Description

Problem: 

One of the major challenges faced by rail transport in Germany is punctuality. In 2023, the DB long-distance trains achieved only a 64.0% punctuality rate, which is low and has been declining year by year (75.2% in 2021) [1]. This not only causes inconvenience and frustration among passengers but also results in financial losses for DB due to compensation claims [2]. 
The main contributing factors to delays include staff shortages (due to a tight labor market), aging infrastructure, intensive construction activities, and unstable construction planning processes. Current strategies (since 2023) are mostly human-driven or process-driven, along with a hybrid of human-AI decision support for traffic dispatching [3].
Are these strategies effective? There has been no observed improvement in punctuality. In this context, agent-based modeling could be used to simulate the effectiveness of such strategies (e.g., testing scheduling and maintenance policies) before they are implemented. 

Goal: 

To develop an Agent-Based Model (ABM) as a proof-of-concept to simulate and assess the effectiveness of various strategies, aimed at improving the punctuality of rail transport on a selected subset of the German rail network. The simulation will focus on modeling the interactions between different agents in the system, such as trains, dispatchers, control centers, and infrastructure, under various conditions (e.g., track closures, staff shortages, and maintenance schedules).
The ABM will enable simulation and evaluation of different operational strategies. For example, they can test how implementing a decentralized dispatcher decision-making strategy, compared to centralized control, affects the punctuality of trains. Currently they are using AI and human decisions for traffic dispatching, and the model can help evaluate how changes in decision structures may improve or reduce punctuality.
The student will define the specific scope of the model and research focus, such as departures and arrivals at Karlsruhe Hauptbahnhof and its connecting routes, leveraging publicly available datasets. Relevant open data sources include Deutsche Bahn's Open Data portal (https://data.deutschebahn.com/opendata) , which provides timetable information, delay statistics, and operational data, as well as aggregated historical punctuality data available through Deutsche Bahn Statistics (https://piebro.github.io/deutsche-bahn-statistics/questions/ ). 
The final scope will be determined based on data availability, computational feasibility, and the specific operational strategies selected for evaluation.

Required Skills and Knowledge: 

-    Familiarity with agent-based modeling and simulation 
-    Tools: Mesa (Python)/ NetLogo/ AnyLogic
-    Data: Synthetic/ Real Data 

Related papers for your reference: 

-    https://ieeexplore.ieee.org/document/10807430  
(Decentralized agent-based model using Belief-Desire-Intention (BDI) agents to simulate proactive and reactive decision-making for railway traffic management in dynamic and uncertain environments)
-    https://scientiairanica.sharif.edu/article_4186_03778e9fe17a0c928b79adedb1bbe229.pdf 
(a simulation-based optimization model to reschedule trains during major disruptions, showing better performance than human or standard software solutions)

References: 

[1] “▶ Punctuality: On-Time Performance. Deutsche Bahn 2023 | Deutsche Bahn Annual Report 2023.” Accessed: Apr. 17, 2025. [Online]. Available: https://ibir.deutschebahn.com/2023/en/combined-management-report/product-quality-and-digitalization/the-customer-is-at-the-center-of-our-actions/punctuality/
[2] “Deutsche Bahn pays €197M in compensation for delays in 2024 – DW – 03/09/2025,” dw.com. Accessed: Apr. 17, 2025. [Online]. Available: https://www.dw.com/en/deutsche-bahn-pays-197m-in-compensation-for-delays-in-2024/a-71872282
[3] “Digitalization | Deutsche Bahn Interim Report 2024.” Accessed: Apr. 17, 2025. [Online]. Available: https://zbir.deutschebahn.com/2024/en/interim-group-management-report-unaudited/product-quality-and-digitalization/digitalization/