On the occasion of the PhD defense of Rolf van Lieshout, ECOPT organizes a workshops on May 12, 2022, at the Erasmus University Rotterdam.

Speakers and program

The speakers in this workshop are:

• Rolf van Lieshout (Technical University Eindhoven)
• Bart Moors (Netherlands Railways) and Bart van Rossum (Erasmus University Rotterdam)
• Rob Goverde (Technical University Delft)
• Layla Martin (Technical University Eindhoven)
• Rowan Hoogervorst (Denmark Technical University)
• Anita Schöbel (Technical University Kaiserslautern)

The program is as follows.

Venue: The event takes place on campus Woudestein of Erasmus University Rotterdam. The morning program is scheduled in the Mandeville Building, Room T3-24. The lunch is served in Restaurant Sienna, in the Tinbergen Building. The afternoon program is in the Theil-Building, Room C2-5. Directions on how to get to the campus and a map of the campus can be found here. 

Registration: Participation in the workshop is free. However, only a limited number of places are available. The regular registration period is closed. For late registration, please send an email to dollevoet@ese.eur.nl.

Rolf van Lieshout: How (not) to Evaluate Passenger Routes, Timetables and Line Plans

Accurate evaluation of the service quality of public transport is imperative for public transport operators, providers of competing mobility services and policy makers. However, there is no consensus on how public transport should be evaluated. We fill this research gap by presenting a structural approach to evaluate three common manifestations of public transport (route sets, timetables and line plans), considering the two predominant route choice models (shortest path routing and logit routing). The measures for service quality that we derive are consistent with the underlying routing models, are easy to interpret, and can be computed efficiently, providing a ready-to-use framework for evaluating public transport. As a byproduct, our analysis reveals multiple managerial insights.

Bart Moors and Bart van Rossum: Operational Railway Crew Planning with Individual Sharing-Sweet-and-Sour Rules: From Strategy to Implementation

In the first talk of this paired presentation, we discuss the necessity and strategic goals of the newly envisioned crew planning process of Netherlands Railways (NS). In the current crew planning process, a highly detailed crew schedule is constructed a year before operation, and this schedule is used to create a crew roster. Due to changes in the timetable and rolling stock schedule, however, the crew schedule and roster are still subject to frequent modifications. This way, it is hard for crew members to schedule their personal lives, and many planning steps are repeated multiple times. Therefore, NS is developing a crew planning process, where a global capacity planning is constructed well in advance, but the exact duty content is determined close to the day of operation. This should lead to more stable and reliable rosters and grant NS more flexibility in replying to demand fluctuations. In addition, the new process allows NS to incorporate individual Sharing-Sweet-and-Sour rules, as compared to the rules on crew-base level that are currently used.

In the second talk, we discuss the development of algorithmic decision-support systems that should enable the implementation of the new process. We present a mathematical formulation of the operational crew planning problem with individual Sharing-Sweet-and-Sour rules. Here, the aim is to construct duties that cover all tasks in the timetable, respect the global capacity planning, and satisfy the individual rules at the end of the planning period. We propose a moving window solution approach that uses a column generation heuristic and feedback mechanism, and evaluate this approach on real-life NS instances covering the period October to November of 2021. We show that our approach attains high satisfaction scores in limited computation times.

Rob Goverde: System performance during railway disruptions

Railway disruptions cause a drop in performance that is theoretically modelled according to a bathtub model in three phases representing the degradation, rescheduled timetable, and recovery. In practice, the performance curves over time have different shapes depending on the conditions and disruption cause. This presentation first discusses the impact of the start and end of a disruption on the three phases based on a timetable rescheduling optimizaton model. Second, real performance curves are presented based on historical traffic and disruption data showing different behavior following different disruption causes. 

Layla Martin: Challenges in Planning and Operating Vehicle Sharing Systems

Vehicle sharing services commonly face the problem of spatially-distributed, unbalanced demand. The number of vehicles rented from a location may not equal the number of vehicles returned to this location. This imbalance influences all of their short- and long-term decisions. On a day-to-day basis, operators must rebalance their fleet, and they must strategically increase their fleet to account for rebalancing, or select a subset of stations to open that reduce the imbalance. During this talk, I will first introduce planning and operating challenges in vehicle sharing, and subsequently deep-dive into different extensions or “complicating factors”: Planning and operating becomes even more complex if we consider aspects from reality such as competition, the ongoing vehicle automation, or budget limitations. For example, considering competition can substantially improve profitability (by up to 40% in a Munich, Germany, based case study). While driverless vehicles are more expensive to procure than conventional, human-driven vehicles, operating them is easier, and thus cheaper. Operators of vehicle sharing systems might thus invest in a mixed fleet consisting of both driverless and human-driven vehicles to increase their profits (by up to 3.9% in a case study using New York City taxi cab data). If the upfront-budget of vehicle sharing systems is limited, they must use accrued profits to eventually increase fleet size and operating area. If operators open the wrong set of stations first, reaching a highly profitable system may take excessively long, or they may even never reach profitability. Further, since vehicle sharing systems promise environmental benefits, but often increase emissions and pollutions, cities may regulate them, e.g., by limiting the operating area, fleet size, or empty mileage, or by imposing a tax. Regulations can have unexpected, unintended, and even counter-intuitive impacts. For example, in a case study, regulating the fleet size reduces empty mileage to a greater extent than directly limiting rebalancing. 

Rowan Hoogervorst: Integrated Rolling Stock and Shunting Driver Rescheduling

In this presentation, we look at the rescheduling of rolling stock and its interaction with the rescheduling of train drivers at the stations. Rolling stock rescheduling focuses on adjusting the assignment of rolling stock to the trips after a disruption occurs. In particular, we want to determine the compositions that train units are in, as they determine the availability of seats for passengers on the trips. These compositions can be changed at stations through shunting actions, which are executed by so-called shunting drivers. Rescheduling the rolling stock thus also requires the rescheduling of the shunting drivers. Traditionally, the rolling stock and shunting drivers are rescheduled sequentially, which can lead to infeasibilities if too many shunting actions are planned compared to the number of available shunting drivers. We thus propose to reschedule the rolling stock and the shunting drivers in an integrated way, for which we suggest both a Benders decomposition and a flow-based approach. We test both approaches on instances of Netherlands Railways (NS).

Anita Schöbel: Line planning for different demand periods

Line planning and frequency setting is a basic planning step in public transportation for which many models and solution approaches exist. The goal of line planning is to offer a good service to the passengers while minimizing the costs for the operator. However, passengers' demand varies within a day: while many passengers want to travel in the morning peak, the demand decreases during the day, and on Sundays or during public holidays demand data is not only smaller than on week-days but also its structure changes significantly. One could accommodate such demand changes by computing a different line concept for each of the given demand periods, but memorizing different line plans with different frequencies within even the same day (lines in the morning peak differ from lines in the afternoon and from lines at night) is not convenient for the passengers. On the other hand, using only one line concept which is able to satisfy the demand during morning peak for the whole day is a waste of resources. Hence, in practice, line concepts differ more or less over the day and the week. In this talk we develop models for finding (possibly different) line concept for each of the demand periods, but under the restriction that these line concepts should be "similar". We propose different definitions for the (dis)similarity of line concepts and add them as constraints, hereby coupling the single line planning problems. We analyze the resulting models and develop solution approaches. In the computational results we demonstrate that maximizing the similarity and minimizing the costs of line concepts are conflicting goals and we show the Pareto front with respect to them.