-hard. A hybrid genetic algorithm is proposed in the solution methodology which can tackle the normal bus line design, limited stop service pattern design, autonomous bus line selection, and frequency setting problems simultaneously. The hybrid algorithm encapsulates a line search heuristic, a pattern design heuristic and an autonomous line selection heuristic to tackle the frequency setting problem, the limited stop service design problem and the autonomous bus lines determination problem respectively. Specific genetic mutation operators are developed to explore lines operated by autonomous buses. Compared with the base case of bus network design, the developed algorithm can generate a solution that can reduce the total passenger travel time and operator cost simultaneously. Numerical experiments are also implemented to demonstrate the effects of the value of travel time.

ABSTRACT. In this study, we focus on practical and quantitative methods for designing rail transit networks. We generalize the corridor characteristics of rail transit networks and develop a potential rail transit corridor search algorithm. Potential rail transit network schemes are generated on the basis of the potential rules and algorithms of rail transit corridor combinations. These network schemes are evaluated by splitting the model and evaluating the total social costs, facilitating the identification of efficient alternatives. The method is implemented and validated in a case study, demonstrating its feasibility, efficacy, and suitability for practical planning applications.

ABSTRACT. With the extension of urbanization, enhancing efficient trunk-feeder transit systems becomes crucial to accommodate the escalating travel demand. A well-designed trunk-feeder network not only improves travel efficiency on trunk lines but also bolsters connectivity and flexibility through feeder services. In this paper, based on the heterogeneity of travel demand, we optimize a heterogeneous trunk-feeder transit network design using continuous models. We consider both the heterogeneous demand density and heterogeneous network design. This study pioneers a model that integrates heterogeneous trunk-line designs and multiple service modes. This model advances beyond previous studies that predominantly focused on homogeneous transit network designs or single service modalities. Our findings provide valuable insights for urban planners and policymakers in developing more adaptive and efficient transit solutions.

ABSTRACT. This study addresses tidal tourist traffic in scenic areas by proposing a bi-level optimization model for customized tourist bus routes and ticket pricing. The upper-level model optimizes routes, frequencies, and ticket prices to maximize operator revenue, while the lower-level model minimizes tourists' travel costs. A case study on Chongming Island demonstrates significant improvements in operational efficiency, vehicle utilization, and passenger waiting time. A sensitivity analysis was also conducted to assess the impact of varying pricing and scheduling strategies on both tourist behavior and operational outcomes, offering practical insights for sustainable transport in tidal tourist areas.

ABSTRACT. On October 3rd, 2024, three stations along the east end of Montreal’s blue metro line were closed, resulting in a seven-day service disruption. While previous studies have examined the operational impacts of such disruptions, their effects on user experiences remain underexplored. To address this gap, we measure the impacts of the closure on user satisfaction and their willingness to recommend transit services. Using data from a bilingual online survey launched the day after the disruption began, we analyzed responses from blue line users (N= 655) by employing ordered probit models. The survey included a treatment group of riders directly impacted by the closure (N = 361) and a control group of those unaffected (N = 294). Additionally, we incorporate data from a secondary survey conducted one prior to the closure, which included riders living close to blue line stations (N = 161), as a secondary control. Our findings reveal a significant decrease in both user satisfaction and willingness to recommend transit services among those impacted by the metro closure. However, these negative impacts can be mitigated when users perceive the availability of reliable and suitable transit alternatives. The findings from this research can be of interest to practitioners and policymakers as they highlight the broader implications of metro disruptions.

ABSTRACT. Passenger service depends on line frequencies and the resulting timetable. Operating costs are largely defined by the vehicle schedule. Traditionally, the frequencies, timetable and vehicle schedule are determined sequentially, which may lead to suboptimal solutions. Moreover, pre-assignment of passengers to routes could lead to further suboptimality. To avoid this, we integrate frequency setting, timetabling, and vehicle scheduling to create a-periodic full-day timetables for medium-to-large urban bus networks, where passenger route choice depends on the final timetable.

Another specificity of our model is the division of the daily timetable into a few periods with different frequencies. This allows for the adaptation of the number of bus services to the time dependency of the demand. Headways are regular within each period to make the timetable easier to memorise for passengers.

To solve this problem, we develop problem-specific low-level heuristics and a hyper-heuristic algorithm based on sequences of low-level heuristics. We present computational results derived from real-life data on the Odense bus network in Denmark. We focus on demonstrating the value of integrating frequency setting, timetabling, vehicle scheduling, and passenger route choice. We also measure the impact of the set of allowed frequencies on passenger and operation costs.

ABSTRACT. The line planning problem has been well studied in the literature. Most models aim to minimize the costs or to maximize the number of direct travellers. But the new bus system Nova Xarxa in Barcelona proves that properly designed transfer-based networks can be very appealing and even attract more demand than their conventional counterparts. This design is based on the analytical model proposed by Daganzo (2010), which has two main limitations. First, the model assumes that origins and destinations are uniformly and independently distributed, and second it produces conceptual plans for particular road network topology in the city such as a grid, hub-and-spoke systems, and a hybrid of both. This paper presents a two-stage discrete model to construct transfer-based bus networks for any demand distribution and for any road network topology. We take a new approach maximizing the passenger flow intensity (PFI) to keep the costs low for the public transportation company. Additionally, this approach considers transfer from one line to another at the same stop to reduce the walk distance. To tackle our problem, we present a reinforcement learning method and suggest a solution approach using MCTS for finding the optimal bus network. Numerical results of real-world instances are presented.

ABSTRACT. Uncertain travel time may break the linkage between interlining bus shifts and lead to low bus service punctuality and unreliable bus utilization. To address this issue, a two-stage robust bus timetabling and scheduling optimization model is proposed. The effectiveness of the model and algorithm is verified through a case study of five bus lines. Results show that compared with the traditional fixed schedule-based interlining scheduling plan, the proposed method shorten the stand-by time by 29.55%. Analysis finds that TCPs reduce the arrival time deviation of buses and the total cost of vehicle use, and improve vehicle work efficiency and punctuality.

ABSTRACT. Most transit network design studies overlook demand variations. We propose a two-stage stochastic model that accounts for stochastic demand, minimizing passenger travel time---including access, egress, in-vehicle, and transfer times---and operating costs. The model designs distinct networks for varying weather and peak/off-peak conditions, maintaining consistent line configurations while adapting frequencies. Using MATSim data for a Zurich subnetwork with 28,000 transit passengers, our results show the stochastic model outperforms the deterministic model in reliability and serving higher demand. Compared to the robust model, the stochastic model achieves similar performance in serving high demand but with lower passenger and operating costs.

ABSTRACT. As a primary mode within urban rail transit systems, metro plays a critical role in supporting urban development. However, operational disruptions due to equipment failures, severe weather conditions, or maintenance activities occur periodically. Emergency buses serve as a timely and reliable alternative, effectively mitigating the adverse impacts of metro service disruptions. This study focuses on optimizing emergency bus dispatch in the context of metro operational disruptions. To prevent congestion at turnaround stations and improve passenger evacuation efficiency, we propose a flexible shuttle strategy based on dynamic passenger demand with deadhead routes, considering both turnaround stations and nearby transfer stations in disrupted areas as potential destinations. By adopting rolling horizon optimization, the model can adapt to variations in passenger demand in real time, dynamically adjusting emergency bus dispatch plans. A multi-objective nonlinear optimization model is developed to minimize total evacuation time, passenger travel delay and penalty costs, which is then linearized and solved using Cplex. The feasibility and effectiveness of the model are validated through a simulation of a sudden disruption during the morning peak on Shenzhen metro line segment. Results show that the proposed scheme outperforms traditional approaches in reducing both total evacuation time and passenger delay.

ABSTRACT. Urban transit is facing challenges on ridership decline in recent years, sparking debate on the appropriate extent of transit service provision. To incorporate internal and external effects of transit, this study proposed a planning framework to minimize total social monetary cost integrating decisions of network design, frequency setting, fleet size, and government subsidies in the multimodal system with transit and automobiles. A Particle-Swarm-Optimization-based algorithm was designed to solve the problem, which exhibited effectiveness on saving social cost and improving transit service quality. Effects of transit speed and value of time were analyzed to better understand transit provision under different scenarios.

ABSTRACT. Seoul’s bus and subway system was modelled as a time-dynamic multiplex graph to simulate cascading failures. Validation against strike and protest logs shows buses are fragile to random shocks, whereas subways are vulnerable to centrality-targeted attacks. When vulnerabilities overlap, a boundary section triggers abrupt network fragmentation. The model accurately reproduced passenger rerouting and segmentation patterns observed in smart-card data. A tiered response protocol integrating timed detours, real-time information and adaptive schedules offers actionable guidance for operators to reinforce resilience and reduce disruption costs.

ABSTRACT. Bus Rapid Transit (BRT) lines provide a fast and reliable alternative to regular bus systems at a lower cost than building rail infrastructure. They can enhance public transportation and thus support more sustainable travel. However, BRT lines often require upgraded stops, meaning additional investments are necessary. BRT lines typically span multiple municipalities responsible for the investments, often with varying interests. In this paper, we address the problem of selecting optimal routes and stops for BRT lines considering municipal interests. Moreover, we evaluate our proposed model through a case study in Greater Copenhagen, highlighting the relevance of considering the different municipalities.

ABSTRACT. Transit reliability is important for the agencies and passengers. Most of the existing literature focused on the reliability experienced by actual passengers or on the accessibility of destinations for potential passengers. However, we can still ask the question what would have been the service reliability for passengers traveling on low-demand origin-destination pairs, and what would be the potential reliability for potential passengers making regular trips across the whole service area? We propose to analyze the potential transit reliability by simulating passenger trips across the entire service area using a routing engine and archived service delivery data. Then we evaluate the planned against the delivered travel times to determine the potential reliability across the region using discrete metrics similar to on-time performance and continuous metrics like buffer times. Finally, we try to discover potential spatial and temporal variations of these calculated potential transit reliability.

ABSTRACT. Transit operators worldwide are transitioning to electric buses to mitigate greenhouse gas emissions. This introduces new challenges for the operators and policymakers as electric buses depend on electricity, increasing the significance of a city’s power grid infrastructure. Any power grid failure has the potential to disrupt the advertised schedule of the electric bus service by introducing delays and trip losses, leading to unmet passenger demand. This study proposes a methodology to design an electric bus system considering the relationship between a city’s power grid and its electric bus system. The proposed framework is formulated as a Mixed Integer Linear Program (MILP) and attempts to minimize the worst-case trip losses in the network with the possibility of spatial failure of power grid substations in a city. The framework is then applied to the bus and power grid network of Delhi, India. Results indicate that the proposed methodology can reduce the potential trip loss in the network by 30 % with marginal additional costs from a cost-optimal electric bus system. The study highlights the need to diverge from the typical cost-optimal design and leads to policy recommendations regarding the power grid infrastructure and electric bus system design of a city.

ABSTRACT. Although traditional public transportation, known as a fixed-route transit (FRT) system, is a cost-efficient transit mode in areas with high demand, it is often perceived as inconvenient due to the lack of flexibility. On the other hand, demand-responsive transit (DRT) systems, known as a flex-route transit system, have a high per-capita operating cost due to their personalized nature. This research proposes a bi-level optimization approach to design a hybrid transit system integrating FRT and DRT systems. At the upper level, the FRT network design is optimized, incorporating the optimized routes and schedules of DRT vehicles determined at the lower level.

ABSTRACT. This paper aims to design a network of bus lanes connecting the key bus stops (bus interchanges) to minimize traffic impact. A Steiner tree of bus lanes would be one that gives maximum benefits to bus passengers (the minimal passenger-km Steiner tree) while minimizing the impact on individual traffic (the minimal additional vehicle hours). This problem is formulated as a bi-level programming with the user equilibrium (UE) condition of individual traffic as a constraint. We provide a new sensitivity analysis method that assumes bush-based UE assignment for this problem, which has a vast solution space, and propose an efficient method to obtain an exact solution.

ABSTRACT. This study examines the trade-offs between ridership-maximization and coverage-maximization approaches in transit network design. Using historical ridership data, smart card fare collection, and service patterns, two network scenarios were developed under constant resource constraints. The ridership-maximization scenario projects 9% ridership growth while serving 48% of the population, whereas the coverage-maximization scenario maintains 74% population access with 6% ridership growth. Both scenarios demonstrate a 4.5% immediate revenue loss compared to the existing network, with potential ridership growth of 9% and 6% respectively, suggesting that network optimization requires careful balancing of efficiency and accessibility objectives.

ABSTRACT. This study investigates the reliability of bus-to-bus transfers in the network of Dutch regional bus operator EBS based on anonymized smartcard, GTFS and AVL data. The reliability of a transfer is calculated by evaluating both the cancellation rate of trips and the punctuality of the two buses. More than 50% of transfers are made within 5 minutes while there are little synchronized transfers in the area. Unreliable transfers are less common while more than 50% of made transfers are feasible in more than 80% of the time. Unreliable transfers have the highest potential waiting time when missing the transfer.

ABSTRACT. This study proposes an integer programming model to optimize demand-responsive connector (DRC) vehicle routing, incorporating multi-passenger requests and route travel time constraints. The model aims to minimize total costs, including fleet size, vehicle operating expenses, and penalties for deviations from passengers' desired boarding times. A tailored meta-heuristic algorithm is developed to solve the model efficiently, ensuring scalability for real-world applications. The model and algorithm are validated through a real-world DRC case study, demonstrating its effectiveness in balancing operational efficiency and passenger satisfaction. The results highlight the model's potential to enhance DRC systems by reducing costs and improving service quality.

ABSTRACT. The electrification of ferry systems introduces substantial operational challenges due to high energy demands, charging requirements, and berth limitations. This paper proposes a Mixed Integer Linear Programming (MILP) model to optimize vessel scheduling and timetabling for electrified public transport ferries. The model minimizes a convex combination of fleet size and unproductive rebalancing time, while explicitly incorporating charging constraints, berth capacities, and crew rest periods. To ensure scalability, we develop problem-specific heuristics and apply them to a real-world case study in Sydney, Australia. The model generates feasible, efficient itineraries that outperform the benchmarks. Notably, results show that with sufficient charging infrastructure, fleet size can be reduced without significantly increasing idle time. On the other hand, electrification of the Sydney ferry network could be unfeasible depending on the charging technology. This framework supports strategic planning for electrified public transport ferries, contributing to the decarbonisation of the transport sector.

ABSTRACT. This paper investigates an integrated dial-a-ride problem combining on-demand vehicles and existing mass transit services to minimize both bus operation costs and customer inconvenience. A hybrid large neighborhood algorithm is developed with problem-specific destroy and repair operators to address the routing complexities involving both modes. The algorithm is benchmarked against a state-of-the-art commercial mixed-integer programming solver on instances with 10-50 customers and two transit lines. On average, our approach delivers solutions of 23.8% higher quality in just 136 seconds, compared with the solver’s eight-hour run time. Moreover, a case study using real microtransit data compares the performance of the proposed service against the microtransit service, public transport and private car. Results indicate that the integrated service reduces vehicle kilometers travelled by 36% compared to microtransit service. It also significantly reduces customer travel time over public transport, while customers traveling by private car remains only 19% faster.

ABSTRACT. Limited and dispersed demand in rural areas challenges traditional public transportation. To address this, the government has introduced demand-responsive services with flexible routing and begun using passenger transport resources for freight to boost revenue. This study proposes an integrated passenger-freight transport model using configurable vehicles to improve efficiency and reduce costs. By adjusting passenger-freight configurations and minimizing reconfiguration inconvenience, this approach leverages freight revenue to enhance rural transport operations. Case studies show that this model improves efficiency and decreases operational costs compared to separate transport methods.

ABSTRACT. In passenger rail operations, train unit and driver scheduling are typically solved separately in a sequential way. This research proposes an integrated model combining these tasks for optimal schedules. A pre-generation strategy for driver shifts is used, but the vast number of potential shifts complicates solving large instances. To address this, a hybrid heuristic iteratively activates key empty-running trips and relief opportunities, reducing complexity while retaining critical solution components. Tests on synthetic timetables compare heuristic results with a direct ILP solver and sequential scheduling. The integrated model achieves cost savings, while the heuristic improves computational efficiency.

ABSTRACT. A multi-period asymmetric transit frequency design problem is formulated for a bus operation strategy, in which a class of buses serves both directions while the other class only serves one direction with high travel demand for each route, to address demand variation and demand asymmetry. An enhanced artificial bee colony algorithm is proposed to solve it. Numerical experiments demonstrate that the proposed algorithm can produce better solutions compared with the modified genetic algorithm and the proposed design outperforms the existing design with less passenger travel time and greater demand satisfaction, operating profit, and social welfare.

ABSTRACT. On the day of operations, disruptions can occur on the railway network, which require the rolling stock plan to be adjusted. In this paper, we consider the problem of rolling stock rescheduling with disruptions for which information about the duration, location and severity is uncertain and becomes available dynamically. In current practice, it is typical that the most recent disruption information is naively used to reschedule the rolling stock, which can lead to unnecessary cancellations or decreased passenger satisfaction in case the disruption turns out different than expected. We propose a rolling horizon stochastic programming approach for real-time rolling stock rescheduling, which has the ability to anticipate changes in the disruption information. The approach takes into account a variety of disruption scenarios and iteratively creates interim rolling stock schedules with a rolling horizon. Computational experiments are conducted on instances of the Dutch railway network. Compared to the naive rolling stock rescheduling approach, the proposed approach successfully creates rolling stock schedules with fewer cancellations and lower rescheduling costs for disruptions with changing information, in short computation times.

ABSTRACT. In China, the government-led approach to Mobility as a Service (MaaS) implementation presents unique challenges and opportunities. This study examines the dynamics between government, bike-sharing companies, and metro operators in China's MaaS implementation using evolutionary game theory, analysing how to balance stakeholder interests. The findings reveal government leadership is essential for stable cooperation, while subsidies speed up but don't determine final outcomes. Increased regulation promotes faster operator cooperation. The study provides policy recommendations including proactive government strategies, targeted subsidies, and public-private partnerships, which are valuable for policymakers with government-led transportation initiatives, offering a framework to balance public oversight with market.

ABSTRACT. Airline planning involves complex, large-scale combinatorial optimization problems that are typically addressed sequentially. However, disruptions in operational conditions introduce new constraints, requiring adjustments to aircraft rotations, tail assignments, and fleet allocation. This study proposes a dynamic assignment model that integrates these elements to enhance operational flexibility while maintaining cost efficiency. The problem is formulated as a mixed-integer linear program using a directed acyclic graph representation. Computational experiments over a two-day planning horizon, based on data from a European airline network, show that the model ensures routing feasibility, meets maintenance requirements, and effectively optimizes refleeting decisions.

ABSTRACT. This study evaluates accessibility via public transport and satisfaction levels to hospitals in urban environments using doubly constrained gravity approaches. By balancing origin demands and destination capacities, these approaches offer a realistic assessment of accessibility. A web-based application was developed to compute and visualize results for Rome, Italy. Findings reveal significant disparities, with central areas having higher accessibility and satisfaction. Moreover, the relaxed approach demonstrated higher satisfaction levels compared to the strict approach by allowing partial utilization of destination capacities and better accounting for real-world constraints. These results highlight the need for targeted interventions to improve equity in underserved zones.

ABSTRACT. Sustainable revenue balance is a critical issue for the global railway industry, requiring collaboration among all organizations. This research develops multi-level programming models based on market roles: the infrastructure manager sets access charges, the train operator determines ticket pricing and service plans, and passengers generate demand. A case study demonstrates the models’ suitability for current market conditions, providing valuable insights for revenue analysis and future railway reforms.

ABSTRACT. This study presents a service zoning framework that integrates Reinforcement Learning with a Genetic Algorithm within a Demand-Responsive Transit network. The objective is to demonstrate the model's effectiveness, particularly in adapting to varying travel demand patterns and improving upon initial solutions. The reward function is designed considering key elements of the DRT cost function, ensuring alignment with operational goals. Evaluation on a 10x10 grid network demonstrates the framework's convergence and adaptability across different scenarios, with optimized zones aligning closely with demand patterns. This data-driven approach offers a practical solution for improving transit operations and holds promising potential for real-world applications.

ABSTRACT. The rapid growth of online shopping has led to an increase in delivery vehicles on urban roads, exacerbating traffic congestion and greenhouse gas emissions. This study introduces the Bus-to- Bicycle (B2B) delivery framework, which integrates parcel transportation into existing public bus networks and leverages reward-based participation from bicyclists for last-mile delivery. The proposed framework addresses a complex multi-agent matching problem involving the assignment of packages to buses, allocation of locker space, and coordination with available bicyclists for final delivery. To reduce computational complexity, the model restricts its search to a set of pre-identified feasible matches. Results demonstrate that the B2B framework reduces vehicle miles traveled, enhances the utilization of underutilized bus capacity, and promotes the use of active transportation modes.

ABSTRACT. This paper studies the problem of scheduling extra trains into an existing timetable where the existing timetable is allowed to change and the schedule for the extra trains has constraints on their departure-time windows. The objectives are to minimise alterations to the existing timetable and the total travel time of extra trains. We model the problem as an Integer Linear Program (ILP) using a space-time network and employ the Alternating Direction Method of Multipliers (ADMM) framework to decompose it into subproblems per train, each solved using a time-dependent shortest-path algorithm. Computational tests on a real-life double-track high-speed railway network demonstrate the effectiveness and efficiency of the proposed approach.

ABSTRACT. The emergence of Automated Mobility-on-Demand (AMoD) services, such as Waymo and Zoox, is reshaping urban transportation. Yet, the increasing demand and expanding robotaxi fleets may worsen traffic congestion. This study presents a novel centralized ride-matching framework designed to improve the scheduling efficiency of Shared-Ride Automated Mobility-on-Demand Services (SRAMODS). The proposed adaptive, multi-objective model incorporates the perspectives of on-site, in-vehicle riders, as well as robotaxi operators, by minimizing on-site waiting times, in-vehicle travel durations, and detour distances. The model operates dynamically within each time epoch, with each robotaxi able to serve up to four riders concurrently. A case study using the 2019 Chicago taxi data demonstrates that varying objective weights yield different match outcomes, while a balanced weighting configuration minimizes total time expenditure. Compared to conventional decentralized approaches, the SRAMODS framework reduces the price of anarchy, measured by distance traveled per rider, by up to 24%. These results underscore the value of centralized coordination in promoting shared robotaxi adoption, offering policy guidance to enhance urban mobility and reduce system inefficiencies and congestion.

ABSTRACT. This study presents a multi-objective optimization model addressing three key problems: timetable design to enhance service regularity (social objective), vehicle scheduling to minimize operational costs (economic objective), and electric vehicle acquisition to maximize electrified kilometers (ecological objective). Based on mixed-integer linear programming and the epsilon constraint technique, the model explores trade-offs between objectives and approximates the Pareto front. Results demonstrate its utility in understanding the impact of electric bus adoption on public transport systems. Future work focuses on larger instances, efficient solution methods, and incorporating factors like battery lifecycle and seasonal operating conditions.

ABSTRACT. Designing routes for community transit systems operated by municipalities often faces the challenge of limited data availability due to constrained budgets and human resources. This study applies the Transit Network Design Model, which assumes a spanning tree structure, and investigates whether the model, initially developed for urban areas, can also be effectively applied to less populated regions. Furthermore, it explores the use of Mobile Device Location (MDL) data to simulate scenarios where conventional usage records are unavailable. The results reveal that network designs derived from MDL data closely resemble those based on actual usage records.

ABSTRACT. In order to reduce the probability of accidents caused by the intrusion of foreign objects at railway crossings, there is great potential for foreign objects detection in advance. This paper uses cameras to collect images of foreign objects at railway crossings and divides the objects around the track into 10 categories through a self-made dataset. In addition, the popular YOLO series models are used for foreign object detection and comparative analysis based on the self-made image dataset. In the experimental results of the dataset proposed in this paper, the mAP50-95 and mAP50 can reach more than 70% and 90% respectively.

ABSTRACT. Successful implementation of on-demand feeder bus services requires thorough analysis to ensure their effectiveness and public acceptance. However, policymakers lack tools to identify suitable locations for introducing such services. This study proposes a utility-based simulation framework to evaluate area attractiveness by assessing service key performance indicators under uncertain demand. We report a case study comparing Bronowice and Skotniki areas in Krakow, Poland, where such services are planned. Based on the results of the simulation experiment, Skotniki area demonstrated greater potential for feeder attractiveness and added value across a wide range of alternative-specific constants, which we assume unknown.

ABSTRACT. Optimizing driver scheduling in passenger railway systems poses a critical operational challenge with direct implications for resource efficiency and service reliability. Given the problem's NP-complete nature, effective solution strategies must balance computational tractability with practical feasibility. This study proposes a novel three-phase methodology integrating depth-first search, a weight-based selection heuristic, and local search optimization. In the first phase, depth-first search generates feasible driver duties that comply with contractual and operational constraints. The second phase assigns duties using a weight-based heuristic to construct complete schedules that cover all required trips. The final phase applies local search to refine solutions through neighborhood-based improvements. Computational experiments using real-world data from a railway system with over 800 trips demonstrate the approach’s scalability and high solution quality. The results confirm the practical applicability of the proposed method to large-scale railway crew scheduling problems.

ABSTRACT. Urban transit systems frequently experience bus bunching, crowding, and idling, which degrade service quality. iROAM is an integrated toolbox for anomaly detection, data preprocessing, and visualization that fuses Automatic Vehicle Location (AVL), Automatic Passenger Counting (APC), and General Transit Feed Specification (GTFS) data. Its configurable pipeline allows fast threshold adjustments for labeling multiple anomalies. Through a case study on Toronto’s Route 29, iROAM demonstrates enhanced diagnostic capabilities and real-time adaptability, informing timely interventions. Future development integrates predictive modeling to provide proactive alerts, further supporting transit agencies in mitigating service disruptions and improving overall operational efficiency.

ABSTRACT. This study develops a dynamic system under an efficient and equitable reservation-based travel demand management scheme, named booking cum rationing (BCR), designed to converge to dual user equilibria: a daily equilibrium and a periodic equilibrium. Under the BCR scheme, travelers have limited opportunities to book restricted links during a period while facing endogenous uncertainty of failing to use restricted routes due to overbooking, where booking demand exceeds link capacity. Cost-minimization behavior and endogenous uncertainty encourage travelers to strategically budget their booking chances to mitigate periodic travel costs. We formulate travelers’ sequential decision-making as a Markov decision process to capture day-to-day flow dynamics within each period. To ensure daily bottleneck capacities, a projection-based quadratic optimization model is proposed to determine the daily success rate for bookings on restricted routes. Additionally, we examine a period-to-period learning method to describe travelers’ learning behaviors under this periodic scheme. The existence of the equilibria is established, and we further explore a range of properties of the equilibria, such as the connections with the static equilibria in the BCR system. Preliminary findings indicate that, under mild conditions, the proposed dynamic system converges to the dual user equilibria.

ABSTRACT. The rapid growth of e-commerce has created significant challenges to last-mile delivery, driving the adoption of drones for fast and cost-efficient service. However, the limited battery capacity of drones necessitates collaboration with trucks or access to charging stations to extend their operational range. Although trucks can provide line-haul transportation from distribution centers to customers, they contribute to traffic congestion and often face restricted access in small communities. Charging stations require substantial investment on infrastructure. This paper proposes an alternative solution, i.e., integrating drones with bus networks. Buses can carry drones, recharge their batteries, and deploy them at bus stops nearest to delivery destinations. Integrating drones with buses could help avoid traffic congestion caused by trucks, reduce the total delivery times, and provide additional revenue opportunities for bus operators. This study addresses operational challenges in this integrated system, such as synchronizing drone operations with bus schedules and routes, determining optimal stops for landing and taking off, and managing capacity constraints in terms of bus space and drone battery usage. We introduce a novel variant of the drone-routing problem, termed the bus-assisted drone routing and charging problem, and develop a branch-and-price-and-check algorithm to find the exact solutions. The proposed method offers a promising solution for deploying drones in last-mile delivery systems.

ABSTRACT. In developing countries, paratransit refers to informal transportation services which cater to populations that either lack access to public transportation altogether or lack access to sufficient, high-quality service. A variety of paratransit modes operated by small-scale operators, legally or illegally continues to play a significant role in the urban transport systems as these forms of transportation are able to respond to shifting market demands and fill the gaps which are unserved by formal public transport services. While their gradual elimination has been suggested as the way forward, paratransit operations may also support more formal public transportation. The study presents the development of prototype machine learning-based planning and decision-making tool using crowdsourced vehicle tracking and ridership data. This system integrates with the pilot deployment of the SafeTravelPH public transport crowdsourcing and information exchange platform in General Santos City, Philippines to ingest the data feeds into an open dashboard which can quickly and intuitively provide visualizations and querying tools in order to assist local government units and transport stakeholders with public transport policymaking. The initial test of the system has garnered approval and support from local government units and public utility jeepney cooperatives towards greater compliance with the Public Utility Vehicle Modernization Program (PUVMP) which the government launched in 2017.

ABSTRACT. The transition from diesel-powered bus fleets to electric vehicles (EVs) is a crucial step toward achieving sustainable urban transport systems, but the adoption of electric buses involves complex decision-making processes, including determining the optimal timing and quantity of EV acquisitions, selecting the most suitable technology, and assigning vehicles to specific bus lines. This study presents a multi-objective optimization approach to address these challenges, incorporating constraints such as budget, maximum vehicle age, total electrification at the end of the planning period, among others. Our problem aims to minimize total costs (economic objective), optimize a gradual fleet electrification (ecologic objective), and fostering an equitable distribution of electric vehicles across different urban regions (social objective). We implement an epsilon-constraint algorithm that effectively approximates the Pareto front, enabling decision-makers to evaluate trade-offs and select solutions that align with their priorities leading to a significant contribution to the growing body of research on sustainable transport systems.

ABSTRACT. Railway line planning and timetabling are closely linked, with their integration improving service quality for operators and passengers. This research focuses on jointly optimising both while considering economic and societal benefits. Beyond operational efficiency, this approach promotes a sustainable, inclusive, and passenger-centric transport system and serves as a decision-support tool for better planning strategies in new railway service development.

ABSTRACT. The Vehicle Scheduling Problem (VSP) assigns vehicles to trips in a transit system to minimize the number of vehicles or the cost of deadheading. This problem is usually formulated assuming all vehicles start and end at depots within a specific time horizon (say for 24 hours). Operations are typically assumed to go through a complete reset at the end of the time horizon. However, there are always some active trips (late night and early morning) in real-world settings. This work focuses primarily on scheduling trips that can be periodically executed for such scenarios. A mixed-integer programming formulation is first derived to capture these features and is compared with the classical VSP. Results from airport shuttles in Bengaluru, India, demonstrate the proposed model's utility.

ABSTRACT. Adaptive capacity is an important aspect of urban rail transit network (URTN) resilience, indicating the ability to mitigate negative effects of disruptions via emergency actions like the operation of complementary shuttle bus services. Complementary shuttle buses help transport metro passengers stranded at disrupted rail lines, but could lead to even more chaos owing to their limited capacity under severe disruptions. This study develops a multi-modal network capacity model to facilitate understanding the adaptive capacity of URTN. Both supply- and demand-side adaptation stresses will be considered, namely the insufficient capacities of shuttle bus services and the heterogeneous adaptation behaviors of metro passengers.

ABSTRACT. This paper presents the multimodal vulnerability network model (MVNM), which determines the critical links to assess the effects of multimodality on disruption management. Therefore, we combine a path-based multicommodity approach with timetable sensitive passenger routing to optimally adjust the operating services under disruptions. The resulting MVNM is solved by combining multi-column generation and row generation, to iteratively identify beneficial passenger routes and disruption management measures. The MNVM is applied on the long-distance air-rail network of Spain. The results show, that multimodality increases the survivability of a network. However, multimodal networks appear to be more vulnerable under few disruptions.

ABSTRACT. Transit network design in the phase of transit route design and frequency setting primarily aims on the minimization of passenger and operator related costs, neglecting vulnerability sourcing from its critical infrastructure. This study focuses on enhancing the resilience of public transit networks by incorporating vulnerability analysis into the transit network design problem. After estimating a set of candidate solutions with their critical infrastructure, we proceed to the optimization of frequencies of the solution that resulted to the lowest disruption cost. Simulated annealing is used to adjust the frequencies of the transit lines subject to fleet availability and capacity constraints. Applied to a toy and a real-world network, the results show that the final solution experiences significantly limited losses from disruption, while offering a result comparable to the other candidate solutions that their objective is limited to cost optimization.

ABSTRACT. In a typical tactical timetabling process, route planners must follow a complex and iterative process to develop new timetables, often relying on trial-and-error methods. To guide practitioners throughout their decision-making process, this study introduces a fragility-based approach to timetable design and shows how route planners can exploit the concept of fragility to design more robust timetables. We propose a MILP model that aims to enhance timetable robustness by focusing on its most critical part. Considering real-life scenarios from a Norwegian railway line, we show that we can improve the fragility of a timetable, even when adopting conservative improvement strategies.

ABSTRACT. Integrated transport hubs are critical node-based infrastructures for achieving sustainable public transportation development. Accurately forecasting the distribution of passenger flows within the hub is essential for enhancing operational efficiency, ensuring passenger safety and effectively responding to extreme passenger volumes and emergencies. To thoroughly investigate the evolution mechanism of passenger activities within hub spaces, this study integrates multimodal data sources, including monitoring video data, passenger behavioral experiment data within a digitally hub environment, train schedules, and ticketing records. By analyzing the topological characteristics of spatial mobility networks of passenger groups and the causal impacts of public transport operational events on regional passenger flow fluctuations, this research develops a passenger flow distribution prediction model (GEME-Net) that incorporates group evolution mechanism and regional network topological features. Furthermore, knowledge distillation techniques are applied to achieve lightweight model deployment, effectively balancing prediction accuracy and real-time inference performance . Experimental results at the Shanghai Hongqiao integrated transport hub demonstrate significantly improved prediction accuracy compared to baseline models, validating the effectiveness and superiority of the proposed spatial evolution mechanism and multimodal data fusion approach.

ABSTRACT. Accurate and timely predicting onboard passenger loading is essential for enhancing public transport efficiency, service quality, and safety. Considering the difficulties of existing model-based approaches in capturing complex spatiotemporal patterns and interconnected network structures, this paper presents a data-driven approach that utilizes dynamic graph convolution, multiple graphs, and relational interdependency to predict passenger loading across multiple urban commuter train lines. The model is evaluated using automatic passenger count and automatic vehicle location data from the Greater Helsinki region's commuter train system. Experimental results demonstrate that the proposed model achieves a very good prediction performance, superior to that of a set of tested baseline approaches.

ABSTRACT. Real-time railway rescheduling is crucial in response to unexpected passenger conditions in a timely and flexible manner. Current passenger-oriented rescheduling strategies primarily focus on offline planning based on static demand data. Uneven distribution of demand over time is often neglected, despite its dynamic nature during long-term disruptions. This study defines a real-time demand-responsive rescheduling problem that addresses disasters at a railway hub station, focusing on five key challenges: uncertain disruption duration, dynamic passenger mobility, rolling stock insufficiency, multi-route balancing, and in-station overcrowding. We propose a novel data-driven approach leveraging real-time mobile data to capture passenger mobility. A hierarchical deep reinforcement learning framework, consisting of two cooperating agents, is introduced to handle the sparse reward environment. A real-world disaster case demonstrates the advantage of our approach in sparse reward navigation and multi-objective balancing, compared to global search, rolling horizon, and single-level deep reinforcement learning algorithms. The online applicability is validated by transferring pretrained agents to new environments with varying rolling stock availability and passenger demand distributions.

ABSTRACT. This study develops a data-driven framework to estimate elderly transit demand in Winnipeg, a single-mode, bus-only city, with a focus on spatial and seasonal variations to inform Demand-Responsive Transit (DRT) zoning strategies. Recognizing the uneven distribution of older adults and seasonal shifts in travel behavior, this study generates monthly Origin-Destination (OD) matrices by combining boarding data with symmetry-based alighting estimations, calibrated through a weighted averaging technique. Spatial and temporal analysis reveal significant seasonal variations in trip distance sensitivity: longer trips dominate during low-sensitivity periods, while localized travel increase during high-sensitivity periods. These findings underscore the need for adaptive DRT zoning, recommending radial layout with smaller, denser zones in warmer season and larger zones in colder season. The proposed approach addresses challenges associated with incomplete data in smaller transit systems and provides actionable insights to support the planning of cost-efficient, equitable, and elderly-focused DRT services.

ABSTRACT. The rapid advancement of artificial intelligence has accelerated the potential for shared autonomous vehicles (SAVs). Dynamic Wireless Charging (DWC) technology has emerged as an ideal solution to enable the continuous operation of SAVs, allowing SAVs to charge while in motion. However, conventional vehicle routing algorithms are inadequate for SAVs' pickup, delivery, and charging problems, lacking flexibility in dynamic traffic. This study introduces a Two-Stage Deep Reinforcement Learning framework (TDRL) to address these challenges. The proposed framework operates in two stages: developing global routing strategies that consider route costs, state of charge (SOC), and route re-planning strategies that can handle dynamic traffic. In the first stage, the TDRL uses a heterogeneous attention mechanism to integrate diverse node information for optimal DWC node selection. In the second stage, using the first stage result as a baseline, the framework reassesses current traffic to enable strategic route adjustments, including deleting and reinserting node pairs under a dynamic Mask scheme. The experimental results show that TDRL not only outperforms existing heuristic algorithms in reducing route costs but also effectively maintains SOC stability. Furthermore, TDRL significantly mitigates the impact of traffic fluctuations on route costs by 16.2\% for the 50-node model and 23.7\% for the 100-node model, as well as on SOC distribution.

ABSTRACT. Max-Pressure (MP) control is a decentralized real-time traffic signal control method that is popular for its simplicity and theoretical stability. However, most existing MP controllers prioritize throughput for private vehicles without accounting for the specific needs of transit services that are essential for sustainable urban mobility. This oversight can exacerbate transit delays and undermine the effectiveness of public transportation systems. To address these challenges, this study introduces a Priority-MP framework that integrates transit signal priority and driver advisory systems into MP control for multi-modal traffic networks. By weighting pressures based on real-time vehicle occupancy, Priority-MP prioritizes high-occupancy transit vehicles while guaranteeing network queue stability. In addition, the framework considers more realistic scenarios with the presence of transit stations and integrates driver advisory systems to provide speed and dwell time recommendations. Simulations on a real-world multi-modal traffic corridor in Amsterdam show that compared to existing MP control methods: 1) Priority-MP significantly reduces average passenger delay when only vehicle occupancy is considered, but significantly increases average vehicle delay; 2) Priority-MP considering transit stations further reduces both vehicle delay and passenger delay while maintaining the network stability; and 3) Priority-MP integrating driver advisory systems further reduces transit queuing counts.

ABSTRACT. This study proposes a novel framework integrating a cell-based passenger dynamics model (CPDM) and simulation-based optimization (SBO) to analyze spatiotemporal passenger flow dynamics on metro platforms. The proposed model achieves high behavioral accuracy by incorporating heterogeneous passenger characteristics and multi-destination patterns. The CPDM is cast into a computational graph to enable calibration by iterative backpropagation (IB) algorithm, that uses automatic differentiation to derive the analytical gradient and iteratively refines model parameters. The methodology demonstrates superior performance in capturing complex passenger behaviors while maintaining computational tractability. The results provide valuable insights for metro station design and crowd management.

ABSTRACT. This paper introduces a novel framework for modelling bus bunching under multiple uncertainties stemming from demand, supply, and exogenous factors, unified within a comprehensive and scalable model featuring microscopic granularity. The methodology is validated through a real-world case study enabling the measurement and quantification of bunching sources and their cascading effects. The findings reveal the model's capability to accurately capture day-to-day variability and differentiate between distinct types of bunching. This advanced approach facilitates informed, targeted and context- specific decision-making by allowing the evaluation of control measures and their effectiveness across dynamic conditions including peak and off-peak periods, varying congestion levels, and fluctuating passenger demand.