The challenges of the sensor web have been well documented, and the use of appropriate semantic web technologies promises to o er potential solutions to some of these challenges (for example, how to represent sensor data, integrate it with other data sets, publish it, and reason with the data streams). To date a large amount of work in this area has focused on sensor networks based on\traditional" hardware sensors. In recent years, citizen sensing has became a relatively well-established approach for incorporating humans as sensors within a system. Often facilitated via some mobile platform, citizen sensing may incorporate observational data generated by hardware (e.g. a GPS device) or directly by the human observer. Such human observations can easily be imperfect (e.g. erroneous or fake), and sensor properties that would typically be used to detect and reason about such data, such as measurements of accuracy and sampling rate do not exist. In this paper we discuss our work as part of the Informed Rural Passenger project, in which the passengers themselves are our main source for transport related sensing (such as vehicle occupancy levels, available facilities). We discuss the challenges of incorporating and using such observational data in a real world system, and describe how we are using semantic web technologies, combined with models of provenance to address them.
The challenges of the sensor web have been well documented in, for example, [
The use of semantic web technologies o er potential solutions to some of these challenges. Ontologies, such as the W3C SSN XG ontology1 provide mod
els for sensors, sensor networks, and observations; and linked data [
To date a large amount of work in this area has focused on sensor networks
based on\traditional" hardware sensors. In recent years, citizen sensing [
One further challenge of citizen sensing, is that for observations generated directly by the human observer, sensor properties that would typically be used to detect and reason about imperfect data, for example measurements of accuracy and sampling rate, do not exist. Similar problems exist with observations generated by the mobile phone's hardware: often the mobile APIs provide few details such as data sheets (describing sensor capabilities), settings used for observations, and, in some cases, which sensor generated an observation3.
This lack of information makes it di cult to perform the necessary assessments of observations produced using citizen sensing. Semantic web technologies potentially have a role to play here by, for example, providing additional contextual information for use in assessment processes.
In this paper we describe an example real-world system which combines citizen sensing with semantic web technologies (section 2); discuss some of the challenges faced by this system (section 3); and describe how we are addressing those challenges (section 4). 2
As part of the Informed Rural Passenger (IRP) project4, we are investigating the challenges of developing a trusted passenger information system (PIS) for rural areas. In our system the passengers themselves are our main source of transport related sensing, performed using a mobile app. The app enables passengers to contribute observations about their journey on public transport, including observations generated directly by the phone (e.g. location, presence of Wi-Fi) and by the passenger (e.g. occupancy level, and perceived vehicle temperature).
Wi-Fi, or GPS sensor to determine location, but does not indicate which was used.
Using linked data principles, this data is then integrated with other relevant data sets, and used as the basis of a PIS, which provides passengers with details, including real-time bus locations, delays, and expected arrival times. This therefore gives the potential for any imperfect data passed as input to the system to adversely e ect its outputs, reducing user trust in the system. 3
In developing the IRP PIS, we have identi ed a set of issues, which extend those de ned for the sensor web, and, we believe, require to be addressed by any system which incorporates humans as a source of sensor data, in order to remain trusted by its users. These challenges are raised due to the potential generation of imperfect data by humans, and the lack of information for identifying and reasoning about it.
Challenge 1, is one of the most pressing: the need to characterise and manage constructs not just of data quality, but also of, for example, reliability, reputation, and trustworthiness, which can use the available types of data.
This gives rise to challenge 2: maximising the data available for making those assessments. Here, identifying and integrating the sensor data with appropriate external data sets can help address this challenge. Related to this are: challenge 3, selecting an appropriate model for describing the citizen sensors and their observations, the possible granularity of which is limited by the lack of information about them; and challenge 4, integrating the qualitative observations generated by humans with the machine generated quantitive observations.
In real-time information systems, short response times are vital; however, processes such as data integration and data assessments potentially con ict with this requirement. Further, the additional data generated by these processes adds to the amount that must be stored and processed. This gives rise to challenge 5: designing a system architecture which uses an appropriate combination of technologies (e.g. for storing and reasoning about the data), which enable the system to perform well while maintaining an acceptable response time.
Finally, challenge 6 relates to ensuring user privacy, especially when sensitive data such as location is being collected and used as the basis of information passed to other users and/or services. Addressing this challenge is made more di cult by the integration with other data sets, which potentially provide additional data which can be used to violate a user's privacy. 4
Within the IRP project we are addressing the above challenges by, rstly exploring the data available within the application domain, and secondly investigating how it can be integrated to form an information ecosystem supporting a range of services which perform PIS functions and data assessments. Whilst the solutions below are outlined within the context of IRP, we believe they are generalisable to other systems incorporating humans as sensors.
The information ecosystem that we are developing to address challenge 2 integrates the passenger observations with various other types of data, including user pro les, social networks, and various types of transport information, such as: operator timetables; NaPTAN and NPTG5 datasets6; road maps; and details of roadworks and travel disruptions7, which themselves link to NPTG. We will also reverse geocode location points obtained from passengers and, where possible, link them to the nearest road/railway line and nearest settlement in the NPTG.
Fig. 1 outlines how we are addressing challenges 3 and 4, by basing our model of citizen sensors and observations on the SSN XG ontology, with domain extensions describing: mobile phones as sensor platforms, people as sensors, and the observations generated via our mobile app (Trips represent timetabled journeys).
Within the ecosystem, we also maintain a provenance record of the data.
Provenance has previously been identi ed as essential in supporting reliability,
discovery, trust [
ssn:Platform
ssn:SensingDevice rdfs:subClassOf
rdfs:subClassOf iPhone
LocationSD ssn:Sensing opmv:Process rdfs:subClassOf opmv:controlledBy some LocationSensing ssn:Observation opmv:Artifact ssn:Sensor foaf:Agent opmv:Agent rdfs:subClassOf AgentSensor iPhoneLocationSD ssn:onPlatform only rdfs:subClassOf ssn:observedBy ssn:sensingMethodUsed only opmv:wasGeneratedBy only rdfs:subClassOf ssn:observedBy only ssn:observes ssn:observesssn:observeLdoPcraotpieorntyObservsastnio:FneatureOfInterest
ssn:featureOfInterest ssn:featureOfInterest ss"nL:oPcraotipoenr"ty ssn:isPropertyOf rdfsT:sruipbFCOlaIssOf stsrinp:isPropertyOf
OccupancyLevelObs ssn:observedProperty
ssn:Property "OccupancyLevel" Trip
One reason for maintaining this information within the ecosystem is to support various types of data assessment, particularly of IQ and trustworthiness.
details of all UK settlements and roads connected to the public transport network.
IQ assessments of data typically analyse various dimensions of the data, and so
the additional information should be bene cial; for example, other members of
our research team are currently investigating the role of provenance in IQ
assessments of linked sensor data [
We will incorporate any data assessments and their results within the ecosystem as part of the provenance record (as subclasses of OPMV Process and Artifact classes respectively). This will allow services/applications (including those making new assessments) to make use of these assessments if appropriate.
The nature of the IRP project requires that it handles large quantities of data and still functions reliably in real time. To help support this and address challenge 5, passenger contributed observations are currently stored in a database, and exposed as linked data using the D2R server9. This setup takes advantage of the strengths of databases (such as scaling to large data sets, and handling multiple concurrent read, update, and delete operations). However, the disadvantage is that it does not exploit many of the advantages of semantic web technologies, such as the ontology based querying and reasoning. 5
In this paper we have identi ed a set of challenges, which we believe, require to be addressed by any system that incorporates humans as a source of sensor data. We propose the use of semantic web technologies to help address these challenges, and illustrate their use in the development of a real-time PIS for rural areas.
We currently have three strands of future work addressing challenges 1, 5, and 6: developing a trust model for the ecosystem; evaluating the performance of di erent options for storing and reasoning about streaming linked sensor data, to determine if a combination can be found that provides (some of) the advantages of semantic web technologies without negatively impacting overall performance; and investigating how we can ensure user privacy.
Acknowledgements The research described here is supported by the award made by the RCUK Digital Economy programme to the dot.rural Digital Economy Hub; award reference: EP/G066051/1