The German Operations Research Society working group for logistics and transportation (GOR AG LuV) invites you to attend our workshop this year.
AG LuV Workshop 2026: 05.-06.03 at DXC Technology in Düsseldorf
Thank you to DXC Technology for hosting this year’s GOR AG LuV workshop at their offices in Düsseldorf!
Directions to the workshop location
The workshop will take place in at DXC Technology, Speditionsstraße 13, 40221 Düsseldorf, Germany.
Attending
Registration is open! Please register here!. Note that the workshop is free for members of the GOR e.V.. If you are not a member, please sign up before registering.
Workshop language
This year we kindly request that all talks are given in English so as to not exclude anyone from the talks.
Preliminary Program
Note that the following program is subject to change. All speakers are requested to leave 5 - 8 minutes for questions and discussion.
| Day | Start | End | Title | Presenter(s) | Affiliation |
|---|---|---|---|---|---|
| Thursday | 13:45 | 14:00 | Arrive | ||
| 05.03.2025 | 14:00 | 14:10 | Welcome and Introduction from DXC Technology | TBA | DXC Technology |
| 14:10 | 14:40 | Strategic Railway Maintenance Planning | Hanno Schülldorf | Deutsche Bahn AG | |
| 14:40 | 15:10 | A Parameterized Algorithm for Real-Time Train Dispatching | Luka Stärk | ZIB / FU Berlin | |
| 15:10 | 15:40 | Coffee break | |||
| 15:40 | 16:10 | Agile Modernization of Mission‑Critical Railway Systems at SBB | Nicki Skujat | DXC Technology | |
| 16:10 | 16:40 | News from the tram simulation project: Converging the undirected infrastructure graph in Openstreetmap into a routable directed graph | Jörn Schönberger | TU Dresden | |
| 16:40 | 17:10 | Tariff Design and Real-Time Acceptance Control in On-Demand Public Transport | Pirmin Fontaine | Catholic University of Eichstätt-Ingolstadt | |
| 17:10 | 17:20 | Elections: AG Logistik und Verkehr | |||
| 18:30 | Group dinner (Location TBA) | ||||
| Friday | 8:45 | 9:00 | Arrive | ||
| 06.03.2025 | 9:00 | 9:30 | Title TBA | TBA | DXC Technology |
| 9:30 | 10:00 | Weather Routing in Maritime Operations: Bridging Theory and Practice | Daniel Wetzel | University of Vienna | |
| 10:00 | 10:30 | A state-expanded network formulation for Multiple-Depot Integrated Vehicle and Crew Scheduling | Jonas Brenker | Paderborn University | |
| 10:30 | 10:45 | Group picture | |||
| 10:45 | 11:00 | Coffee break | |||
| 11:00 | 11:30 | The Truck-Drone Hurdle Relay Problem in a Euclidean Space | Christin Münch | University Duisburg-Essen | |
| 11:30 | 12:00 | Logical Mapping of Physical Production and Material Flow Networks | Wilmjakob Herlyn | University of Magdeburg | |
| 12:00 | 12:30 | VRPAgent: LLM-Driven Discovery of Heuristic Operators for Vehicle Routing Problems | Kevin Tierney | University of Vienna | |
| 12:30 | 12:35 | Farewell and thanks | |||
| Group lunch |
Presentation abstracts
``Agile Modernization of Mission‑Critical Railway Systems at SBB’’ - From Order‑to‑Cash at SBB Cargo to Wagon Management in Marshalling Yards at SBB Infra
**Nicki Skujat
SBB Cargo has successfully modernized and significantly simplified an IT landscape that had grown increasingly complex over decades. Over a period of five years, a new cloud‑based IT core solution, ORCA (based on DXC-RCMS), was introduced step by step during
ongoing operations. This approach enabled a substantial reduction in operational complexity and IT costs.
In parallel, SBB Infrastructure introduced WaVe (based on DXC-RCMS), a modern wagon management system based on RCMS. With real‑time integration into hump interlocking systems, WaVe supports efficient train separation and formation in SBB marshalling yards.
This talk shares practical insights into agile delivery, modular architecture, and cross‑functional collaboration required to modernize mission‑critical railway systems without disrupting daily operations.
Tariff Design and Real-Time Acceptance Control in On-Demand Public Transport
Pirmin Fontaine (Catholic University of Eichstätt-Ingolstadt)
Demand-responsive public transport systems must balance uncertain demand with limited vehicle capacity. Besides routing and scheduling, two design levers critically affect system performance: tariff structures that influence perceived fairness and demand levels, and real-time acceptance decisions that allocate scarce operational resources.
We consider these two decision layers in complementary studies. First, we evaluate alternative tariff systems with respect to transparency, fairness, and price sensitivity using large-scale survey and choice-based conjoint data. Estimated demand elasticities indicate that tariff design directly shapes request volumes and operational load, particularly for on-demand services. Second, we develop a supervised machine learning framework to predict the operational fit of incoming requests in dynamic dial-a-ride problems. Embedded in an online acceptance policy, the predictor anticipates downstream capacity effects while ensuring real-time feasibility.
Computational experiments with real-world data show that informed acceptance decisions increase served passengers and improve resource utilization. Together, the results highlight fair tariff design and data-driven decision support as key levers for efficient on-demand transport systems.
Logical Mapping of Physical Production and Material Flow Networks
Wilmjakob Herlyn (University of Magdeburg)
The presentation offers a new approach to map production and material flow networks on the mathematical theory of Boolean Algebra. The production and material flow of products is normally not arbitrary but sequential oriented, from one location (source) to another location (sink). Relationships between sources and sinks are node-edge-relations that can be represented as vectors or different kinds of graphs (graphs theory). This way of mapping source-sink-relation is normally applied for routes between physical locations. These models can’t represent hierarchical, alternative, and bonded and branched structure of material flow in manufacturing, transporting, storing, handling, sorting etc. For mapping such complex network, the sequential interval structure of Boolean algebra comprises three fundamental mapping dimensions.
- Sequential-hierarchical dimension
- Sequential-bonded and branched dimension
- Sequential-parallel dimension
The logical network is a virtual representation of the world and must be related to the real physical world and supplemented by an equidistant unit of measure, e.g. the lead-time. On this basis, various methods of arithmetic calculation, simulation, and optimization can be performed.
The Truck-Drone Hurdle Relay Problem in a Euclidean Space
Christin Münch (University Duisburg-Essen)
In this talk, we present the Truck-Drone Hurdle Relay Problem in a Euclidean space (TDHRP), which is particularly relevant in the context of disaster relief. In this problem, a drone is tasked with the delivery of relief supplies in a disaster-stricken area, while being supported by trucks along its way with transportation and battery exchanges. Trucks and drone need to work together, as the partially damaged road network prevents the trucks from reaching the destination, and as the distance from the depot to the destination is far too great for the drone to cover in one flight. We develop a mixed-integer linear program (MILP) for the TDHRP, where we apply a geometric approach that includes trajectory planning for the drone, as opposed to the common assumption in truck-drone delivery that the vehicles operate on a graph. The geometric approach allows the drone to be launched and recovered on any position along the roads, to model the drone velocities as continuous variables, and to ensure collision avoidance. To reduce computation times, we develop an exact row-generation solution method with six different strategies on adding rows. In an extensive computational study with instances with up to ten trucks and 300 obstacles or no-fly zones, two strategies are shown to yield faster median computation times then the complete MILP.
News from the tram simulation project: Converging the undirected infrastructure graph in Openstreetmap into a routable directed graph
Jörn Schönberger (TU Dresden)
News from the tram simulation project: Converging the undirected infrastructure graph in Openstreetmap into a routable directed graph.
Openstreetmap (OSM) hosts a quite elaborated tram infrastructure data-package for most of the cities who have such a public transport mode in operation. On top of this infrastructure, tram services (routes) are defined. It is straightforward to use this “data treasure” in the context of evaluating OR-based algorithms. However, since the tram infrastructure storage concepts follow common OSM concepts, track tracks are represented as a huge collection of undirected edge sequences. In addition, important information is not directly stored in OSM (e.g. which turnout/switches can be used in a route). Finally, the overall quality and completeness of the stored data is unclear. In this presentation, a procedure is presented that reconstructs a routable direct graph from the OSM raw data. It derives missing but important data by analyzing “the data stored around”. Finally, it is able to detect obvious data errors but also collects suspicious data (to be handed over to a human-made data inspection and data correction). The abilities and limitations is demonstrated by reconstructing tram network from several European cities.
Strategic Railway Maintenance Planning
Hanno Schülldorf (Deutsche Bahn AG)
We present a strategic planning problem for railway infrastructure and discus different modelling aspects. We propose a MIP model and (very first) computational results.
A Parameterized Algorithm for Real-Time Train Dispatching
Luka Stärk (ZIB / FU Berlin)
We present a parameterized depth-first-search branch-and-bound algorithm developed for the DISPLIB 2025 competition on real-time train dispatching. Solving the problem requires making routing and ordering decisions for trains to minimize delays. We model the problem using a generalized disjunctive graph, where nodes represent train operations and arcs encode precedence constraints.
The key contribution is the integration of train path selection into conflict resolution, leading to a branching degree of at most 4. Specifically, a conflict can be resolved by deciding precedence or by removing the involved operations of one train. Hence, we show a parameterized running time of O(4^k poly(n)), where k is the number of conflicts, and n is the number of operations.
For this work, we adapt and integrate several algorithmic components: (1) conflict detection using interval graphs, and grouping adjacent conflicts with mutually implied precedence decisions, (2) incremental longest path computation of operation start times, and (3) memory-augmented node selection informed by previous iterations, which additionally provides lower bounds for branching cuts.
To quickly construct feasible solutions and upper bounds, we introduce train-by-train and rolling-horizon heuristics that reuse the branch-and-bound algorithm.
Experiments demonstrate the effectiveness of our approach: we find feasible solutions for all 112 competition instances and prove optimality for 50 of them within 1 minute.
A state-expanded network formulation for Multiple-Depot Integrated Vehicle and Crew Scheduling
Jonas Brenker (Paderborn University)
Joint work with Guido Schryen (Paderborn Universit)
In public transport systems, both vehicle deployment and crew assignment constitute major planning challenges for urban, metropolitan, and intercity services. The integrated vehicle and crew scheduling problem (VCSP) addresses both challenges simultaneously. Given a set of timetabled trips, it seeks to determine minimum-cost schedules for vehicles and crews that are feasible and compatible with each other. As both vehicle scheduling in the multiple-depot case and crew scheduling are NP-hard, existing research mainly relies on heuristic solution methods, while exact approaches have thus far been limited to smaller instances. To address this gap, we propose a novel exact reformulation of the multiple-depot VCSP in which crew duties are constructed as paths in a state-expanded network, with the goal of reducing overall run time, in particular for large-scale instances. We solve the resulting mixed-integer programming model using a standard solver and conduct experiments on well-known benchmark instances from the literature to compare solution quality and run time against state-of-the-art exact and heuristic methods.
VRPAgent: LLM-Driven Discovery of Heuristic Operators for Vehicle Routing Problems
Kevin Tierney (University of Vienna)
Designing high-performing heuristics for vehicle routing problems (VRPs) is a complex task that requires both intuition and deep domain knowledge. Large language model (LLM)-based code generation has recently shown promise across many domains, but it still falls short of producing heuristics that rival those crafted by human experts. In this paper, we propose VRPAgent, a framework that integrates LLM-generated components into a metaheuristic and refines them through a novel genetic search. By using the LLM to generate problem-specific operators, embedded within a generic metaheuristic framework, VRPAgent keeps tasks manageable, guarantees correctness, and still enables the discovery of novel and powerful strategies. Across multiple problems, including the capacitated VRP, the VRP with time windows, and the prize-collecting VRP, our method discovers heuristic operators that outperform handcrafted methods and recent learning-based approaches while requiring only a single CPU core. To our knowledge, \VRPAgent is the first LLM-based paradigm to advance the state-of-the-art in VRPs, highlighting a promising future for automated heuristics discovery.
Weather Routing in Maritime Operations: Bridging Theory and Practice
Daniel Wetzel (University of Vienna)
This talk revisits the work of Stefan Kuhlemann and Kevin Tierney on a genetic algorithm for computing realistic sea routes under dynamic weather conditions and examines how such research can be translated into a practical prototype for real-world maritime operations. We focus on the technical and organizational challenges encountered when adapting an algorithm from research into a deployable tool: handling incomplete or uncertain data, integrating with existing ship and fleet systems, ensuring computational efficiency at scale, and aligning model assumptions with operational constraints such as safety margins, regulatory requirements, and commercial priorities. The presentation highlights which assumptions hold in practice, which must be revised, and how differing operational areas and future customer needs influence design decisions. By tracing the path from theory to practice, we aim to provide insights into turning advanced optimization methods into usable decision-support systems.
built using Jekyll and GitHub Pages
GOR e.V. logo copyright GOR e.V., DXC logo copyright DXC Technologies, content: cc-by-sa Kevin Tierney 2025 (get source code). Last build date: 2026-02-20.
