Growing demand for rail services is driving investments to improve sustainability, efficiency, and safety. Traditionally, these goals were pursued through large-scale infrastructure projects, but they are complex, costly, and prone to delays and budget overruns. Technologies such as Automatic Train Operation (ATO) are emerging as potential alternatives or complements, promising gains in punctuality, efficiency, and capacity. Limited research exists on the comparative economic feasibility of infrastructure and ATO. This paper develops a framework for comparing the costs and benefits of infrastructure investments and ATO, using key performance indicators (KPIs) that relate costs, both per kilometer and at project level, to capacity and punctuality improvements. The framework is applied to Norwegian cases, drawing on front-end analyses of infrastructure projects and ATO use cases. The results show that infrastructure remains more cost-effective for large-scale capacity increases, while ATO is more favorable for punctuality improvements and incremental efficiency, particularly in dense corridors. Findings suggest that infrastructure and ATO should be viewed as complementary: infrastructure providing the foundation for capacity expansion, and ATO enhancing reliability and efficiency. The study contributes a structured, KPI-based framework for comparative analysis and emphasizes the need for future research to incorporate broader benefits and reduce uncertainty around ATO costs.
This paper presents an optimization framework for microscopic railway timetabling that combines Logic-Based Benders Decomposition with Conflict Discovery (CD-LBBD) alongside a Soft-Conflict Mixed-Integer Programming (SC-MIP) formulation. The CD-LBBD method separates high-level coordination from microscopic feasibility, employing a subproblem based on satisfiability modulo theories to dynamically detect and resolve resource conflicts. The SC-MIP formulation relaxes disjunctive precedence constraints into soft penalties and iteratively minimizes conflict violations to obtain feasible schedules, which are subsequently refined to optimality. Both approaches operate on a strengthened continuous-time event-activity formulation.The framework runs CD-LBBD and SC-MIP in parallel and is evaluated on the complete DISPLIB 2025 benchmark set, comprising 112 real-world-based instances. Within a 10-minute competition limit per instance, CD-LBBD and SC-MIP jointly solve 65 instances to optimality, while the largest instances remain challenging for both methods. The parallelized solver achieves 91 best-known solutions – approximately 81% of the benchmark – and obtained the highest overall score in the DISPLIB 2025 competition. We evaluate the incremental contribution of the model’s strengthening techniques and show that they enable solving nearly four times as many instances to proven optimality. These results demonstrate that combining exact methods with complementary algorithmic structures provides strong and scalable performance for large-scale, microscopic railway timetabling.
Efficient train platforming is essential for maintaining feasible network timetables at the station level while minimizing operational disruptions. The train platforming problem (TPP) involves complex decision-making processes, including train rerouting, cancellations, and shunting operations within railway stations. This paper formulates the TPP as a mixed-integer linear programming (MILP) model, which integrates tactical adjustments to station operations to ensure alignment with the network timetable. To overcome the computational challenges inherent in solving the TPP problem, we propose a three-level iterative heuristic algorithm that efficiently resolves conflicts and facilitates train platforming and shunting operation adjustments. The algorithm ensures timetable feasibility across the rail network but also identifies problematic trains. The proposed approach is validated through computational experiments using instances derived from a real-world railway station and network timetable, demonstrating its capability to enhance station-level operations and network robustness while achieving efficient conflict resolution and operational stability.
Under the constraint of fixed maximum train capacity, we propose a nested optimization model that consists of inner and outer layers, designed to minimize the net energy consumption (NEC) with the concurrent objective of reducing the average passenger waiting time (PWT). The inner-layer optimization model incorporates penalty functions into its objective function to optimize the inter-station position-velocity curves. In contrast, the outer-layer optimization model aims to establish a Pareto front between NEC and average PWT by optimizing the spatiotemporal interactions of trains, accounting for dynamic passenger flow curves (PFCs). The inner and outer models are solved using single-objective and multi-objective optimization algorithms, respectively, and parallel computing is utilized to improve computational efficiency. Finally, numerical experiments are carried out on a real-world metro line, demonstrating that the nested model provides a range of compromise solutions for decision-makers.
Railways are essential for a fossil-free transportation system, yet the strategic planning tools for freight transportation on a long-term horizon remain scarce. Strategic freight transport models usually predict annual transportation demand, but the implications for railway capacity usage remain unclear. This missing link between national freight transportation models and strategic railway capacity management is a barrier for strategic decision-making in national transport and railway system planning. We propose a computational framework that includes a macroscopic Mixed Integer Linear Programming (MILP) model integrating fleet-sizing, flow assignment, and train scheduling problems, addressing this gap in strategic planning. The framework generates a cyclic national-scale rail freight transportation plan containing the flows of locomotives and loaded and empty wagons to fulfill a given transport demand. Using the entire Swedish rail network with real-world demand data, we show that our framework can produce national freight transportation plans that fulfill all demands and bounds on railway capacity and vehicle circulations. The framework allows for comparison of scenarios with different transport demand and capacity supply. The results highlight its potential as a tool for strategic decision-making in national freight transport planning.
Railway timetabling in Europe is shifting towards passenger-centered, connection-optimized network timetables with synchronized stops at major nodes. This changes structural properties, affecting capacity utilization and corridor robustness. This paper analyzes the effects resulting from this new timetabling paradigm by assessing the relationship between timetable structure and corridor capacity. It establishes an approach to systematically relate structural tendencies of timetable concepts to their implications for performance by examining distributions of timetable ensembles for specific concepts in a structural feature space. Using a heuristic ensemble generation method, we study timetables with varying structural properties. Features quantifying service characteristics are defined and evaluated for their impact on stability using Kendall’s correlation and regression. Two subclasses — periodic timetables with regular patterns and integrated (timed-transfer) timetables — are compared through their ensemble distributions in the feature space. For 3106 timetables, running time margins emerge as the dominant factor in corridor stability, especially when converting to timed-transfer timetables on existing, non-travel-time-optimized infrastructure. Results suggest that, with more structured timetables, running time supplements rather than tightly constrained buffer times deserve greater attention in capacity assessment and the dimensioning of service concepts on rail corridors.
The balance between budgeting and retaining the optimum performance of railway infrastructure has gained significant importance in recent years. Due to the expansion of railway networks, high maintenance costs, and limited budgets, prioritizing maintenance operations takes time and effort. On the other hand, derailments are one of the most essential types of rail accidents worldwide. Derailments typically result from track conditions, machinery issues, human error, etc. One approach to railway track maintenance budgeting is to allocate the available budget with the purpose of minimizing the financial consequences of derailments considering the accidents' occurrence probability. In this regard, the probability of accident occurrence must be calculated, and the financial consequences of derailment must be estimated. Then, the maintenance plan could be determined by using optimization methods and considering related constraints. In this paper, to calculate the probability of track-caused derailment accident occurrence, the Track Performance Index (TPI) presented by Janatabadi et al. (2020) is employed. The TPI metric is calculated for each block based on the track geometric data of curves in the railway network. Then, various Machine Learning (ML) models are trained and tested in order to estimate the mentioned consequences and the best one with highest precision is selected. The proposed ML model estimates the financial consequences based on the blocks' characteristics with a Normalized Root Mean Square Error (NRMSE) of 18.2%. In the next step, the Branch and Bound (B&B) method is utilized to optimize railway track renewal budget allocation in order to minimize the total probable consequences of accidents in the railway network. To validate the efficiency of the presented method, it is implemented on Iran's railway network. Compared to other methods, the results indicated that the proposed method decreased the network loss more and saved an amount of the maintenance budget.
Netherlands Railways is considering a shift towards a novel crew planning process featuring individual Sharing-Sweet-and-Sour rules, designed to provide each crew member with a fair and varied work schedule. This approach replaces the current, crew base-level fairness mechanisms. In the new process, template-based rosters specify generic time windows, to which personalised duties are assigned in the operational phase. To evaluate the feasibility of the novel process, we develop a scalable simulation framework that accurately models the operational crew planning phase. The framework features a column generation heuristic to construct personalised duties, a network decomposition strategy to reduce computing times, and stochastically generated disruptions to simulate daily operations. Simulating the process for 3265 guards throughout the year 2024, we find high conformance with the proposed individual Sharing-Sweet-and-Sour rules, with over 93% compliance for four out of six attributes. Our results yield valuable inputs for ongoing discussions with the works council and provide company experts with a powerful strategic tool. These insights are relevant to other transport operators, showing that fair and attractive individual work schedules can be constructed and highlighting the practical benefits of tailored simulation tools.
Rail transport is a strategic solution for moving goods and passengers, particularly in countries with high logistical demands. However, ensuring safety at grade crossings, where railways and roadways intersect at the same level, is a significant challenge for this sector. This paper presents a systematic literature review of the risks associated with grade crossings and the methodologies used to assess them. The study emphasizes the importance of an integrated approach to risk assessment that considers infrastructure characteristics, user behavior, and physical aspects of crossings. Through an analysis of related papers published within the last decade, the review identifies gaps in the literature, particularly regarding the integration of behavioral, technical, and physical variables. The review also emphasizes adapting methodologies to regional circumstances, especially in developing countries with limited infrastructure and resources. This research advances railway safety by critically analyzing recent studies on grade crossings, highlighting areas needing more attention, and suggesting directions for future investigations. The results provide valuable insights for developing more effective risk assessment models using tools such as multicriteria analysis, computational simulations, and artificial intelligence. The results pave the way for further research to improve safety at grade crossings with theoretical and practical applications in different contexts.
The Railway Container System under Passenger Transport Mode (RCSPTM) is an organizational strategy that applies passenger train operating practices to container transport, aiming to enhance transport efficiency. Its efficient operation requires the integrated coordination of three key planning tasks: train stop planning, timetabling, and maintenance planning. As these plans are highly interdependent, sequential optimization often leads to suboptimal results. To address this challenge, this paper proposes an integrated optimization approach, formulated as a mixed-integer linear programming model that embeds stop decisions into the timetabling process and represents maintenance operations as a “virtual train” scheduled together with real trains. The model addresses three aspects of train operations: it allocates container demand across services, avoids scheduling conflicts among trains, and minimizes disruptions from maintenance activities. The effectiveness of the model is validated through computational experiments using the CPLEX solver on both small-scale cases and the Beijing-Shanghai railway corridor. Compared with pre-planned maintenance approaches, the integrated optimization method eliminates unnecessary stops and reduces computation time by 47.5%, demonstrating its computational and operational efficiency. More importantly, it shortens the system-wide committable transit period from 3 days to 2 days, thereby enabling guaranteed delivery within stricter time windows for time-sensitive shipments.
Previous railway efficiency studies largely rely on aggregated data, which limits insight into the operational drivers of performance. Using a unique line-level dataset from a French region (117 observations), this paper leverages fine-grained information to explain within-region heterogeneity by linking cost efficiency to rolling-stock characteristics and maintenance practices. Results identify rolling stock as a central efficiency determinant. Lines operated with older, heterogeneous fleets are systematically less efficient, consistent with higher maintenance complexity and weaker standardization. Operating density also remains a key correlate, confirming economies of density. These findings are critical for the transition to competitive tendering. As networks fragment into bundles, line-level benchmarking aligns with procurement. The proposed framework distinguishes non-discretionary constraints from discretionary levers under the PTA’s and operator’s responsibility. Ultimately, these metrics provide a quantitative foundation for contract negotiation and investment planning, facilitating the arbitration between operational efficiency gains and the capital costs of fleet transformation.
Accurate prediction of train delays and their potential impacts is essential for effective dispatch adjustments and plays an important role in high-speed railway daily operation. Therefore, a hybrid method combining Spatial clustering, Cuckoo search, and Memory networks (SCM) is proposed to improve both the efficiency and accuracy of train delay prediction. This method uses hierarchical density-based spatial clustering of applications with noise (HDBSCAN) and k-nearest neighbour (KNN) classification algorithms to identify train delay evolution patterns. In each train delay evolution pattern, the cuckoo search algorithm, which ensures global convergence, was used to optimize the model hyperparameters, and the long short-term memory network predicted the train delay. On the basis of a real-world case study, with approximately 230,000 records of actual operations of the Beijing-Guangzhou high-speed railway in one year, the experimental results show that the train delay prediction accuracy can reach 93.495% with an allowable error of 1 min, which significantly improves the prediction accuracy and efficiency, and has practical application significance.
High-speed rail (HSR) faces growing competition from aviation. This study proposes a mixedinteger fare optimization model that jointly represents trains and flights on a corridor over a given time window. The model is designed to maximize operator revenue subject to passenger utility constraints, thereby treating passenger behavior as a binding condition rather than a direct optimization objective. Unlike bilevel models that separate upper-level fare setting from lowerlevel passenger choice, enhanced bounded-rationality constraints are embedded directly in the optimization, integrating pricing and behavior. A multi-layer spatio-temporal price-time network captures competitive interactions, service attributes, and passenger choices across multiple operators and services. Using Beijing-Shanghai data, the model attains an objective value of 8,617,549 in 2984 s. Compared with fixed pricing, HSR revenue increases by 33.38%. The policy also balances flows by raising off-peak seat occupancy (e.g., G103: 87.71%-*98.30%) and reducing peak congestion (e.g., G115: 100.00%-*89.02%), demonstrating the managerial value of well-designed floating fares for peak-off-peak load balancing and guiding demand-responsive fare design and capacity allocation under intermodal competition.
Railways play a critical role in the modern transportation system, with infrastructure increasingly shifting towards computer-based signalling systems to enhance operational efficiency and safety. As part of this evolution, the design and visualization of the man-machine interface (MMI) in these systems become essential, since it serves as a key tool for human operators, such as dispatchers, to monitor and control railway traffic. While these computer-based schematics offer improved information density and user interaction, they also impose a significant design workload. Incorrectly designed schematics can lead to operator errors, potentially jeopardizing safety. Hence, high-quality design is imperative. This paper introduces a novel approach aimed at automating the process of railway schematic design by employing mathematical optimization, specifically through integer linear programming (ILP). The central challenge of schematic layout design is framed as a mathematical modelling problem, where the placement of elements such as signals, points, and track circuits is optimized under a set of constraints. The proposed method automates the placement of these elements while allowing for fine-tuned manual control, ensuring the legibility and ergonomics of the final output. By formalizing the layout problem as an ILP, this approach ensures an efficient and structured solution process, taking into account both the technical constraints of the railway system and the human factors that influence operator decision-making. The method’s effectiveness is demonstrated through the development of software capable of generating high-quality railway schematics with minimal manual intervention. The results highlight the potential of mathematical modelling techniques, particularly ILP, in automating complex design tasks while maintaining a high standard of usability and safety. This contribution not only reduces the workload for designers but also improves the overall quality and reliability of railway control systems.
Dwell time is a critical component of railway operations, influencing network capacity, service reliability, and passenger experience. Despite extensive methodological development, discussion of how dwell time models are operationalised in practice remains limited in the publicly available literature, a situation that may partly reflect commercial confidentiality in real-world applications. To address this, the paper proposes a novel lifecycle-oriented, systems-theoretic framework to support the selection, calibration, and operationalisation of dwell time models in alignment with institutional capabilities, data environments, and planning objectives. The framework is informed by a structured, non-exhaustive review of railway dwell time modelling approaches, synthesizing statistical-based, simulation-based, and advanced models to examine how passenger behaviour, operational constraints, and uncertainty are represented across different operational contexts. Unlike prior reviews that focus on individual modelling paradigms in isolation, this study integrates insights across major approaches and aligns them with practical deployment considerations. By introducing a six-part lifecycle framework, this work provides a structured, actionable pathway for translating dwell time models into real-world applications. By bridging academic rigor with real-world applicability, the proposed framework offers a pragmatic pathway for agencies to leverage data-driven modelling for improved dwell time management by advancing the operational maturity and responsiveness of railway systems.
We study the Track Maintenance Possession Problem which schedules maintenance works on railway tracks so that the maintenance machines are efficiently assigned and traffic restrictions are not violated. In this particular version of the problem, the maintenance works should be assigned to predefined time slots, so-called containers. We are provided real-world data for the maintenance demands, the containers, the available machines and the traffic restrictions for the whole railway network in Germany for one year. We present a Mixed Integer Program formulation for the problem and give a mathematical proof that it is NP-hard to solve. In order to be able to solve the problem on the large instance size nonetheless, we propose a Maximum Satisfiability encoding and solve the problem with a state-of-the-art solver. In our result, 95% of the maintenance demands are fulfilled which is close to an upper bound we can provide.
Due to a lack of seat availability information regarding metro train carriages, passengers constantly move on platforms or inside carriages in search of seats, leading to overcrowding and safety concerns. Informing passengers of available seats and guiding them to less crowded carriages are beneficial for riding. The number of available seats at the next station is determined not only by the number of passengers currently occupying the carriage but also by the number of passengers alighting at the station. This study proposes a predictive framework that uses an optimized random forest model to estimate seat availability by analyzing metro passenger action features to predict alighting behavior. A dataset comprising 2009 passenger samples, encompassing 14 categories and totaling 9228 action features, was collected from video recordings in Changsha metro carriages. SHapley additive exPlanations (SHAP) analysis demonstrated that passengers’ actions as they approach the next station—particularly moving toward the carriage exit and standing up—are strong indicators for predicting alighting behavior. Certain combinations of actions, such as “watching the information display” and “standing up” showed superior predictive effectiveness compared with individual occurrences. The predictive model achieves high accuracy in assessing seat availability, contributing to improved service and boarding efficiency.
The escalating incidences of train-wildlife collisions, especially with elephants, are a real blow to forest-based wildlife conservation efforts. To enable real-time detection of wild animals such as elephants and promptly alert train operators, this study utilizes advanced deep learning algorithms integrated with camera systems installed at strategic locations along the railway tracks. The research proposes the use of WildlifeRailGuard as an innovative solution to address this issue effectively. To enable real-time detection of wild animals, such as elephants, and promptly alert train operators, this study utilizes advanced deep learning algorithms integrated with camera systems installed at strategic locations along railway tracks. Upon receiving alerts, train operators can immediately reduce speed, ensuring the safety of both passengers and wildlife. The proposed WildlifeRailGuard system also contributes to wildlife conservation by leveraging data analytics to generate valuable insights into animal behavior and movement patterns. The use of data analytics tools helps mitigate the negative effects of railway expansion on various animal species, fostering hope for achieving a sustainable balance between railway development and forest conservation.
In densely operated European railway networks the trade-off between network utilization and service quality leads to pondering traffic volume against the reliability of services. The perspective on service reliability, however, tends to vary between stakeholders. In this paper, we investigate and assess reliability in railway networks, particularly focusing on the comparison between railway operators and users in freight transport systems: While operators mainly focus on the punctuality of trains, users typically perceive reliability on the level of transportation chains. To combine both perspectives, a new agent-based simulation approach is developed that combines a microscopic domain-specific traffic control simulation incorporating the essential properties of the signaling system with a logistical perspective featuring the flow of goods within the network. The study outlines the methodical steps of model-setup, as well as the implementation on the backbone of an agent-based, open-source simulation environment. The methodology is applied to the German railway network and effects on the two different reliability perspectives are assessed and compared. A significant deviation of transport reliability from train punctuality is observed, highlighting the benefit of incorporating the end-customer perspective in reliability evaluation. It is also found that higher amounts of segments per transportation chain lead to significantly lower reliability.
One of the major causes of dwell time delays is the behaviour of passengers during the alighting and boarding processes, such as an uneven spread of boarding passengers and queues formed in front of the doors. Therefore, changing the behaviour of passengers so that it does not negatively affect the time needed for alighting and boarding is one of the ways to reduce the risk of dwell time delays. Platform management interventions have the potential to induce such a behavioural change, but the real-world impact is not well studied. To fill this gap, the study presented here investigates the effects of a sticker-based platform intervention using video observations from several hundred trains halting at Lund Central Station. We find that the alighting flow rates slightly increased under intervention conditions, with results suggesting that this is due to fewer overlaps in alighting and boarding passenger flows. No statistically significant effects of the intervention on the spread of boarding passengers were found. Our results suggest that changing the behaviour of passengers is likely to be a slow process, requiring additional efforts such as information provision and ensuring that the halting position of a train accurately reflects the information provided by an intervention.
