SemanticRSM is a semantic reinterpretation of UIC's Rail System Model (RSM) based on direct RDF/OWL modeling.
The project revisits the original RSM conceptual model and expresses it natively using Semantic Web technologies, enabling its use in knowledge graph environments while remaining compatible with previous UML-based versions.
RSM (Rail System Model) is an evolution of an UIC International Railway Standard (IRS) originally released in 2016 under the name RailTopoModel (RTM).
RTM and subsequent RSM versions were defined using UML conceptual modeling, following widely accepted conceptual modeling principles.
The SemanticRSM project revisits these models and represents them directly in RDF/OWL, taking advantage of the expressive capabilities of Semantic Web technologies.
More information about the RSM standard can be found at:
Previous work successfully transformed RSM 1.2 UML models into OWL ontologies using the Ontorail toolset developed by UIC.
However, OWL has different modeling capabilities than UML class diagrams.
When used directly rather than through automated translation, OWL enables models that are:
- more compact
- more expressive
- more suitable for semantic reasoning
SemanticRSM therefore represents a native OWL reinterpretation of RSM, designed to preserve compatibility with previous UML-based models while improving semantic expressiveness.
The project follows several guiding principles.
Data conforming to earlier RTM or RSM versions should be convertible with minimal effort, enabling automated transformation where possible.
The goal is simplification without loss of meaning, taking advantage of OWL expressiveness to remove unnecessary modeling complexity.
The model is divided into small, coherent vocabularies following the principle:
high cohesion, low dependency
This facilitates ontology maintenance and reuse.
Several enhancements have been introduced compared with earlier UML models:
- navigability expressed as a transitive property
- internal navigability within non-linear elements such as stations and yards
- more flexible composition of network elements
Where relevant, SemanticRSM relies on widely adopted ontologies, including:
- SOSA / SSN
- GeoSPARQL
- W3C Time
- ontologies for quantities and units
This improves interoperability with other semantic models.
The model allows path determination under constraints using:
- SPARQL queries
- inference engines
Dedicated algorithms may still be preferable for performance reasons in operational systems.
The design of SemanticRSM is influenced by several practical use cases and initiatives.
Key inputs include:
- the RINF (Register of Infrastructure) use case, with emphasis on micro-level topology and geographic referencing
- System Pillar requirements related to rolling stock typology
- FP5 TRANSF4M-R requirements concerning the description of infrastructure “last mile” and rolling stock defects
- other FPx MOTIONAL initiatives
Relevant concepts may also originate from European railway legislation, such as TAF TSI, or from ongoing EU research projects.
Despite these inputs, the model prioritizes generality and long-term usability, avoiding ad-hoc solutions.
Two demonstration workflows illustrate the generation of SemanticRSM data:
- generating an RDF/Turtle representation from a schematic track plan created in draw.io
- generating SemanticRSM data from OpenStreetMap queries
The demonstrations are available in the Flask folder.
To run the demo server locally: