ONTOLOGY-BASED RAILWAY INFRASTRUCTURE

VERIFICATION

Planning Beneﬁts

Michael Lodemann and Norbert Luttenberger

Dept. of Computer Science, Communication Systems Research Group, CAU of Kiel

Christian-Albrechts-Platz 4, Kiel, Germany

Keywords:

Ontology, Owl, SWRL, Veriﬁcation, Railway domain.

Abstract:

Planning new railway infrastructures is a complex process. We present an approach where the formalization

of expert knowledge regarding the railway domain is motivated in order to improve the planning process. By

applying ontologies as a representation of railway related knowledge we are able to make the coherencies of

infrastructural elements explicit. Furthermore the integration of an ontology-based rule language provides the

possibility of a semi-automated integrity veriﬁcation of static infrastructure and safety components. Seman-

tical inconsistencies potentially leading to unsafe conditions regarding train operations can be spotted within

this veriﬁcation process. This combination of conceptualization and correlation rules tends to be applicable

for the creation of a formal and consistent model of speciﬁc railway infrastructures which are to be planned.

1 INTRODUCTION

While planning new railroad lines a lot of issues re-

garding the railway domain have to be addressed. The

deployment and alignment of physical elements such

as tracks, signals and so forth as well as their coheren-

cies with abstract elements like train routes have to

be considered during the planning process in which

many parties are involved. On the one side there

are federal ofﬁces commissioned with approvals, ver-

iﬁcation and certiﬁcation issues. On the other side

there are manufacturers producing physical parts of

the railroad line. Interlocking manufacturers consti-

tute an elevated status within the overall planning pro-

cess. They produce interlocking blocks as important

components of every railway station. An Interlock-

ing block a central unit, where all train movements

and security elements e.g. signals and switches are

controlled. A station inspector uses the interlocking

block to assign a speciﬁc track sequence to a spe-

ciﬁc train, so that sequence can exclusively be used

by this train. Such a so called route is set up temporar-

ily and blocks the corresponding tracks for exclusive

usage. Security elements are used for the blocking

procedure. For example switches nearby the route

are moved into an opposite position in order to pro-

vide ﬂank protection to the train occupying the route.

These security operations are all to be done in order

to prevent other trains from moving into the route and

causing an accident. In short words: The correct ap-

plication of signaling and safety technologies cooper-

ating with the interlocking block is an important task

of the railway infrastructure planning process in order

to provide safe train operations. An importance re-

sides in the correct and formalized planning process

of the static infrastructure in general and the safety

components in detail.

1.1 Motivation

Due to traditionally long cycles of innovation regard-

ing the railway domain the planning process of new

railway infrastructures is only supported by a min-

imum amount of computerization, automation and

software tool support. Especially the railway infras-

tructure itself is not represented as a formalized com-

putational model which can be shared and communi-

cated among involved companies and authorities and

most important: which can be used for computer-

aided veriﬁcation tasks. Almost all agreements of

technical kind are exchanged as paper based docu-

ments. This fact implies an administrative overhead

and a low integrability with respect to reusability.

Apart from these issues, the planning company is

highly dependent on the correctness of the so far de-

veloped railway infrastructure. The company has to

176

Lodemann M. and Luttenberger N..

ONTOLOGY-BASED RAILWAY INFRASTRUCTURE VERIFICATION - Planning Beneﬁts.

DOI: 10.5220/0003071801760181

In Proceedings of the International Conference on Knowledge Management and Information Sharing (KMIS-2010), pages 176-181

ISBN: 978-989-8425-30-0

 2010 SCITEPRESS (Science and Technology Publications, Lda.)

ensure that the current work represents a consistent in-

frastructure regarding integrity and safety. The infras-

tructure veriﬁcation is not realized within the plan-

ning company, but at the EBA

in a manual manner.

Nowadays for veriﬁcation tasks the planning docu-

ments have to be transformed into a speciﬁc table-

based document format. Afterwards they have to be

printed and sent to the EBA, where the infrastructure

plans are veriﬁed against the legal guideline. After the

manual veriﬁcation process an error report is written

and all documents are sent back to the planning com-

pany. Now the veriﬁed documents have to be digi-

talized again, error checked and transformed into the

company speciﬁc format. This veriﬁcation process is

highly inefﬁcient.

The motivation of our work is to develop a possi-

bility for a semi-automated veriﬁcation process of the

static railway infrastructure with focus on the safety

components. We are achieving this by the devel-

opment of an ontological framework where implicit

expert knowledge of the railway domain as well as

the legal German guideline for railway infrastructures

and their correlations are made explicit and stored in

a conceptual model in order. It is obvious that the

German guideline for railway infrastructure planning

is a huge and complex collection of rules and stan-

dards. Our work does not claim to represent the com-

plete guideline. It can be seen as a proof-of-concept

to encode such a guideline into an ontological model

in order to provide a method for veriﬁcation issues.

In general, the choice to use an ontological model re-

sides in the fact that ontologies are achieving inter-

operability between multiple representations of real-

ity [...] and between such representations and reality,

namely human users and their perception of reality.

(Hepp, 2007)

2 RELATED WORK

There are several approaches for infrastructural infor-

mation representation in a domain speciﬁc ontologi-

cal knowledge base. In (M

etral et al., 2007) a method-

ology for modeling urban infrastructures within an

ontology is introduced. The work is mainly focused

on deﬁning an ontological model as a communication

contract between the parties involved in urban plan-

ning processes.

InteGRail is a topic related project founded by

the EU and has the ambition to homogenize the in-

formation retrieval and management within the rail-

way world in order to enable optimization of decision

Federal Railway Authority: german: Eisenbahnbunde-

samt.

making for improved performance on railway tracks.

It uses an ontological model named ”Railway Do-

main Ontology” to solve the integration challenge of

the railway environment. This model is applied al-

lowing a predictive maintenance strategy in railway

components (Shingler et al., 2008) in comparison to

common maintenance strategies like RCM (Reliabil-

ity Centered Maintainance). Another similar publi-

cation of Cristina de Ambrosi (Ambrosi et al., 2009)

discusses the integration of an embedded ontological

system into railway vehicles. This system performs

fault classiﬁcations within the vehicle. InteGRail is a

good example for modeling railway related concepts

using ontologies, but in contrast to our work in Inte-

GRail the model is used as a data structure contract

while we focus on the semantic veriﬁcation of rail-

way infrastructures.

Regarding the formalization of legal railway

guidelines there are several works dealing with au-

tomated ontology extraction (Amato et al., 2008) and

the representation of legislation in ontologies (Boer

and Boer, 2003), but in contrast to them our work

intends to make use of the ontology rule language

SWRL for the representation of legal railway guide-

lines.

3 APPROACH

The railway infrastructure is modeled within an on-

tology by using the ontology description language

OWL (W3C, 2004a). The correlation rules are

phrased in the Semantic Web Rule Language (SWRL)

(W3C, 2004b) syntax. SWRL rules follow the horn

clause syntax, therefore they consist of antecedent-

consequent pairs, and are stored as OWL individuals

within the ontological model to which they are ap-

plied to. During the modeling process we enriched

SWRL with so called built-ins which provide railway

speciﬁc enhancements to the language while main-

taining consistency and decidability.

We favor this ontological representation of the

railway infrastructure planning data because of its ex-

pressiveness. It has the ability to represent the seman-

tics of the railway domain in a more detailed way in

comparison to a simple, syntax-only XML represen-

tation. Apart from semantical correlations expressed

within the OWL-model, the integration of complex

SWRL rules allows the creation of a more realistic

model which as far as possible is in accordance with

railway directive semantics and which can be used for

the planning data veriﬁcation task.

We followed a bottom up approach in formalizing

railway domain knowledge into an ontological model.

ONTOLOGY-BASED RAILWAY INFRASTRUCTURE VERIFICATION - Planning Benefits

177

As a transfer format between the infrastructure plan-

ning tools and the ontological model we make use of

an open source XML-schema named railML (Nash

et al., 2004), which offers a structured description

format for railway related content. We extended the

railML schema in order to cover a majority of infras-

tructural elements and their attributes as they are de-

ﬁned in legal German directives for railway planning.

Then we adapted the concepts and correlations of

railML and created an initial conceptual OWL model.

During fruitful discussions this model is extended and

enriched by profound knowledge of railway experts.

3.1 railML - an XML Schema for the

Railway Domain

The railML XML schema has been developed since

2001 by the railML initiative, intending to establish

railML as an open-source standard format for data

exchange between diverse railway applications. This

includes tools for route simulation and disposition of

trains as well as timetable systems and applications

for railway infrastructure planning.

RailML consists of three subschemata covering

different aspects of the railway domain. There is

a subschema rolling stock in which elements and

attributes regarding train constellations and wagon

conﬁgurations are accumulated. A second is the

timetable subschema which is developed in order to

support the modeling of departure and arrival times

and the whole scheduling of train movements. At last

there is an infrastructure subschema. We focused on

this subschema of railML in which the representation

of physical elements concerning the railway route re-

sides. The element deﬁnitions range from tracks, sig-

nals and level crossings to security related parts like

train detection circuits or transponders for train de-

celeration and many more. Thus the infrastructure

subschema offers a good starting point for the under-

standing of infrastructural concepts and their correla-

tions. During our work we extended this subschema

in order to be capable to formalize the description of

interlocking elements as well.

3.2 Evolving a Formal Railway

Infrastructure Domain Model

We translated the hierarchical railway representation

of the railML infrastructure subschema into an en-

hanced ontological, meshed representation where the

hierarchical structure is broken open. RailML has

limitations thus is not comprehensive enough in or-

der to represent all infrastructural elements and their

attributes requested by , directives. In the ontological

model most of the original concepts and relations of

railML remain, but are augmented with sophisticated

interrelations tending to represent the legal directives.

As a concrete example we can have a look at the

concept signal in railML. A signal can be of the types

main-signal or distant-signal among others. In the

German railway infrastructure guidelines the correla-

tions between main and distant signals are deﬁned in

a very detailed way. A distant signal, for instance, is

always placed before a main signal in order to notify

the engine driver what he has to be aware of. The

correct distances between the signals depend on the

maximum speed on the track as well as on the visibil-

ity of the signals, for example when they are placed in

bends. These directives cannot be encoded and veri-

ﬁed using an XML schema, whereas the expressive-

ness of OWL and SWRL as ontology modeling lan-

guages permit the representation of such directives.

Our starting point was the key concept track,

where all other physical elements are aligned with.

Tracks as well as switches have a spatial extent and

can be interconnected among each other. All other

physical elements like signals, balises, train detection

elements and so forth are point-shaped and related to

tracks or switches. There do exist virtual elements

like routes and railway control centers, but these el-

ements are an overlay of physical elements and sub-

sume their spatial extent.

At ﬁgure 1 an excerpt of the class ontology is

shown. Classes are represented as ellipses. Indi-

viduals as red squares. Whereas datatype properties

are represented as named connections from classes to

blue squares showing their data type. At last object

properties are represented as named connections be-

tween classes. Base is the class where all other el-

ements are derived from. It contains the data type

properties id and name which provide a unique iden-

tiﬁer and a human readable name for all instantiated

objects. Direct subclasses of Base are the classes Re-

lationalObject, DirectedPointObject and VirtualOb-

ject. RelationalObject is a base class for all phys-

ical elements which have a spatial extend namely

Tracks and Switches. DirectedPointObjects are uni-

dimensional aligned along the RelationalObjects via

the object property isOnRO and its corresponding in-

verse property hasDPO. As an example the Signal

class is shown at ﬁgure 1. It features a relation to

the class SignalType which basically consists of an

enumeration of possible signal types as deﬁned in-

dividuals. In extracts the classiﬁcation classes for

the veriﬁcation tasks are shown. E.g. after a veri-

ﬁcation process the class Correct S Placement con-

tains all signals which are correctly positioned within

the bounds of their corresponding track elements re-

KMIS 2010 - International Conference on Knowledge Management and Information Sharing

178

Figure 1: Class ontology (excerpt).

garding the ﬁrst SWRL rule described in the follow-

ing chapter 3.3. VirtualObjects are all non-physical

objects like Route and InterlockingBlock. These ob-

jects are mapped to an accumulation of RelationalOb-

jects. Route elements, as it can be seen on the ﬁgure,

have a direct relation to RelationalObjects and Direct-

edPointObjects via the object properties ﬂankProtec-

tion. With this construct route deﬁnitions are aug-

mented with an additional security concept which had

to be formalized while modeling railway infrastruc-

ture related context. In reality a route is secured by

ﬂank protection elements in order to ensure no other

train can cross the route. A ﬂank protection element

can be for example a signal signaling stop or a switch

which is locked in a position not leading into the

route.

3.3 Reﬁne the Model with Semantic

Rules

OWL allows the deﬁnition of restrictions on classes.

E.g. a common signal containing attributes for a spe-

cialized signal (e.g. entry signal) can be classiﬁed as

such by deﬁning the necessary restrictions and ap-

plying the OWL reasoner. These restriction deﬁni-

tions are a powerful tool within OWL, but they have

limitations. Complex correlations among different at-

tributes and classes cannot be expressed within these

restrictions. We decided to enhance our model not

only with restrictions but also with semantic rules

written in SWRL and stored directly within the on-

tological model. SWRL rules can be deﬁned in horn

clause syntax as antecedent-consequent pairs. This al-

lows the creation of a realistic model which is tends to

be in accordance with railway directive semantics and

which appears to be applicable for veriﬁcation tasks

of railway infrastructure planning data.

As a rule example, the following directive shall

be considered: A signal needs to be placed within the

range of the corresponding track it is assigned to. The

SWRL rule deﬁned in an abstract syntax is as follows:

Signal(?s) ˆ Track(?t) ˆ to(?t, ?to) ˆ

signalIsOnTrack(?s, ?t) ˆ from(?t, ?from) ˆ

SignalPosition(?s, ?pos) ˆ

swrlb:greaterThanOrEqual(?pos, ?from) ˆ

swrlb:lessThanOrEqual(?pos, ?to)

-> Correct_S_Placement(?s)

Rule consequents (results) can be integrated into

antecedents of other rules, thus allowing a hierar-

chization of rules and a reduction of complexity.

Thereby rules can make use of signals which are al-

ready assigned to the Correct S Placement class only,

hence incorrectly placed signals will be omitted. As

shown in the example, SWRL allows the deﬁnition of

ﬁner constraints and the expression of complex corre-

lations among infrastructural elements. With SWRL

an extensive formalization of German legal guidelines

is possible.

As a speciﬁc German legal guideline directive the

following sentence shall be focused: A station area

must be secured by entry signals. This directive can

be formalized with following SWRL rule (simpliﬁed):

ONTOLOGY-BASED RAILWAY INFRASTRUCTURE VERIFICATION - Planning Benefits

179

Track(?t1) ˆ Track(?t2) ˆ

hasConnection(?t1, ?t2) ˆ

hasTrackType(?t1, ’open’) ˆ

hasTrackType(?t2, ’station’) ˆ

CorrectPlacedSignal(?s) ˆ

hasSignalType(?s, ’entry’) ˆ

signalIsOnTrack(?s, ?t1)

-> CorrectStationSecurity(?t2)

The rule is interpreted as follows: ”If there are two

tracks which are connected to each other and between

which there is the (imaginary) border of a station, the

track abutting to the station area has to have a signal

which has to be an entry signal in order to secure the

station area.” Note that in the context of this rule not

only a signal is used but a Correct S Placement. This

shows the hierarchization mechanism and implies that

only correctly placed signals are considers to be part

of this rule.

The power of SWRL and especially the prot

(Stanford-University, 2009) implementation can be

shown with the following rule which represents the

veriﬁcation of the overlap dimension of train routes.

An overlap is a track segment behind a train route

which has to be blocked because of security reasons.

If a train is not able to stop at the end of a train route in

time it runs into the overlap. It is a legal directive that

the dimension of an overlap is (among other factors)

dependent on the maximum speed at which a train is

allowed to move on the speciﬁc route. To formalize

such a directive, mathematical operations are required

within the corresponding SWRL rule. Such opera-

tions can be performed by using the prot

e built-in

eval.

Route(?r) ˆ Overlap(?o) ˆ

hasOverlap(?r, ?o) ˆ vMax(?r, ?v) ˆ

overlapLenght(?o, ?l) ˆ

eval(?result, "v * 3.6 * 2", ?v) ˆ

greaterThanOrEqual(?l, result) ˆ

-> CorrectOverlapPerRoute(?r)

In natural language this (simpliﬁed) rule can be

phrased as follows: ”For every route which has a

vMax in km/h and which has an overlap of a length in

meters this length must be at least the distance a train

can move at vMax during two seconds. Only when

this conditions hold, a route is classiﬁed as a Correc-

tOverlapPerRoute”. The eval method is very power-

ful. A lot of mathematical expressions like sine, co-

sine, abs, ﬂoor, sqrt and many more can be used. The

range of SWRL expressiveness is sufﬁcient for the

purposes of the formalization of German legal guide-

lines for railway infrastructure planning.

4 MODEL PARTITIONING

Figure 2: Ontological veriﬁcation system.

We segmented our veriﬁcation system into different

logical and physical parts. Figure 2 shows the com-

ponents of the veriﬁcation system. At the bottom re-

sides the class ontology which contains the railway

concepts taken from the railML schema augmented

by knowledge taken from railway experts. Apart

from that the class ontology contains categorization

classes. These classes are used to mark veriﬁed ele-

ments like ”[In]CorrectlyPlacedSignals” or ”[In]Cor-

rectStationSecurities” etc.

The rule ontology resides at the upper level and in-

cludes the class ontology so that it is able to make use

of its concept deﬁnitions. These concepts are used

within the rules in order to represent more complex

semantic correlations of the classes and properties.

As mentioned before, the rules are mostly formalized

infrastructure planning guidelines representing legal

directives for German infrastructure planning.

At the uppermost level of the knowledge base the

individual ontology is located. It is generated by ap-

plying XSLT scripts to a document in (the extended)

railML format. This document contain the actual

planning data such as precisely deﬁned infrastructure

objects extracted from infrastructure planning tools.

As it can be seen the individual ontology is data spe-

ciﬁc, whereas the class and rule ontologies build the

unmodiﬁable knowledge base. Invoking the veriﬁca-

tion service, the objects within the individual ontol-

ogy are to be veriﬁed automatically against the Ger-

man railway directives modeled in the class and rule

KMIS 2010 - International Conference on Knowledge Management and Information Sharing

180

ontologies.

At the application side of the system the veriﬁca-

tion service is a web service which wraps the ontology

and reasoning framework. It is invoked by a client

program, which communicates with the service via

standard SOAP mechanisms. A user can upload plan-

ning data in the (enhanced) railML format via a client.

The planning data is transformed via an XSLT script

into the individual ontology in OWL language. Using

an import statement the individual ontology includes

the rule and conceptual ontologies. At this point the

veriﬁcation service performs the actual veriﬁcation

using the ontological knowledge base and generates a

correctness report containing the veriﬁcation results.

This report is transfered to the client, again via SOAP,

that displays the veriﬁcation results to the user. The

client also provides an interface for changing as well

as grouping and (de)activating rules. This mechanism

allows the veriﬁcation of only parts of the planning

data for example if they reﬂect an early stage of the

planning process where not all data is available yet.

5 CONCLUSIONS

The application of ontologies for veriﬁcation issues

especially within the railway domain is inconvenient.

Our experiences with the approach described in this

paper seem to be very promising. The separation of

concepts, rules and individuals provides a loose cou-

pling which is common application design for years.

The ascertainment of expert knowledge regarding the

railway domain into a formal environment tends to be

a step towards the standardization of railway infras-

tructure planning. With straightening the heteroge-

neous communication and veriﬁcation processes, de-

velopment periods can be shortened and optimized.

Similarly the formalization of legal German infras-

tructure planning directives seems to be necessary and

tend be achievable by applying SWRL. Although we

only focused on a small excerpt of all guidelines for

infrastructure planning, we gave examples to show

that formalization can be accomplished with complex

issues as well.

REFERENCES

Amato, F., Mazzeo, A., Penta, A., and Picariello, A. (2008).

Building rdf ontologies from semi-structured legal

documents. In Complex, Intelligent and Software In-

tensive Systems, 2008. CISIS 2008. International Con-

ference on, pages 997 –1002.

Ambrosi, C. D., Ghersi, C., and Tacchella, A. (2009). An

ontology-based condition analyzer for fault classiﬁca-

tion on railway vehicles. In IEA/AIE, pages 449–458.

Boer, A. and Boer, A. (2003). Using ontologies for compar-

ing and harmonizing legislation. In In Proceedings of

the International Conference on Artiﬁcial Intelligence

and Law (ICAIL, pages 60–69. UK), ACM Press.

Hepp, M. (2007). Ontologies: State of the art, business po-

tential, and grand challenges. In Martin Hepp, Pieter

De Leenheer, A. d. M. and Sure, Y., editors, Ontology

Management: Semantic Web, Semantic Web Services,

and Business Application, pages 3–22. Springer.

etral, C., Falquet, G., and Vonlanthen, M. (2007). An

ontology-based model for urban planning communi-

cation. In Ontologies for Urban Development, pages

61–72. Springer.

Nash, A., Huerlimann, D., Schuette, J., and Krauss, V.

(2004). Railml - a standard data interface for rail-

road applications. Proc. of the Ninth International

Conference on Computer in Railways (Comprail IX),

15(5):233–240.

Shingler, R., Fadin, G., and Umiliacchi, P. (2008). From

rcm to predictive maintenance: The integrail ap-

proach. In Railway Condition Monitoring, 2008 4th

IET International Conference on, pages 1–5.

Stanford-University (2009). Prot

e - an open source on-

tology editor. http://protege.stanford.edu/.

W3C (2004a). Owl web ontology language reference.

http://www.w3.org/TR/owl-ref/.

W3C (2004b). Swrl: A semantic web rule language com-

bining owl and ruleml. http://www.w3.org/Submissi

on/SWRL/.

ONTOLOGY-BASED RAILWAY INFRASTRUCTURE VERIFICATION - Planning Benefits

181