Open wsw2016 opened 2 months ago
RTM1.1, RSM1.x (consumed by EULYNX DP and, in altered form, by railML(R)3.x) all allow both representations.
But a track level (MICRO level), and by convention, only linear elements are used for the topology graph.
This is to ensure that navigabilities are transitive and can be used to determine possible paths between an origin and a destination.
In the present version (3.1.0) of the ERA ontology, navigabilities are expressed from a linear element to a linear element. So the case you represent is excluded by design.
Note: it is recommended to also instantiate non-navigable relations (e.g. between NE2 and NE3) as RDF/OWL assumes an open world and monotonic reasoning (i.e. not all data may be available, but adding missing data would not disprove previously made inferences). Missing navigability information does not mean "non-navigable", it just means "not documented". In the case of ordinary switches, no problem, we can infer N3 (between NE2 and NE3) to be false. In the case of crossings, this even becomes mandatory, as there are no rules to derive non-navigability from known navigabilities.
As long as the era:from and era:to do not explicitly refer to an end of a LinearElement, we can connect the middle of one LinearElement with an end of another LinearElement, and the position of this linking can be defined via NonLinearElement.
If this would be made impossible by referencing era:from to LinearElementEnd (or so), the reason for the NonLinearElement is unclear: it is not used for the graph definition. It is using the graph and so, can be defined in some other package.
Indeed, the nonlinear element can be dispensed of. In the ERA proposal, it is used as the "carrier" of navigabilities: possible, but not necessary.
The notion of orientation of the linear element seems to have disappeared. This was given by the axiom that a net relation connects the two linear elements head-to-tail, head-to-head etc. In other words, the net relation being attached to the start/end of a linear element.
Did I miss an episode ?
The orientation should not disappear indeed, lest the pathfinding become impossible using topology alone.
New updates of the model diagram here https://linkedvocabs.org/data/era-ontology/3.1.0/doc/resources/era-ontology-3.1.0.jpg. You can also see the Turtle file in the repository.
In the previous version of ERA-vocabulary the topology graph was defined by two classes NetElement and NetRelation. With them we could model topologies, where the NetElements are connected by their ends.
In the new version an additional object is introduced “NonLinearElement”, which can be placed on a LinearElement and link to a connection (era:Navigability). Now the following graph is representable.
Is it by intension? Both, RailML3 and Eulynx dataprep models support only the first kind of the topology graph.