Optimized personnel deployment of conducters in local transport by pooling (partner: DB Regio AG)
In the German local public transport, state-owned operation authorities define, design, and tender every transport network separately. The tender documents describe precise network-specific requirements such as a meticulous timetable or minimum attendance rates for conductors in the trains. In this project, we focus on the latter: cost-effective crew scheduling as one key to railway companies’ competitive position and operational profit.
Different local transport networks are often interconnected. More precisely, the intersection of two sets of operated train stations by network A and B is nonempty in many cases. Yet, at DB Regio AG, our project partner, crew schedules are optimized for each network separately. The aim of the project is to explore whether crew pooling, which describes a joint crew schedule and personnel deployment across multiple networks in a defined region, can achieve additional cost savings. We expect that crew pooling reduces the total paid working time (total number of shifts x average paid working time) and generates economies of scale (personnel fix costs, cost for depots and others).
The complexity of the railway crew scheduling problem (RCSP) increases exponentially with its size; in mathematical terms, the problem is known to be NP-hard. Solving multiple networks simultaneously imposes not only an increase in size, but also adds network-specific restrictions to the problem’s complexity. Most of the present optimization methods and algorithms can solve large-scale but only one single network with globally valid restrictions at a time. One example is the SINA project, which develops a software solution for crew scheduling with attendance rates for conductors in one local transport network. Building on the results of SINA, the project at hand widens the scope of the crew scheduling problem with attendance rates as to solve multiple networks simultaneously. The aim of the project is to develop and implement an algorithm which optimizes the large-scale RCSP for multiple networks, which vary in size and must fulfil network-specific restrictions. For this purpose, we use powerful Operation Research methods and heuristics, e.g. column generation.
Partner
Time period
October 2016 - September 2018
Involved Persons
Julia Heil
Prod. Dr. Udo Buscher
Dr. Janis Neufeld
M. Sc. Felix Tamke
Martin Scheffler
Dipl.-Wi.-Ing. Michael Hölscher