DB InfraGO Rail Data Challenge

Intelligent Data Transfer from Measurement Trains - Modern measurement trains are a central building block for a high-performing, safe, and future-proof rail infrastructure. They capture the condition of tracks, overhead lines, and other elements of rail infrastructure with high precision – generating very large volumes of image, video, and sensor data in the process.

The challenge is not only collecting this data, but transferring it reliably. Along the route, available network coverage is often limited or nonexistent, while operations generate large data volumes that cannot be transmitted continuously.

With the Rail Data Challenge, we invite external experts and startups, together with DB InfraGO AG, to develop new paths and new systemic approaches for handling this data.

Measurement trains generate data volumes in the terabyte range per trip, including image, video, and sensor data. At the same time, available connectivity along the route is highly variable or temporarily unavailable. Even modern mobile communication technologies are not sufficient to permanently and completely transmit these data volumes during operation.

In the future, selected, particularly time-critical data shall be transmitted during operation via mobile networks. However, this is not part of this challenge and not a mandatory criterion for application.

The majority of measurement data is stored locally on the train. In order to handle this data efficiently, intelligent preprocessing directly at the point of origin is required – for example through structuring, prioritization, and suitable forms of data reduction.

At regular intervals, the measurement train is parked in a depot for several hours or overnight. These idle times shall be used to transfer the stored data in bundled form and as efficiently as possible into downstream systems.

This is exactly where the Rail Data Challenge comes in:

We are looking for concepts and technical approaches for reliably, economically, and transparently transferring very large data volumes from a measurement train to backend systems in a depot setting – taking into account bandwidth, stability, resume capability after interruptions, and realistic operational conditions.

The central question therefore is:

How can an overall system be designed that reliably and efficiently processes and transfers very large volumes of measurement data under realistic operating conditions?

The goal of the Rail Data Challenge is to gain a deeper understanding of

• which system architectures and technological approaches are suitable today for handling very large measurement data volumes,
• which combinations of preprocessing, compression, prioritization, and data transfer are realistically feasible today, and
• which technological developments and perspectives may become relevant for long-term system build-up (e.g., 2030+).

The moving measurement train serves as the reference scenario for the target system. The practical execution of the challenge deliberately takes place in a controlled environment (e.g., depot, simulations, assumptions) in order to enable comparable evaluation of concepts.

We are looking for startups and teams with experience in large data volumes, distributed systems, or challenging connectivity conditions – regardless of industry.

Particularly relevant competencies include:

• Transfer of large data volumes
  (image, video, or sensor data in the GB–TB range)
• Edge computing & edge data intelligence
  (preprocessing, prioritization, and data reduction directly at the system)
• Hybrid connectivity
  (combination of different transmission paths such as mobile networks, Wi-Fi, or stationary offload points)
• Mobile or distributed systems
  (experience with unstable or temporarily unavailable connections)

Rail experience is not required. Partial solutions are explicitly welcome.

The challenge is not a pure upload speed competition.

The focus is on:

• Optimization of data throughput within limited time windows (e.g., during depot dwell times)
• Intelligent system and architecture design
• Handling unstable or dropping connections
• Prioritization of relevant data
• Data integrity and traceability
• Lossless or technically valid data reduction (e.g., compression) with full data restorability
• Realistic assumptions for later operation

Future-oriented approaches are welcome but are not a mandatory evaluation criterion.
 

The DB InfraGO Rail Data Challenge is conducted as a multi-stage selection and working process. The aim is to select, in a structured procedure, the teams that best understand the technical problem and can progressively deepen their approach.

Application Phase

Startups can apply with their solution approach until 22 March 2026.

Feedback on participation in the challenge days will be provided by 01 April 2026 at the latest.

Required Application Documents

Applications should in particular include the following information:

• A short description of the company and team (e.g., year founded, team size, relevant competencies)
• A brief description of the solution or concept approach and problem understanding in the context of the use case
• A clear presentation of the basic system idea (e.g., text or simple sketch)
• An assessment of typical data volumes and boundary conditions the approach can handle (e.g., orders of magnitude such as how many TB the system can transmit in x hours and under which conditions)
• A description of how very large data volumes are generally handled (e.g., prioritization, preprocessing, or data reduction)

Phase I: Virtual Concept Phase (Top 5) – 27–30 April 2026

The challenge starts with a concept phase that takes place exclusively online.

• Virtual onboarding and technical deep dive with DB experts (27 April)
• Independent development of a technical overall concept incl. Q&A session and sparring sessions with DB InfraGO subject-matter experts (28 & 29 April)
• Optional: presentation of long-term technological perspectives (e.g., 2030+)
• Online presentation of concepts to a jury on 30 April

→ Selection of three teams

Phase II: Two-Week Challenge Phase (Top 3) – May (tba)

• Further development of the concepts
• Implementation of a mini-prototype or simulation of key elements
• Focus on system logic, feasibility, and robustness
• Predominantly remote, with occasional joint sessions

At the end of the challenge, teams present their results in person to subject-matter experts. One team will be selected for the subsequent proof of concept.

100-Day Proof of Concept – starting 08 June 2026

The winning team becomes part of the 100-day proof-of-concept program of DB mindbox. In this phase, starting on 08 June 2026, the solution approach will be further developed and tested in a near-real environment together with Deutsche Bahn.

Financial support for participation in the DB InfraGO Rail Data Challenge is based on fixed amounts:

• Each selected team (Top 5) receives a one-time payment of €1,500 for participation in the challenge days.
• Each team (Top 3) that additionally completes the two-week working phase and participates in the result presentation (Selection Day) receives an additional one-time payment of €4,500.

In addition, travel expenses (transportation, hotel accommodation) for in-person appointments within the scope of the challenge can be reimbursed up to €500 per team upon submission of receipts.

The selected team participates in the 100-day proof-of-concept program of DB mindbox, including:

• €25,000 funding (no equity taken)
• Personal support by a DB mindbox startup manager
• Access to experts, specialist departments, and relevant contacts
• Optional use of the DB mindbox coworking space in Berlin

For questions regarding the challenge and the application, we offer regular exchange formats. Technical and organizational questions can be sent to dbmindbox@deutschebahn.com. In addition, we offer an open office hour every Friday from 12:00 to 12:45 CET. You can join here.