=Paper=
{{Paper
|id=Vol-2548/paper-09
|storemode=property
|title=Decentralized Route Planning Across the Web of Data
|pdfUrl=https://ceur-ws.org/Vol-2548/paper-09.pdf
|volume=Vol-2548
|authors=Julian Rojas
}}
==Decentralized Route Planning Across the Web of Data==
https://ceur-ws.org/Vol-2548/paper-09.pdf
Decentralized Route Planning across the Web of Data
Abstract. Wheelchair users looking for accessible public transport routes,
tourists discovering attractive routes to go around a new city, or bicycle users
trying to avoid highly polluted routes are some examples where highly
individualized route planning is needed. Current route planning applications lack
query flexibility. The types of queries supported by a route planner are only
determined at design time and heavily depend on centralized pre-selected data
sources. Integrating a new data source such as another transport mode, a different
road network or wheelchair accessibility is not straightforward as it generally
requires human intervention to extend the subjacent data model and route
planning algorithm implementation. I investigate how relevant data sources
available on the Web can be dynamically reused for answering custom queries,
and thus allow creating more flexible and personalized route planning
applications. Semantic Web and Linked Data technologies provide a common
framework for data integration. Yet it is still unclear how relevant data can be
automatically reused while remaining independent from specific route planning
algorithm implementations. Preliminary work (i) tests the feasibility of solving
route planning queries over live and static public transport data sources on the
Web, (ii) explores the trade-offs of different Web APIs for publishing and
consuming live data streams on the Web and (iii) introduces a Linked Data based
approach for publishing road networks data.
1. Relevance
Computer-based route planning is nowadays used by millions of people to obtain
journey directions [1]. Public transport providers around the world offer route
planning services to keep their users better informed. Public authorities also recognize
route planning as a mechanism to improve the mobility conditions of their cities and
regions, by enabling citizens and visitors to move around more efficiently [2, 3, 4], as
it occurs in the case of Portland1, London2, Antwerp3, among others. In recent years,
multimodal route planning has received special attention [5, 6, 7]. Applications like
Google Maps4 or Citymapper5, allow users to plan their journeys using multiple
transport modes, e.g. bus, metro, bicycle or walking. They focus on incorporating
mobility related data about different transport modes into their applications to offer
more diverse route alternatives.
Meanwhile, millions of datasets have been published on the Web [8]. Following
Open Data and Smart City initiatives, national and regional governments, private
companies and others, publish heterogeneous data from multiple domains. Many of
these datasets contain information relevant for route planning purposes, e.g. parking
alternatives, road works or street occupancy; which opens the door for creating richer
applications capable of addressing more diverse or specific needs. Considering
accessibility constraints is a very typical scenario for route planning. Enabling users
to specify for example, if they are on a wheelchair [9] or if they are visually
impaired [10], are features that are getting included in existing route planners.
Copyright © 2019 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
Different types of data available on the Web provide virtually unlimited
possibilities for extending route planners: Environmental data (calculate a route that
avoids biking through highly polluted areas), meteorological data (calculate a route
that limits walking outside under bad weather conditions), security-related data
(calculate a route that avoid areas with high criminality indexes) or even general
knowledge bases such as DBPedia [11] or Wikidata [12] (calculate a route that pass
by the historic monuments of a city) are some examples of data that could be used to
create enhanced route planners.
2. Problem Statement
Currently available route planning applications are built following a centralized
data strategy. A route planner6 calculating routes using both the bus and metro public
transport services in a certain city:
1. Collects datasets of each transport mode, containing the minimum necessary
information for route planning, i.e. station locations and vehicle schedules.
2. Integrates the data following a predefined (usually graph-based) data model in
a centralized data store.
3. Calculates available routes using a route planning algorithm tailored to run
over the predefined data model.
The route planner service offered by Google Maps for example, follows this
approach. By encouraging public transport operators around the world to publish their
data using the GTFS [13] format, Google collects these datasets and integrates them
into their route planning application.
However, centralized approaches incur in high costs in terms of computational
infrastructure. Hosting, integrating and keeping up to date the data of every transport
alternative that wants to be supported, and effectively processing route planning
queries from a scaling number of users, requires big data storages and powerful
processors, specially if aiming on providing a global route planning solution.
Furthermore, they also limit the types of queries supported due to the difficulties that
are faced when dealing with heterogeneous data. As long as data are homogeneous in
terms of structure and format, i.e. dealing only with public transport data using a
common format as GTFS, integrating new data sources into a route planning
application can be easily automated and the same route planning algorithm
implementation will remain useful without requiring many adaptations. Other types of
data make necessary to manually adapt the data model and the route planning
algorithm that runs on top of it. This could explain why route planning applications
such as Google Maps or Open Trip Planner do not allow querying for routes that
consider, for example air quality, even though this type of data can be found published
on the Web as Open Data7.
Problem Statement - Integrating new heterogeneous data sources to extend route
planners capabilities requires adapting the subjacent data models and route planning
algorithm implementations, resulting in ad-hoc and hence not reusable solutions.
3. Related Work
Including new data sources to allow more diverse and specific queries in route
planning applications can be generalized as a data integration problem. The Semantic
Web [14] and Linked Data [15] technologies provide a common environment where
data is given a well-defined meaning, better allowing machines to comprehend and
work with heterogeneous data from different sources. Moreover, the Web offers a
common platform for data sharing where different data publishing strategies may
facilitate and influence a decentralized approach for route planning applications.
In this section I present an overview of different aspects related to the problem of
dynamically integrating new data sources to enable more diverse and specific queries
in route planning applications, such as data publishing and route planning algorithms.
3.1. Data Publishing on the Web
Millions of datasets are published on the Web [8]. Following the right strategy
when publishing data is key to increase the interoperability and reusability of data,
and foster the creation of new and innovative services. In this direction, Tim Berners-
Lee introduced a set of principles8 for publishing data as Linked Data on the Web.
Extending this principles, the W3C published a document of best practices for
publishing Linked Data [16], which provides a set of guidelines and design principles
aimed at data publishers.
Traditionally, data is often published on the Web either as a data dump or through
an API. In the public transport domain, GTFS [13], which is regarded as the de-facto
standard for describing and exchanging transit schedules, is a common example of
data dump publishing. The European Committee for Standardization released the
NeTEx standard [17] as a general purpose format for exchanging public transport
schedules and related data. NeTEx is the standard selected by the European Union,
under the Directive 2010/40/EU9, for the provision of an EU-wide multimodal travel
information service, where every member state will publish their public transport-
related datasets through a National Access Point.
Transport related data is also offered on the Web as route planning APIs.
Navitia.io10, Plannerstack11, CityMapper or the open source Open Trip Planner are
some examples. Geospatial data is also fundamental for route planning applications.
Different initiatives exist for publishing geospatial data on the Web. For instance,
OpenStreetMap [18](OSM) stands as a community-driven rich source of freely
available spatial data. LinkedGeoData [19] publishes a spatial knowledge base
derived from OSM that uses RDF as its data model. Also GeoNames [20] publishes
over 25 million geographical names using stable URIs and a semantically annotated
vocabulary12.
On the one hand, data dumps offer full query flexibility over the data but are
expensive for clients who need to deal with integration of complete datasets. On the
other hand, APIs offer simpler access to data, but are expensive for data publishers
and limit query flexibility for clients. The Linked Data Fragments (LDF)
framework [21] explores the different trade-offs in terms of query flexibility and
computational costs of in between Web interfaces determined by fragmenting datasets
and publishing them on the Web. Based on the LDF concept, the Linked Connections
specification [22] was introduced as a light-weight data interface for publishing public
transit schedules that allows to perform route planning on the client-side. It uses the
Linked Connections Ontology13 and the Linked GTFS vocabulary14 to describe
vehicle departures in public transport networks [23].
Data summarization techniques are also used when publishing data on the Web to
facilitate its access and speed up querying processes. Taelman et al. introduced
Multidimensional Interfaces [24] as mechanism to semantically describe ordinal
ranges within datasets, e.g. time-based or geospatial-based ranges, which are inherent
to the nature of commonly used route planning related datasets, such as public
transport network topologies (geospatial range) and their schedules (time range).
Graph-based summarization techniques such as Contraction Hierarchies [25], K-
means [26] or connectivity-based clustering [27] are also commonly used in the route
planning domain to reduce the amount of data that needs to be processed when
solving route planning queries.
Hypermedia plays also an important role when publishing data on the Web.
Providing declarative descriptions about how data can be consumed, helps clients to
understand the means of accessing data sources and increases their interoperability.
The Hydra vocabulary15 aims on providing a common terminology for describing
hypermedia-driven Web APIs to create generic API clients. Hypermedia is also used
to semantically describe responses of Web APIs in Tealman et al. [28]. A shape-based
approach using SHACL [29], is used to define a parameterized structure of API
responses.
3.2. Route Planning
Route Planning has been extensively studied throughout the years. Multiple
algorithms were proposed to calculate routes over different types of networks (e.g.
road, public transport, multimodal, etc), and using different criteria (e.g. shortest path,
travel time, number of transfers, etc). Bast et al. [30] and Pajor [31] present a
comparative analysis of multiple route planning algorithms for road, public transit and
multimodal networks. Most algorithms use subjacent graph-based data models to
represent the different transportation networks over which they operate, and are
defined as extensions of Dijkstra’s algorithm [32]. In recent years, different
approaches that disregard Dijkstra-like graph algorithms, have been introduced for
public transport and multimodal networks. RAPTOR [33], CSA [34], Transfer
Patters [35], Trip-based routing [36] are among the approaches that exploit the basic
elements of public transport networks to calculate routes directly on the timetables.
Different solutions that use some of the aforementioned route planning algorithms
are available. For instance Open Trip Planner16 provides an open source
implementation of Dijkstra-based and RAPTOR algorithms for planning routes over
public transport and road networks. OsmAnd17 is also an open source route planner
for road networks that implements a Dijkstra-based algorithm on top of OSM data.
Being open source allows these solutions to be extended to include new different
types of data sources for route planning. However, extending them requires
substantial effort to adapt the algorithms and the data models over which they operate.
There is no consensus about the criteria that route planning algorithms should
support. The parameters supported by route planning algorithms change from one
approach to another depending on the data model, the algorithm implementation and
use-case. Kelly et al. [6] presents a list of data requirements for ideal multimodal
route planners. It identifies 22 different parameters from traditional data sources such
as GTFS feeds and OSM. The NeTEx standard defines also a Trip Plan Query
Model18 that considers among others, accessibility, via and trip fares parameters. A
extensive analysis of formal languages using regular expressions to define constrained
shortest path queries is presented in Barret et al. [37]. Furthermore, the Web of Data
opens the door for new possibilities and use-cases that need to be supported by route
planning algorithms. For instance, the Multi-Criteria RAPTOR algorithm supports
route calculation with an arbitrary number of input parameters. However, additional
criteria has a significant impact on the algorithm performance, causing it to become
too slow for practical use [33].
Given that Linked Data is also described using a graph-based model, the problem
of finding routes between nodes have been previously studied. SPARQL 1.1
introduced the Property Paths (PP) syntax that allows query engines to test for path
existence between nodes. Savenkov et al. [38] propose an extension for PP syntax to
compute the top-k shortest paths of a given path expression between two nodes. An
extension of the query semantics of PP is also presented in Hartig et al. [39] to
support navigation of PP over the Web of Data instead of over single RDF graphs. A
more expressive alternative is later introduced as the Linked Data Query Language
(LDQL) [40] that allows expressing query graph patterns and navigation paths
independently. Another approach is introduced in Rutgers et al. [41] as an extension
of the property graph query language CypherQL19, which allows for top-k shortest
paths queries, calculated path weights and filtering on path edges and nodes. Lastly,
multimodal route plans are calculated using a GraphQL interfaces published by the
Finish20 and Norwegian21 Transit Agencies.
4. Research Questions
Given the related work, the need for a structured and declarative approach to
express route planning queries, that allows identifying specific data requirements to
resolve them, becomes apparent. A publishing strategy of heterogeneous data on the
Web, that allows for decentralized data access and efficiently solving route planning
queries using state of the art algorithms is also a need. I thus investigate the following
research questions:
Research Question 1 - How can route planning queries be expressed in a use-case
independent and declarative way that allows identifying data requirements for
resolving given queries?
Research Question 2 - How do different ways of publishing linked open data on the
Web affect decentralized route planning query evaluation in terms of performance?
5. Hypotheses
The first research question is related to the need of having a declarative and use-
case independent mechanism to express route planning queries, that allows including
arbitrary constraints, while being self-describing about the type of data needed to
answer a query. Existing solutions are usually tightly coupled to the subjacent data
model and algorithm implementation, limiting the type of queries that can be
expressed. Also, implementation independent approaches such as regular expression-
based syntaxes do not allow for automated identification of data requirements.
Therefore, I define the following Hypothesis related to Research Question 1:
Hypothesis 1 - A semantic model for expressing route planning queries is at least as
expressive as state of the art approaches and allows for complete and automatic
identification of query solving data requirements.
The second research question relates to the problem of how data should be
published on the Web to allow for efficient decentralized route planning solutions.
Current Linked Data based approaches for publishing data on the Web allow for
interoperable data integration. Also, data summarization techniques provide data
publishing alternatives for clients to access just the necessary data to solve a certain
query, accelerating query evaluation times. Considering this, I define the following
Hypothesis related to Research Question 2:
Hypothesis 2 - A data publishing strategy based on data summarization techniques
allows for decentralized route planning query evaluation to be as performant as state
of the art centralized approaches such as Open Trip Planner.
6. Approach
My approach consists of three steps.
1. Semantic modeling of route planning queries - A flexible semantic model to
describe route planning queries that considers different constraints and
parameters for route calculations. By using domain vocabularies in a structured
way, automatic data requirements identification for a given query becomes
possible.
2. Hypermedia-driven Linked Data summaries - Definition of general
requirements to create and publish data summaries on the Web, based on spatial
and temporal dimensions and including declarative hypermedia controls. By
interpreting the hypermedia controls, route planning applications are able to
integrate subsets of relevant data sources on the fly.
7. Evaluation Plan
For evaluating my approach I propose the following plan:
1. Hypothesis 1 - An evaluation that (i) identifies available approaches in the
literature to express route planning queries and the different types of queries
supported by each approach. Then (ii) a qualitative comparison between the
proposed semantic model and the literature to verify that the proposed approach
is at least as expressive as the state of the art.
2. Hypothesis 2 - A performance evaluation in terms of response time and
answer completeness will compare different data summarization approaches
through an multimodal route planning use-case and compare its results to a
centralized solution such as Open Trip Planner.
8. Preliminary Results22
Preliminary work related to Hypotheses 2 and 3, has tested the feasibility of
performing decentralized route planning over public transport networks including live
data. Based on the non Dijkstra-based algorithm CSA [34] and a data publishing
strategy extended from the proposal given in [22], clients are able to execute the route
planning algorithm on the client-side and use hypermedia controls to discover and
integrate the data during query execution.
Other preliminary work, related to Hypothesis 2, evaluates computational cost and
query performance for push and pull-based Web interfaces for publishing data
streams. Also introduce a modular architecture for publishing data summaries based
on the concepts introduced in [24]. Initial results show that push based approaches
perform better for high frequency data streams. Also related to Hypothesis 2, a Linked
Data based a approach for publishing road networks is introduced. OSM road network
data is published in tiles allowing clients to dynamically fetch the correct tiles to
calculate a route.
9. Reflections
Route planning applications are often showcased as a success story for the Open
Data movement. By publishing public transport schedules as Open Data on the Web, a
wave of innovation was triggered, leading to the creation of applications used by
millions. Today the Web offers a much more diverse world of data with virtually
unlimited possibilities. Bringing the power of the Semantic Web and Linked Data to
route planning could open the door for the next innovation wave. Automating data
integration in route planners sets the bases for building more flexible and personalized
applications, able to consider and learn from the needs of their users to provide more
opportune and significant answers.
References
1. Delling, D., Dibbelt, J., Pajor, T.: Fast and Exact Public Transit Routing with
Restricted Pareto Sets. In: 2019 Proceedings of the Twenty-First Workshop on
Algorithm Engineering and Experiments (ALENEX). pp. 54–65
2. Rode, P., Hoffmann, C., Kandt, J., Graff, A., Smith, D.: Towards new urban
mobility: the case of London and Berlin. LSE Cities - The London School of
Economics and Political Science, http://eprints.lse.ac.uk/id/eprint/68875 (2015).
3. Berry, R.: London Assembly - Future transport: How is London responding to
technological innovation? Mayor of London - Transport Commitee, https://
www.london.gov.uk/sites/default/files/future_transport_report_-_final.pdf (2018).
4. Ahlers, D., Akerkar, R., Krogstie, J., Opdahl, A.L., Tessem, B., Zhang, W.:
Harnessing Mobility Data in Cities: A Case Study From the Bergen Region.
Presented at the (2018).
5. Dibbelt, J., Pajor, T., Wagner, D.: User-Constrained Multimodal Route Planning.
J. Exp. Algorithmics. 19, 1–19 (2015).
6. Kelly, T., Miller, R., Datz, A., Hillsman, E., Barbeau, S.: Enabling Cost-Effective
Multimodal Trip Planners through Open Transit Data. National Center for Transit
Research, https://www.nctr.usf.edu/wp-content/uploads/2011/06/77926.pdf
(2011).
7. Kujala, R., Weckström, C., Mladenović, M.N., Saramäki, J.: Travel times and
transfers in public transport: Comprehensive accessibility analysis based on
Pareto-optimal journeys. Computers, Environment and Urban Systems. 67, 41–54
(2018).
8. Noy, N., Burgess, M., Brickley, D.: Google Dataset Search: Building a search
engine for datasets in an open Web ecosystem. In: 28th Web Conference
(WebConf 2019) (2019).
9. Ding, C., Wald, M., Wills, G.: A Survey of Open Accessibility Data. In:
Proceedings of the 11th Web for All Conference. pp. 37:1–37:4. ACM, New York,
NY, USA (2014).
10. Dornhofer, M., Bischof, W., Krajnc, E.: Comparison of Open Source routing
services with OpenStreetMap Data for blind pedestrians. Presented at the (2014).
11. Auer, S., Bizer, C., Kobilarov, G., Lehmann, J., Cyganiak, R., Ives, Z.: DBpedia:
A Nucleus for a Web of Open Data. In: The Semantic Web. pp. 722–735. Springer
Berlin Heidelberg, Berlin, Heidelberg (2007).
12. Vrandečić, D., Krötzsch, M.: Wikidata: A Free Collaborative Knowledge Base.
Communications of the ACM. 57, 78–85 (2014).
13. General Transit Feed Specification. https://developers.google.com/transit/gtfs/
(2006).
14. Berners-Lee, T., Hendler, J., Lassila, O.: The Semantic Web. Scientific American.
284, 34–43 (2001).
15. Heath, T., Bizer, C.: Linked Data: Evolving the Web into a Global Data Space.
(2011).
16. Hyland, B., Atemezing, G., Villazón-Terrazas, B.: Best Practices for Publishing
Linked Data. World Wide Web Consortium, https://www.w3.org/TR/ld-bp/
(2014).
17. Network and Timetable Exchange. http://netex-cen.eu/ (2014).
18. OpenStreetMap. https://www.openstreetmap.org/ (2004).
19. Auer, S., Lehmann, J., Hellmann, S.: LinkedGeoData: Adding a Spatial
Dimension to the Web of Data. In: The Semantic Web - ISWC 2009. pp. 731–746.
Springer Berlin Heidelberg (2009).
20. GeoNames. https://www.geonames.org/
21. Verborgh, R., Vander Sande, M., Colpaert, P., Coppens, S., Mannens, E., Van de
Walle, R.: Web-Scale Querying through Linked Data Fragments. In: Bizer, C.,
Heath, T., Auer, S., and Berners-Lee, T. (eds.) Proceedings of the 7th Workshop
on Linked Data on the Web (2014).
22. Colpaert, P., Verborgh, R., Mannens, E.: Public Transit Route Planning Through
Lightweight Linked Data Interfaces. In: Cabot, J., De Virgilio, R., and Torlone, R.
(eds.) Web Engineering. pp. 403–411. Springer International Publishing (2017).
23. Colpaert, P.: Publishing Transport Data for Maximum Reuse. https://
phd.pietercolpaert.be (2017).
24. Taelman, R., Colpaert, P., Verborgh, R., Mannens, E.: Multidimensional Interfaces
for Selecting Data within Ordinal Ranges. In: Hartig, O., Sequeda, J., and Hogan,
A. (eds.) Proceedings of the 7th International Workshop on Consuming Linked
Data (2016).
25. Geisberger, R., Sanders, P., Schultes, D., Vetter, C.: Exact Routing in Large Road
Networks Using Contraction Hierarchies. Transportation Science. 46, 388–404
(2012).
26. Hartigan, J.A., Wong, M.A.: Algorithm AS 136: A K-Means Clustering
Algorithm. Journal of the Royal Statistical Society. Series C (Applied Statistics).
28, 100–108 (1979).
27. Flinsenberg, I., Van Der Horst, M., Lukkien, J., Verriet, J.: Creating graph
partitions for fast optimum route planning. WSEAS. 3, 569–574 (2004).
28. Taelman, R., Verborgh, R.: Declaratively Describing Responses of Hypermedia-
Driven Web APIs. In: Proceedings of the 9th International Conference on
Knowledge Capture (2017).
29. Knublauch, H., Dimitris, K.: Shapes Constraint Language (SHACL). https://
www.w3.org/TR/2017/REC-shacl-20170720/ (2017).
30. Bast, H., Delling, D., Goldberg, A.V., Müller-Hannemann, M., Pajor, T., Sanders,
P., Wagner, D., Werneck, R.F.: Route Planning in Transportation Networks.
CoRR. abs/1504.05140, (2015).
31. Pajor, T.: Algorithm Engineering for Realistic Journey Planning in Transportation
Networks. https://d-nb.info/1058165240/34 (2013).
32. Dijkstra, E.W.: A note on two problems in connexion with graphs. Numerische
Mathematik. 1, 269–271 (1959).
33. Delling, D., Dibbelt, J., Pajor, T.: Fast and Exact Public Transit Routing with
Restricted Pareto Sets. In: Proceedings of the Twenty-First Workshop on
Algorithm Engineering and Experiments, ALENEX 2019, San Diego, CA, USA,
January 7-8, 2019. pp. 54–65 (2019).
34. Dibbelt, J., Pajor, T., Strasser, B., Wagner, D.: Connection Scan Algorithm. J.
Exp. Algorithmics. 23, 1.7:1–1.7:56 (2018).
35. Bast, H., Hertel, M., Storandt, S.: Scalable Transfer Patterns. In: 2016
Proceedings of the Eighteenth Workshop on Algorithm Engineering and
Experiments (ALENEX). pp. 15–29
36. Witt, S.: Trip-Based Public Transit Routing Using Condensed Search Trees. In:
ATMOS (2016).
37. Barrett, C., Jacob, R., Marathe, M.: Formal-Language-Constrained Path
Problems. SIAM Journal on Computing. 30, 809–837 (2000).
38. Savenkov, V., Mehmood, Q., Umbrich, J., Polleres, A.: Counting to K or How
SPARQL1.1 Property Paths Can Be Extended to Top-k Path Queries. In:
Proceedings of the 13th International Conference on Semantic Systems. pp. 97–
103. ACM, New York, NY, USA (2017).
39. Hartig, O., Pirrò, G.: A Context-Based Semantics for SPARQL Property Paths
Over the Web. In: The Semantic Web. Latest Advances and New Domains. pp.
71–87. Springer International Publishing, Cham (2015).
40. Hartig, O., Pérez, J.: LDQL: A query language for the Web of Linked Data.
Journal of Web Semantics. 41, 9–29 (2016).
41. Rutgers, P., Martella, C., Voulgaris, S., Boncz, P.: Powerful and Efficient Bulk
Shortest-path Queries: Cypher Language Extension & Giraph Implementation. In:
Proceedings of the Fourth International Workshop on Graph Data Management
Experiences and Systems. pp. 6:1–6:7. ACM, New York, NY, USA (2016).
1 https://trimet.org/#/planner
2 https://tfl.gov.uk/plan-a-journey/
3 https://www.slimnaarantwerpen.be/en/home
4 http://maps.google.com/landing/transit/index.html
5 https://citymapper.com/
6 See Open Trip Planner’s case, a widely used open source route planning application:
https://github.com/opentripplanner/OpenTripPlanner/blob/dev-
1.x/docs/Configuration.md
7 Latest measurements from Europe’s air quality monitoring network: https://
www.eea.europa.eu/data-and-maps/explore-interactive-maps/up-to-date-air-quality-
data
8 https://www.w3.org/DesignIssues/LinkedData.html
9 http://data.europa.eu/eli/dir/2010/40/oj
10 http://navitia.io/
11 http://www.plannerstack.org/
12 http://www.geonames.org/ontology/ontology_v3.1.rdf
13 http://semweb.mmlab.be/ns/linkedconnections#
14 http://vocab.gtfs.org/gtfs.ttl#
15 https://www.hydra-cg.com/spec/latest/core/
16 https://github.com/opentripplanner/OpenTripPlanner
17 https://github.com/osmandapp/Osmand
18 http://www.netex-cen.eu/model/conceptual/passenger_info/index.htm
19 https://neo4j.com/developer/cypher-query-language/
20 https://digitransit.fi/en/developers/apis/1-routing-api/itinerary-planning/
21 https://api.entur.io/journey-planner/v2/ide/
22 To comply with the double-blind review process, references to publications on
related preliminary work are not provided.