=Paper=
{{Paper
|id=None
|storemode=property
|title=Short Paper: Enabling Lightweight Semantic Sensor Networks on Android Devices
|pdfUrl=https://ceur-ws.org/Vol-839/daquin.pdf
|volume=Vol-839
|dblpUrl=https://dblp.org/rec/conf/semweb/dAquinNM11
}}
==Short Paper: Enabling Lightweight Semantic Sensor Networks on Android Devices==
https://ceur-ws.org/Vol-839/daquin.pdf
Short Paper: Enabling Lightweight Semantic
Sensor Networks on Android Devices?
Mathieu d’Aquin, Andriy Nikolov, Enrico Motta
Knowledge Media Institute, The Open University, Milton Keynes, UK
{m.daquin, a.nikolov, e.motta}@open.ac.uk
Abstract. In this short paper, we present an architecture to deploy
lightweight Semantic Sensor Networks easily based on widely available
Android Devices. This approach essentially relies on deploying a SPARQL
endpoint on the device, and federating queries to multiple devices to
build Semantic Sensor Network applications.
1 Introduction
Research into semantic sensor networks has been focusing on the treatment and
processing of data aggregated from large networks of sensors, often based on
specialised equipments geographically distributed in large areas. [1] discusses a
number of challenges related to Semantic Sensor Networks in such scenarios.
The challenge we are particularly interested in relates to the ability for “rapid
development of applications” that make use of Semantic Sensor Network. We
especially look at applications in scenarios where it is needed to set-up networks
of simple sensors quickly and easily (e.g., school projects, small experiments).
In recent years, the Android platform1 became a de-facto standard for dif-
ferent types of mobile devices from several manufacturers. These devices possess
several types of embedded sensors such as a camera, an accelerometer, a GPS
sensor and a microphone. On the other hand, as shown by our previous work [2],
the computational power of these devices already allows efficient processing of
small to medium volumes of semantic data.
In this short paper, we describe a lightweight architecture to create small
scale sensor networks based on Android devices, and an application that makes
use of such a network of Android devices/sensors. To realise this, we adapt
a triple store to be deployed on an Android device, which provides a shared
repository populated through sensor-aware applications on the device. The in-
formation gathered in this shared repository is exposed through an externally
accessible SPARQL endpoint, making it possible to build applications exploiting
data collected from a network of devices, through query federation.
2 Overview
The idea on which this paper is based is very simple: exposing the sensors at-
tached to an Android Device through a SPARQL endpoint and using SPARQL
query federation so that the information gathered through these sensors can be
exploited as the product of a Semantic Sensor Network (see Figure 1).
?
Part of this research has been funded under the EC 7th Framework Programme, in
the context of the SmartProducts project (231204).
1
http://www.android.com/
Fig. 1. Overview of the approach to create Semantic Sensor Networks out of Android
devices.
In this approach, the use of the Semantic Sensor Network ontology is essen-
tial, as it allows to integrate the information coming from various devices and
sensors. The idea here is that applications developed for the Android platform
directly produce information using this ontology from sensors attached to the
device. In this way, no post-processing is need for external applications to employ
this information directly out of the SPARQL endpoint.
3 Implementation
At the core of our approach is the deployment of a triple store on the Android
device, which is shared by the mobile applications populating it, using the Se-
mantic Sensor Network ontology, and by the SPARQL endpoint deployed in the
device. As discussed, in [2], Sesame2 is, amongst the available options, the one
that best fits an environment where only limited resources are available. We
therefore adapted Sesame so that it can be deployable as an Android library.
The Android environment is based on a specialised Java Virtual Machine, and
Sesame being developed entirely in Java, most components of Sesame did not
require any adaptation. Access to files is however different on the Android plat-
form than it is on a usual computer. We therefore extended Sesame so that it
provides a persistent RDF store using the shared, external storage available on
most Android Devices (in SmartPhones, it corresponds to the SD card). In other
terms, a shared persistent repository is installed on the Android device that is
accessible, and can be populated, by applications accessing the device’s sensors.
2
http://www.openrdf.org/
The other element to be included on the Android device is a Web interface
giving access, through the SPARQL protocol, to the content of the shared triple
store. One of the difficulties here is to deploy a Web server on the such as device,
being accessible externally. Luckily, the popular Web application server Jetty3
has been ported to work on the Android Platform in iJetty4 , providing both a
Web server and a servlet environment. We therefore implemented (a simplified
version of) a SPARQL endpoint as a web application relying on our Android-
adapted version of the Sesame API.
Finally, a mechanism is needed for the federation of SPARQL queries over
multiple Android devices. In the cases where the data comes only from isolated
and independent sensors, this federation mechanism can be very simple, as it only
requires concatenating the results obtained from queries to different devices. In
more complex scenarios where information from different sensors can be linked
(such as discussed later), a more sophisticated mechanism is needed. We rely here
on our own implementation of a distributed SPARQL query endpoint based on
federating queries to multiple other SPARQL endpoints [3].
4 Example Application
To illustrate the benefits of the architecture we are proposing for lightweight
Semantic Sensor Networks with Android, we developed a simple application
used to collaboratively “map” a geographical area using pictures (for example,
to give an idea of the views at certain points of a path, in an area that Google
Streetview does not cover, such as a University Campus).
Fig. 2. Extension of the SSN ontology used in our example application.
We develop an application that can take pictures and represent the informa-
tion about the picture and its location as an Observation using the extension
of the Semantic Sensor Network ontology shown in Figure 2. This application
records the path of the picture on the device, the location of the device at the
time of taking it, as well as the time and the identifier of the device representing
the sensor used to make the observation. We then used this application on several
3
http://jetty.codehaus.org/jetty/
4
http://code.google.com/p/i-jetty/
different SmartPhones. Using the simple SPARQL federation method described
above, we implemented a Javascript application that displays the pictures taken
from this network of phones on a map of the covered area (see Figure 3).
Fig. 3. Screenshot of the application mapping pictures taken from a network of Android
phones using federated SPARQL queries.
5 Related Work
An architecture supporting a semantic sensor network commonly includes three
layers: 1- Data layer: sensors producing observations; 2- Middleware layer: ser-
vices responsible for sensor discovery and semantic data integration; 3- Service
layer: services providing integrated data to the end-user applications. Imple-
menting such an infrastructure for a specific use case requires dealing with five
main challenges [1]:
– Choosing an appropriate abstraction level for sensor data representation.
– Adequate characterisation and management of the quality of service, includ-
ing desired levels of data availability, completeness, and response time.
– Realizing integration and fusion of data from heterogeneous sensing devices.
– Identification and location of relevant sensor data sources.
– Supporting rapid development of applications handling integrated sensor
data.
Depending on the requirements of the targeted use case scenarios, these issues
can be addressed to a different extent, and a trade-off between them can be
required. For example, ensuring high quality of service can make the architec-
ture more complex and expensive and complicates deployment and application
development.
Among existing solutions, the SemsorGrid4Env project5 focuses on building
large-scale semantic sensor networks for environmental management, in particu-
lar, for such critical scenarios as fire prevention and flood control. The proposed
solution involves a multi-tier service-oriented architecture [4], which utilises a
range of web services to integrate streaming data coming from sensor networks
with data stored in static repositories. The nature of the scenarios requires the
architecture to pay particular attention to such problems as quality of service,
information latency, and security. The Sense2Web approach [5] uses the linked
data principles to make sensor data publicly available via the Web. It allows the
user to publish semantic descriptions of sensors and link them to other linked
data resources (such as location URIs from DBPedia). The LinkedSensorData
repository [6] implements a wrapper over meteorological data provided by sen-
sors in the O&M XML format, combines the data in a single repository and
makes it accessible with SPARQL queries. Similarly, the SensorMashup plat-
form [7] assumes the existence of sensors producing streaming data and provides
a semantic infrastructure for composing mashups over these data.
These architectures primarily concentrate on large-scale geographically dis-
tributed networks for industrial scenarios. We, on the other hand, look at more
ad-hoc scenarios in which the deployment of a dedicated sensor network can be
too complex and expensive. Other systems have been developed that provide
RDF-based storing and querying of information on Android devices [8, 9]. They
focus on the storage, querying and manipulation of RDF on the device, while
we focus on how exposing information collected on the device through SPARQL
can enable lightweight networks of devices as sensors.
6 Discussion
The architecture presented here is simple and lightweight by nature. It has how-
ever a number of advantages that favour the rapid development of applications
relying on Semantic Sensor Networks, where the sensors are provided by one of
the most available and affordable platforms. We presented an application that
shows how somebody can easily set up a set of devices that act as a sensor net-
work, providing information through a SPARQL interface for the application to
integrate and use.
We can imagine many possible extensions to this application and other ap-
plications where a similar set-up could be considered, focusing on different types
of sensor. For example, we could use the same application in a school trip ded-
icated to bird-watching. Groups of pupils would be given an Android phone
where they could record with a picture seeing a particular type of bird. In this
case, we could extend the application to also use the microphone of the phone to
record the sound of the birds. Thinking about other sensors that could be used
on an Android device, we could set-up a network of devices at different fixed po-
sitions in a building to record the vibrations on the floor of the building during
the day, together with the sound intensity (for example, to find out about the
impact of some building work). We could use a more complete combination of
5
http://www.semsorgrid4env.eu/
sensors (camera, accelerometer, microphone) to check how busy different areas
of a museum are during an opening day, and derive from this information the
flow of visitors going from exhibitions to exhibitions.
In all these examples, the common aspect is that the sensor infrastructure
is simple and lightweight. There is a need for an architecture that can be easily
set-up, is highly re-usable, and is affordable. Also, the need for easy ways to
semantically integrate data is very clear in such scenarios. In our application, we
could annotate the pictures with the buildings they represent, and similarly, with
the species of birds considered in the school trip scenario. Both the vibration
and museum scenarios should be connected to a semantic model of the buildings
being considered and to events happening during the day.
The approach using query federation makes such integrations easier. Indeed,
an annotation service could be set up that provide an interface for users to
annotate the pictures taken from the different phones with buildings or bird
species. This service would not need to aggregate all the data in one place, but
could simply use query federation to access information about the pictures and
expose the annotations through its own SPARQL endpoint, therefore enriching
the network with more information. Similarly, a SPARQL endpoint can be set-up
that delivers data regarding the buildings and events in which a network is set-
up. An interesting element is that such a SPARQL federation approach makes it
easier to realise hierarchical networks: networks made of sub-networks. The use
of the Semantic Sensor Ontology and of a coherent URI scheme would allow in
this case to put together sensor networks that are heterogenous in nature and
infrastructure.
References
1. Corcho, O., Garcia-Castro, R.: Five challenges for the semantic sensor web. Semantic
Web Journal 1(1) (2010)
2. d’Aquin, M., Nikolov, A., Motta, E.: How much semantic data on small devices?
In: EKAW. (2010)
3. Miche, M., Erlenbusch, V., Allocca, C., Nikolov, A., Mascolo, J.E., Golenzer, J.: Fi-
nal concept for storing, distributing, and maintaining proactive knowledge securely.
Technical Report D4.1.3, SmartProducts Consortium (2011)
4. Gray, A.J.G., Garca-Castro, R., Kyzirakos, K., Karpathiotakis, M., Calbimonte,
J.P., Page, K., Sadler, J., Frazer, A., Galpin, I., Fernandes, A., Paton, N., Corcho,
O., Koubarakis, M., De Roure, D., Martinez, K., Gmez-Prez, A.: A semantically
enabled service architecture for mashups over streaming and stored data. In: ESWC.
(2011)
5. Barnaghi, P., Presser, M., Moessner, K.: Publishing linked sensor data. In: Semantic
Sensor Networks, SSN. (2010)
6. Patni, H., Henson, C., Sheth, A.: Linked sensor data. In: Collaborative Technologies
and Systems, CTS. (2010)
7. Le-Phuoc, D., Hauswirth, M.: Linked open data in sensor data mashups. In: Se-
mantic Sensor Networks, SSN. (2009)
8. David, J., Euzenat, J.: Linked data from your pocket: The Android RDFContent-
Provider. In: Demo, ISWC. (2010)
9. Le-Phuoc, D., Parreira, J.X., Reynolds, V., Hauswirth, M.: Rdf on the go: An rdf
storage and query processor for mobile devices. In: Demo, ISWC. (2010)