PERCCOM Master Program
Master's Thesis in
PERvasive Computing & COMmunications for sustainable development
Md Anowarul Abedin
MOBILE GLOBAL SENSOR NETWORK FOR ENHANCED SENSOR MANAGEMENT
2016
Supervisors: Doctor Arkady Zaslavsky - CSIRO, Australia
Doctor Karan Mitra - Luleå University of Technology, Sweden Doctor Prem Prakash Jayaraman -RMIT University, Australia Examiners: Doctor Jari Porras - Lappeeranta University of Technology
Doctor Eric Rondeau - University of Lorraine
ment.
Lappeenranta University of Technology School of Business and Management PERCCOM Master Program
Md Anowarul Abedin
Mobile Global Sensor Network for Enhanced Sensor Management Master's Thesis - 2016.
52 pages, 12 gures, 1 table and 4 appendices.
Keywords: Internet of Things, OpenIoT, Data Annotation, XGSN, Big Data Internet of Things introduces many applications and services where sensor data need to be easily accessible. There are a number of open source middleware platforms to collect and publish sensor data. Some of those platforms support data annotation which provides the context of the data and make them meaningful. However, current architectures of those platforms facilitate the annotation of the sensor data to happen in centralized servers. This process requires the sensor data to travel to the cloud- based server in order to be annotated. There are some research works that focus on mobile devices e.g., smart phones, tablets, mobile routers etc. to be able to run an instance of such platforms so that data acquisition becomes easier in a distributed manner. But in these approaches also, data annotation, if supported, happens only after pushing them to the server. Whereas, the ability to annotate sensor data close to the source would improve the ability to understand the data and could play important role in data ltering and provenance. This thesis focuses on annotating the sensor data in the mobile devices. It proposes Mobile Global Sensor Network (MGSN) that can collect data from sensors, annotate them, make them available for other applications and push them to the server for further use. It also can combine on-board sensor data with external sensor data and capture user behavior by annotating the on-board data. Thus MGSN opens up opportunities for context aware applications and can help end-users practice sustainable behavior. It can manage sensor life time by suggesting power saving mode based on the annotated data. Also, data can be ltered in the mobile devices to reduce uplink trac.
All praises to the Almighty Allah, who gave me the strength to continue my work during the dicult time of last six months.
I would like to express my gratitude to my supervisor Arkady Zaslavsky for the condence that he has placed in me and for his guidance during this research work.
It is my honor to accomplish this master thesis under his supervision.
Also, I would like to thank Karan Mitra for his continuous support, guidance and Stigs coee time. Thanks to Prem Prakash Jayaraman as well for his advices and knowledge in the domain of Internet of Things. They are my co-supervisors.
Thanks to Karl Anderson and Robert Brannstrom for their presence, their accessi- bility and their assistance during my thesis work.
Thanks to Eric Rondeau, PERCCOM coordinator, and every member of the PER- CCOM family for the two years of Master studies. Also, a special "Thank You a Lot" to Jari Porras for believing in me and encouraging me when I was almost lost.
At last, I would like to thank my parents, younger brother and my wife, Rud- mila, for their continuous support and inspiration. Family is everything to me and this thesis is for them.
Dhaka, September 15, 2016
Md Anowarul Abedin
CONTENTS
1 INTRODUCTION 9
1.1 Context . . . 9
1.2 Motivation . . . 11
1.3 Research Objective and Challenges . . . 13
1.4 Research Methodology . . . 14
1.5 Thesis Contribution . . . 14
1.6 Thesis Outline . . . 14
2 BACKGROUND AND RELATED WORK 16 2.1 IoT - WSN Integration . . . 16
2.2 Semantic Annotation . . . 17
2.3 Mobile Sensing . . . 20
2.4 IoT Platforms . . . 23
2.4.1 FIWARE . . . 23
2.4.2 OpenIoT . . . 24
2.5 Summary . . . 25
3 MGSN for Enhanced Sensor Management 26 3.1 MGSN Details . . . 26
3.2 Components of MGSN . . . 28
3.3 Communication Model of MGSN . . . 29
3.4 Execution Flow . . . 30
3.5 Summary . . . 31
4 IMPLEMENTATION 32 4.1 Installing OpenIoT Middleware . . . 32
4.2 Development of MGSN . . . 35
4.3 Summary . . . 36
5 CONCLUSIONS AND FUTURE WORK 37 5.1 Summary . . . 37
5.2 Limitations . . . 38
5.3 Future Work . . . 38
REFERENCES 38
APPENDICES
Appendix 1: Installation of OpenIoT Middleware Appendix 2: MGSN Source Code
List of Figures
1 An Agricultural Scenario - Smartphones Collecting Sensor Data [10] 10
2 The Three Pillars of Sustainable Development [10] . . . 12
3 A Sample RDF File of a Geo-Positioning Sensor . . . 18
4 A Sample Metadata of an Environmental Sensor . . . 20
5 Proposed MGSN Block Diagram . . . 29
6 MGSN Interaction with OpenIoT . . . 30
7 JSON API of the Sense Smart City Weather Sensor . . . 33
8 Sensor - XGSN Interaction Sequence Diagram (High-level) . . . 33
9 Sensor - XGSN Interaction Process Flow (Detail) . . . 34
A1.1 OpenIoT Installation . . . 45
A1.2 OpenIoT Wrapper File Conguration of X-GSN . . . 45
A1.3 OpenIoT Execution Command . . . 46
A1.4 OpenIoT Query to Show Linked Data from Sensors . . . 46
A1.5 Wrapper Code to Push Sensor Data in OpenIoT from JSON API - 1 47 A1.6 Wrapper Code to Push Sensor Data in OpenIoT from JSON API - 2 48 A1.7 Wrapper Code to Push Sensor Data in OpenIoT from JSON API - 3 49 A1.8 Wrapper Code to Push Sensor Data in OpenIoT from JSON API - 4 50 A1.9 Metadata Information for the JSON API . . . 51
A2.1 MGSN Code to Mitigate the Internal Resource Problem . . . 52
List of Tables
1 Comparison among MOSDEN, Tiny-GSN and Thread Mobile Appli- cation: . . . 22
ABBREVIATIONS AND SYMBOLS
(Alphabetic order)
API Application Programming Interface GPS Global Positioning System
GSN Global Sensor Network IoT Internet of Things
NGO Non-Governmental Organization WSN Wireless Sensor Network
XGSN Extended Global Sensor Network
1 INTRODUCTION
This research focuses on collecting and annotating the sensor data on the mobile devices e.g., smartphones. It proposes Mobile Global Sensor Network (MGSN) that collects data from sensors, annotates them and pushes them directly to internet for further use. Thus MGSN enables semantic annotation in close proximity of data generation.
This chapter introduces the problem domain and the context of this research. It also discusses the motivation of the research both from technological and sustainability aspect. Then it denes the objective, challenges and contribution of this research.
Finally, it draws the outline of the rest of this thesis.
1.1 Context
Wireless sensor networks (WSNs) are being widely deployed in dierent sectors for many applications and services of our everyday life. The implementation of Internet of Things (IoT) requires more sensor deployment to facilitate the introduction of more innovative applications and services [1]. But WSNs in IoT paradigm need some enhanced sensor management abilities to enable such applications [2].
Various middleware platforms have been developed to facilitate WSN data usage in IoT environment. These middleware collect data from WSNs and enable other applications and services to use them. These middleware can be used for introducing dierent sensor management abilities, e.g., sensor life time calculation, controlled data collection from sensors etc. [3].
But in most of the cases, sensor data are not linked together and thus they cannot be used eciently. Thus, recent researches have focused on annotating the sensor data in order to make a linked sensor database to provide more intelligent and context- aware services [4]. Those researches mainly extend the previously built middleware and enable them with linked data support. The data are stored in the server as before, but in addition to that, the metadata are also stored.
This thesis takes OpenIoT as the middleware for its implementation. It aims at enabling sensor data annotation in close proximity of the WSN. Thus it plans to modify one of the components of OpenIoT, X-GSN, to be working in mobile devices
e.g., smartphones [5]. Implementation of such an application can enable a number of new services for mobile users including context-aware applications [6], [7]. It also can open up scopes for additional sensor management tasks, i.e., energy usage of sensors, suggesting power saving mode etc.
One of the use cases could be the agricultural sector. Use of dierent sensors in agricultural elds is necessary to improve food production. Digital agriculture and its dierent applications are addressed in [8], [9]. Temperature, water level, soil moisture, soil salinity, level of oxygen and nitrogen in soil, nutrition level of soil etc.
can be continuously monitored in an agricultural eld if sensors are deployed. These data can help decide emergency action or formulate policy-level decision.
Figure 1. An Agricultural Scenario - Smartphones Collecting Sensor Data [10]
On top of these data, the on-board sensors e.g., GPS, camera etc. can also be helpful. [11] states that GPS and camera are the on-board sensors used mostly in agricultural sector. The pictures of the diseases of the plants taken by the camera can be helpful to diagnose the disease and decide which pesticides to use; GPS info can locate the area and with the help of the external sensor data it can be suggested whether the disease has the possibility to spread to other lands. So, the combination of on-board and external sensor data can be very helpful in this sector and may be interesting to the agricultural specialists.
A smartphone of a farmer, as shown in Figure 1, can collect the sensor data op- portunistically. An application will be running in that mobile phone in order to collect the data, annotate and process them, and send relevant data to the server.
An IEEE 802.15.4 supportable chipset in a smartphone can enable mobile phones to communicate directly with sensors and collect data from them [12]. But if the data could be pre-processed and ltered in the mobile phone based on certain criteria, that would help reduce overall trac ow in the uplink (to the server).
In addition to that, annotation of sensor data can be helpful to compare nearby farmers' collected data and bring out interesting ndings related to sensor manage- ment. Especially, annotation of sensors' power consumption data can help decide dynamically when data collection should be reduced or when data collection should be timed. For example, if a temperature sensor of a land has low energy and sen- sors from adjacent lands are sending temperature data, then data collection from that specic sensor can be stopped, reduced or timed. Thus it opens up scope for dierent enhanced sensor management abilities.
Also, the working pattern or behavior of the farmers can be analyzed based on the annotated data. If the nitrogen level of a particular land is lower than other lands of that same region, it might be the eect of not using certain kind of fertilizers by the farmer of that land. So he may be advised to use the fertilizers or take other measures. Thus, farmers can be suggested more sustainable farming practices.
1.2 Motivation
With the growth of IoT, Big Data is also available with its potentials and challenges.
One way to address the big data challenge is to lter data. Meaningful data could be helpful to understand relevant data from irrelevant data. For this purpose, original data is annotated with metadata to add context around the data. For example, tem- perature data from elds where experiments are not conducted could be discarded at the data ingestion stage. This could reduce the amount of unwanted data stored in the server.
There are a number of implementations which collect the information in mobile devices and send them to a centralized database. But so far, during literature review, no mobile device based implementation has been found that supports data annotation on-board for WSN. This would be a new work and can open up a number
of scopes for future research and implementation. On top of that, annotated data is very important and can bring dierent dimensions for analysis. From technical perspective, this is a good motivation to start with.
After starting the work, and the selection of the framework and middleware, it is observed that semantic sensing at the close proximity of the data generation can help in many other aspects in future, e.g., Big Stream management, mobile device based opportunistic services etc. This is why working in this domain comes up with a lot of excitement.
There is another very motivating factor. As a student of PERCCOM, I have al- ways looked for a research scope that can ensure sustainable development. Having sustainability as one of the primary focuses of our two years Master study, this re- search work gives me an opportunity to work with the three pillars of sustainable development as in Figure 2.
Figure 2. The Three Pillars of Sustainable Development [10]
As my research output has the potential to capture user behavior and manage or suggest sustainable practices to the end users, I can work with all three pillars - environmental (less energy use), economical (less battery use, more food production) and social (farmers get more help through such an app, this research can be the benchmark for measuring how ecient other apps are).
Thus, from both technical and a PERCCOM perspective, I nd my research work to be very much exciting.
1.3 Research Objective and Challenges
The objective of this thesis is to propose a good model in order to enable semantic sensing in mobile devices. Also the implementation of the proposed model is one of the objectives. However, it is to make clear that the full implementation of the model will be part of a bigger project. Rather, the scope of this thesis is limited to only the implementation of the semantic annotation portion of the proposed MGSN model.
On-board semantic annotation is one of the research challenges of this research work. A mobile phone can collect data from outside sensors via 802.15.4 interfaces (if any) [12], [13], [14] or it can collect information from on-board sensors. In any way, the sensor data has to be annotated semantically. How semantic annotation will be implemented in mobile devices is the rst key question for the progress of this research.
Semantic annotation can be done based on a specic ontology for wider meaning, acceptability, use and reuse. However, there are dierent ontologies and a number of approaches for the choice of ontology. This also depends on the underlying mid- dleware or platforms that are used for data management. Which ontology to use in this research is one of the major criteria for the successful completion of the project.
Thus the choice of ontology is another research challenge.
For dening an end-to-end model of the proposed solution - collecting data from on- board or external sensors, annotating the data, using them based on the annotation and sending relevant data to the central database if necessary - these are the two major research challenges. This thesis addresses these two challenges and tries to clarify them in Chapter-3.
1.4 Research Methodology
Design Science research methodology has been followed to carry out this thesis.
This research has a specic problem domain and an artifact is proposed, which is a new solution to this domain. This artifact is expected to provide better output compared to the existing ones. Thus this thesis focuses on proposing a model of the artifact and its implementation. Based on the implementation, necessary discussion and review are necessary to conclude if the implementation is good enough or not or whether the research contribution is strong or not.
1.5 Thesis Contribution
This thesis studies the state-of-the-art techniques for collecting sensor data in mobile devices. It then proposes a mobile application based model that will collect on-board and o-board sensor data and send it to the upstream for further use. It includes on-board data annotation mechanism. However, the mobile application has not been completely developed.
The proposed architecture allows users to send their mobile usage data and other data related to the energy usage of the mobile phones, sensors etc. As a result, these data could be used for enhanced sensor management techniques and eective use of mobile devices. Thus the proposed solution, if implemented fully, can be used for practicing and suggesting sustainable behavior, creating context-aware applications, reducing network trac and saving energy of the sensors and devices .
1.6 Thesis Outline
This section gives an overview of the thesis structure with a brief introduction to the following chapters:
Chapter 2 Background and Related Work
Chapter two presents a literature study of the integration of wireless sensor network and internet of things. Further, semantic annotation and mobile sensing are analyzed in detail, and related works on them are presented. The chapter is concluded with a background on two IoT platforms and their features.
Chapter 3 MGSN for Enhanced Sensor Management
Chapter three introduces the MGSN application. At rst, the app requirements are presented. Then, the dierent components of MGSN are described. Further, the package diagram, the execution ow of MGSN and its interactions with OpenIoT are explained. At last, the semantic annotation part of MGSN is demonstrated.
Chapter 4 Implementation
Chapter four presents the Implementation of MGSN. The implementation is not complete. However, the works that have been done are demonstrated here. The problems that are faced during implementation are also discussed.
Chapter 5 Conclusions and Future Work
Chapter ve summarises the thesis contribution and discusses the limitations of the current implementation. The chapter concludes this Master thesis with future work scopes in the domain of semantic mobile sensing.
2 BACKGROUND AND RELATED WORK
This chapter presents the literature review of this thesis. The rst part discusses Internet of Things (IoT) paradigm in conjunction with cloud and wireless sensor network. The second part discusses semantic annotation in detail. Then, mobile sensing is discussed in relation with dierent components of OpenIoT framework.
At the end, an overview of two of the IoT platforms has been added.
2.1 IoT - WSN Integration
As WSN is a key component of IoT, integration of WSN with traditional cloud or internet is a major step. A lot of research works are focused on the integration of WSN and internet. [5] proposes a middleware based solution, Global Sensor Network (GSN), which can collect data from any sensor from any vendor or network and process it before storing it to a database [15], [16]. [17] extends the functionality of GSN with semantic sensing (X-GSN) from the context of web of things.
[18] provides an integrated platform, OpenIoT, for end to end communication of meaningful sensor data and using that data for other applications. OpenIoT uses XGSN [19], CloUd-based Publish/Subscribe middleware (CUPUS), Linked Stream Middleware (LSM), RDF database based on SSN Ontology and a Manager that manages service delivery and utility [20]. The problem with current OpenIoT ar- chitecture is: sensor data must travel through the entire network via a number of cloud-based middleware, which may be hosted in a distributed manner, to become available to be accessed by other applications.
[21] opens another interesting dimension of WSN. Deploying sensor network is another problem as there is the presence of gateway router or base station or hub, which needs to sink with sensor networks and with the internet infrastructure. Mass level users do not purchase and use their sensors as per their need unlike what they do for computers, mobile phones etc. because of the concerned network deployment problems. As [21] proposes smart phones to collect data from WSN, it is also possible for the mobile phones to operate as a Base Station (BS) / Hub. In that case, there will be no setup issues regarding hub or BS, sensors will communicate directly with the mobile phone using existing WiFi, Bluetooth or 802.15.4 [13], [22].
2.2 Semantic Annotation
Semantics and syntax are related with each other. When syntax is how someone says something, semantics is the meaning behind what he/she says. Two syntactically dierent statements can mean the same thing semantically.
HTML is the current syntax for the web/internet, which is used to display any docu- ment hosted somewhere in the internet. Using that syntax, devices can communicate between each other and exchange the information among them, but they do not un- derstand the meaning behind those information. The internet is not designed to enable the devices to understand the meaning.
Semantic Web is the concept for enabling the devices to know the meaning behind the webpages they are showing us [23], [24], [25]. It helps us introduce many in- novative applications more easily. To make that happen, there are Ontologies and RDFs. Ontology denes the grammar of dierent "terms" and "actions" in a partic- ular domain. RDF is the language to describe the relationship between two objects through annotated metadata, based on the rules dened in that particular ontology.
Devices check the RDF and understand which object is connected with which object by which actions. The objects are expressed as URI (Uniform Resource Identier).
CURI (Compact URI) is also used to express the objects, which is shorter version of long URIs.
As an example: "Abedin reads Hamlet" - this statement represents an RDF. Each RDF statement is called a Triple. Here "Abedin" and "Hamlet" are objects, so they must have URIs. The word "reads" is the action that connects "Abedin" and
"Hamlet". "Abedin" can have relations with other objects through other Triples.
But the URIs for the objects must be unique, i.e., there cannot be one URI for two dierent objects.
How Triples will be written in a particular domain is dened by the ontology of that domain. Ontology is a knowledge base in a particular domain; it is also called the vocabulary. It is a precise denition of terms, actions and reasoning in a subject area.
Triples follow the dened rules from the specic ontology to describe a relationship between two objects (URIs).
For the above mentioned Triple, we can assume that the concerned ontology is
"LibraryOntology". For that Triple to be correct, there must be a denition in the "LibraryOntology", which species two objects with "reads" action. Such a
denition could be: "Person reads Book". Here, "Person" and "Book" are classes and "read" is the action to connect them. In that case, "Abedin" is an object of
"Person" class and "Hamlet" is an object of "Book" class. It is to be mentioned that the "LibraryOntology" can have many such denitions.
In xhtml webpage, the metadata about the above Triple could be written as below:
<body xmlns:library=http://xmlns.com/LibraryOntology/>
<span about="#abedin" rel="library:reads" resource="#hamlet">
Abedin reads Hamlet
</span>
</body>
Thus in the browser, it will display only "Abedin reads Hamlet". But the metadata specied through the Triple will be read by every devices and could be used for further use. Here, "#abedin" is the URI for "Abedin" and "#hamlet" is the URI for "Hamlet". So they must be unique.
Figure 3. A Sample RDF File of a Geo-Positioning Sensor
URLs can be used for URI as well. And that gives the real strength of semantic web. In this example, if "#hamlet" actually refers to a real URL where the web- page about the book Hamlet is located, then some interesting information could be retrieved. If the webpage of Hamlet holds a Triple based on the "LibraryOntology", which states - "Shakespeare wrote Hamlet", where "Shakespeare" and "Hamlet" are objects and "Hamlet" has the same URI as the previously mentioned Triple, then
"Which Shakespeare written book does Abedin read?" - this query can be answered by internet. The devices will check both Triples and nd that Abedin reads Hamlet which is written by Shakespeare, though it is not mentioned directly in any web page. Current internet cannot provide this answer to this query. It looks in the database or the text displayed by the browser and produce search result based on that. But it misses the links of the data and thus cannot gure out the relation.
When annotation is done at the source or close to the source i.e. WSN, it is termed as semantic annotation. It is the ability to annotate incoming data right at the source of the data improving the ability to understand the data. This approach also plays a signicant role in data provenance as the annotated data could be used to backtrack to the data source. Another important aspect is, semantic annotation allows data ltering to be done close to the source which will result in reduced network trac to the server.
Semantic sensing is the concept to enable IoT to be integrated with Semantic Web to form Web of Things (WoT) [26], [27], [28]. Sensor data are semantically annotated.
It will ensure that the meaning of the sensor data will be more easily understood and managed by the intermediate devices. So sensor data can be queried and retrieved easily. Also it will enable many new applications based on wireless sensor networks including sensing as a service [29].
Semantic Sensor Network (SSN) Ontology has been dened for semantic sensing [30].
OpenIoT Ontology is also available, which is an extension of SSN Ontology. Once the data is received, annotation can be done and thus global relation of the sensor data can be established.
Figure 4. A Sample Metadata of an Environmental Sensor
2.3 Mobile Sensing
As mobile devices (e.g., smart phones, tablets, mobile routers etc.) have good computing resources now a days, a number of preprocessing operations could be done in mobile devices. Especially, the way smart phones are gaining computational ability; it will soon start serving as equally as normal computers do. Moreover, recharge ability, mobility and the access of internet are making mobile phones more available in cloud/internet than computers.
Crowd sensing and mobile crowdsourcing are another two popular and new domains that are being explored by researchers in [31], [32], [33]. This brings in a lot of attraction all together to the use of mobile phones or other devices, on-board sensors of the mobile devices and data generated by those devices.
That is why, using mobile devices as the integrator between WSN and internet is a nice idea. [34] discusses a distributed platform for mobile data analysis. [21]
implements a mobile version of GSN, MOSDEN, to run on smart phones. It collects data in mobile phones, from any sensor and when network is available, sends it to the OpenIoT [35]. It proposes to open the collected data for opportunistic sensing applications after preprocessing and storage. It also proposes the possibility of dynamic sensor discovery, mobile distributed load-balancing and task allocation, though the implementation of them is out of the scope of [21].
The limitation of MOSDEN is: it cannot do semantic sensing. It actually sends data to a cloud based X-GSN and then X-GSN does the rest of the processing and sends it to OpenIoT. By this way, MOSDEN basically just works as an initial interpreter. But mobile phones' use as the integrator between WSN and internet could be extended to further extent. It is possible to enable mobile phone to annotate sensor data based on SSN or OpenIoT Ontology and store meaningful data in local storage.
Moreover, MOSDEN does not manage the mobility of sensors. So when a sensor moves from one network to another network, the management needs to be done [36].
In OpenIoT, it is done by CUPUS middleware [37] which is running in an upper layer than MOSDEN. MOSDEN sends the data to XGSN and XGSN sends the data to CUPUS for mobility management. It is also possible to manage the mobility of sensors and BSs by another lightweight application like CUPUS running on mobile devices. By this way, it will also store the data based on the semantic annotation and will support the mobility of the sensors.
There are a number of mobile sensing applications available. In this thesis, three of them are surveyed: Thread [38], [39], Tiny-GSN [40] and MOSDEN. A comparative analysis among them has been highlighted in Table-1:
Table 1. Comparison among MOSDEN, Tiny-GSN and Thread Mobile Application:
Name of the fea-
tures MOSDEN Tiny-GSN Thread (Google)
Description Android app App built as a platform for opportunistic sensing apps
Android app to get access to meaningful sensor data in everyday life
Android app developed
specically fo- cusing Thread project
Function Through Plugin Through Wrap-
per NA
Data Collection Both 4m Sensors
and Hubs Both 4m Sensors
and Hubs 4m Hubs
Data Transfer GSN pulls the
data Subscribe to
GSN stream, push mecha- nism is under development
NA
Mobility No No Yes (BS does
that, built in the protocol stack) Semantic Sens-
ing No No No
Load Balancing No No No
Sensor Auto-
Discovery No No Yes (BS does
that, built in the protocol stack)
Pre-processing Yes Yes No
Applicability All type of sen-
sors All type of sen-
sors Thread specic
sensors. Other sensors are sup- ported through software update
Visualization Yes Yes Yes
Security and
Privacy No Yes Yes (Layer-2 se-
curity of data) App Develop-
ment Scope Yes Yes Yes
Advantages Plugins, so no need to recom- pile
Privacy pro- tection library for obfuscating location data
Product Specic
Disadvantages NA Wrappers, so
recompilation needed for every modication
Only 250 devices support
2.4 IoT Platforms
There are a number of IoT platforms available. Among them OpenIoT and FIWARE have been studied for this thesis. These two platforms provide support for IoT solutions and WSN integration. Data collection, presentation and management are supported by both of the platforms along with semantic annotation of the sensor data.
2.4.1 FIWARE
FIWARE is a european project for developing and deploying IoT applications [41], [42]. It is sponsored by the Future Internet Public Private Partnership (FI-PPP) program, created by the European Commission [43]. It oers an opensource cloud platform targeting mainly small and medium entrepreneurs. It provides open API for the developers in a variety of areas e.g., smart cities, sustainable transport, logistics, renewable energy, healthcare and environmental sustainability [44], [45].
But the current main focus is IoT and Smart Cities.
FIWARE provides many reusable software modules called Generic Enablers (GE) [46], [47], [42]. GE is the fundamental building block of FIWARE. The GEs are shared among dierent sectors to ensure a range of usage areas. Each GE oers many general functionalities along with suitable protocols and opensource APIs [43].
To develop an application in FIWARE means choosing the appropriate GE and building a software system composed of those GEs and non-GE software elements [48]. In general, the FIWARE architecture consists of several technical chapters.
The chapters consists of domain and each domain can have a number of GEs. As an example, the FIWARE IoT chapter consists of two domians - IoT Backend and IoT Edge. IoT GEs are spread over these two domains [46].
FIWARE doesn not have any specic format for data representation. It supports built-in basic data types of most programming languages. A data element is a triplet representation of name, type and value. So, every data element has a type where the meta data can be stored. Thus FIWARE supports semantic data as well. Similarly, FIWARE also supports the storage of contexts and events [49]. Even FIWARE release 3 has specic GEs for ontology support as well [50]. Based on the need of the applications, developers can use the specic GEs of semantic data, contexts or
events.
2.4.2 OpenIoT
OpenIoT is an open source middleware that supports data from any physical or virtual sensor, regardless of the manufacturer. It enables semantic annotation of sensor data based on predened Ontology. Thus, OpenIoT follows a cloud based paradigm to link all sensor data with the internet [18].
Some basic features of OpenIoT are mentioned below [51]:
• Collecting and processing data from virtually any sensor in the world, includ- ing physical devices, sensor processing algorithms, social media processing algorithms and more.
• Semantically annotating sensor data, according to the W3C Semantic Sensor Networks (SSN) specications.
• Streaming the data of the various sensors to a cloud computing infrastructure.
• Dynamically discovering/querying sensors and their data.
• Composing and delivering IoT services that comprise data from multiple sen- sors.
• Visualizing IoT data based on appropriate mashups (charts, graphs, maps etc.)
• Optimizing resources within the OpenIoT middleware and cloud computing infrastructure.
There are a number of components in OpenIoT. One of them is XGSN which sup- ports semantic sensing. As discussed in Section 2.1, This is an extension of GSN that supports semantic annotation of sensor data by adding metadata. XGSN works similar to GSN; additionally XGSN annotates the original sensor data [17]. Sensors send the data to the Wrappers of XGSN. These Wrappers are similar to GSN Wrap- pers. However, XGSN now annotates the data with metadata and then stores both data and metadata in LSM (Linked Stream Middleware) Database, another compo- nent of OpenIoT middleware, rather than publishing them in GSN. The metadata from the LSM can be used for other applications.
2.5 Summary
In this chapter, most impactful works have been presented on Internet of Things, Semantic Annotation, Mobile Sensing and IoT platforms. These are the basic of this thesis and anyone will get the required knowledge about these topics from this chapter.
3 MGSN for Enhanced Sensor Management
This chapter presents MGSN in detail in terms of functionalities and implementa- tion. As mentioned in Chapter-1 that the major research challenges are the im- plementation of semantic annotation on-device and the choice of ontology. So the proposed MGSN model follows the path to answer these questions rst.
From Chapter-2, it is observed that MOSDEN is a mobile based application for similar requirements though it can not do on-board annotation. It can be termed as the mobile version of GSN. It followed similar design principle and code structures.
As a result, working with MOSDEN to extend it in to MGSN is an easy decision concerning the rst research challenge. The solution would be to add semantic annotation ability with MOSDEN, which is similar to the semantic annotation part of XGSN. Thus, MGSN would be an extension of MOSDEN as like XGSN is an extension of GSN.
Concerning the second research challenge also, MOSDEN can be helpful. The code structure of MOSDEN follows that of GSN which is a part of OpenIoT platform.
So the choice of OpenIoT as the IoT platform for this thesis would be very help- ful. In such case, previous code structures of MOSDEN could be followed and reused. Moreover, FIWARE introduces the support for ontology only in the recent version [50]; previously FIWARE supported annotation dierently without the use of ontology. Keeping that in mind, OpenIoT is the platform and MOSDEN is the code base to start our implementation. As OpenIoT follows SSN Ontology [30] for semantic annotation, complying with OpenIoT platform, this thesis works with the same SSN Ontology.
3.1 MGSN Details
If a mobile device running an app like MOSDEN can operate as a Base Station, then there will be no setup issues regarding hub or BS. This will make sensor deployment as easier. However, the BS needs to be capable of sensor auto discovery and a secure, robust authentication process.
This thesis proposes Mobile Global Sensor Network (MGSN), an extension of MOS- DEN, with semantic annotation, local storage and mobility management ability. It will ensure distributed OpenIoT infrastructure with close proximity of WSN. As
MGSN can semantically sense the data and store them in local LSM based storage, it can then also perform a lot of processing on that data before sending it to cloud based storage. It can run initial classication or clustering algorithm on the col- lected data and can send those meaningful information to the internet for storage.
One of the use cases of such computation is the Big Data Stream that is going to be a reality soon.
Proposed MGSN provides a number of ecient approaches. As to mention, it enables OpenIoT services to be available very close to WSN. It will ensure ecient data processing and information management scopes. In addition to that, storing every sensor data in cloud based database not only requires huge storage but also requires huge bandwidth for transmission. When billions of sensors and IoT devices will be sending data to cloud based storages for advanced services and applications, it will make more sense when the nearby base stations (mobile devices running MGSN) will do initial processing and ltering before sending the data to the cloud.
Such preprocessing often needs task division and allocation with other devices for faster, ecient computation. Moreover, nearby BSs or MGSNs are supposed to have similar type of data, meaning they may need to share their data as well. Often, location based services or applications need to process data across multiple devices.
Current OpenIoT enables MOSDEN or XGSN to push the data vertically only to the cloud based LSM storage. But horizontal communication with other MOSDEN instances is not possible currently, though it was proposed by [21].
Proposed MGSN will support processing in ad-hoc mobile cloud. Unlike MOSDEN, MGSN will do horizontal data and process sharing to run classication/clustering algorithms and heavy parallel processing among the shared data. Even, multi- MGSN based classier algorithm can run on this shared environment - named as MGSN-cloud - formed with nearby MGSN instances. There are a number of imple- mentations for mobile phone based cloud. [52] proposes and simulates MapReduce based IoT cloud. It uses the IoT devices to form an ad-hoc cloud and divide tasks.
But the proposal has two major limitations - (i) IoT devices are still limited in hard- ware resources (power and computation) and (ii) lack of an authentication model.
However, their idea of having a private and on-demand mobile cloud can be used in our model.
This thesis proposes of using such an ad-hoc cloud model with the MGSNs which are running on mobile devices. Mobile phones, tablets, routers, Base Stations etc. are capable of performing complex operations. Most importantly, these devices are often
connected to a power outlet or rechargeable. So an ecient task allocation could be achieved based on [53], [54], which is implemented on Hadoop. But the choice of localized computation against MGSN-cloud computation and OpenIoT computation can be a problem. [55] proposes to solve such network choice problems in a mobile cloud environment. Based on the implementation of [55], MGSN could send the data to OpenIoT or could, perform shared operation in on-demand MGSN-cloud, or even choose to do on-board computation.
The strength of implementing MGSN cloud is that MapReduce can run policy based authentication and task sharing scheme on the mobile devices. Sharing of data and process will happen based on the policy which could be greedy, stochastic or opportunistic. Moreover, [55] supports adaptability to the dynamic changes in the MGSN-cloud. So, when MGSN nodes will go out of the proximity or enter in to the network, our proposed model will be able to adapt the changes during computation process.
Our proposed model can run on any mobile device operating system e.g., Arduino, TinyOS, Raspberry pi, Android etc. It gives wider choice to use MGSN as the integrator of WSN and internet.
3.2 Components of MGSN
As MGSN will make OpenIoT functionalities (e.g., semantic sensing, storing and processing) available in the BS along with the additional ability (e.g., formation of MGSN-cloud, policy based task management and authentication), big data stream processing will be more feasible. The ad-hoc cloud can be used for many innovative apps that involve heavy and complex processing e.g., Augmented Reality, Video processing and analysis, Video Surveillance System etc.
As in Figure-5, the proposed MGSN will have ve components in it: Ad-hoc Cloud, Semantic Sensing, Mobility Management, Data Pre-processing and Mini LSM Sor- age. The Ad-hoc Cloud will manage the formation of the cloud and dynamic man- agement of the members based on the work of [52], as discussed in the previous section. It will also manage the policies for the choice of network based on the work of [55]. Semantic Sensing works with the annotation part and the annotated data are stored temporarily in Mini LSM Storage. Mobility Management supports mo- bility of sensors and hubs from WSN to WSN and auto discovery of them based on
Figure 5. Proposed MGSN Block Diagram
CUPUS, a component of OpenIoT [37]. Data Pre-processing will do all clustering, classication and ltering.
However, as mentioned in Chapter-1, the scope of this thesis is to focus on the Semantic Sensing part only. Implementation of other components and their perfor- mance will be addressed in future works.
3.3 Communication Model of MGSN
For every kind of communication with MGSN (with sensors, servers or other ele- ments), there should be a concerned plugin. Based on the plugin, the communi- cation mode will be decided. The communication protocols and methods will be implemented in the plugins. These plugins are similar to the plugins of MOSDEN, which are similar to the wrappers of GSN.
The interaction of MGSN with OpenIoT and rest of the internet is autonomous and via the network the mobile device is connected to. The communication between WSN and MGSN will happen through an IEEE 802.15.4 supportable interface, if that is installed in the mobile device. If such support is not available, MGSN will be able to collect only on-board sensors' data; external sensors can not be com- municated normally. However, if the sensors have IP addresses and are congured properly, normal web-based communication may be possible. But such investiga- tions have not been done in this thesis.
Figure 6. MGSN Interaction with OpenIoT
For ad-hoc cloud, the communication has to be via any wireless protocol. But discussion on that is subject to further research.
3.4 Execution Flow
When sensor data will be received by the MGSN, MGSN will annotate the data with RDF, i.e., what the sensor data means, its unit, what network it belongs to, what the sensor hardware is and where to store the data. Then MGSN will store it in the local Mini-LSM and sends the processed data to the appropriate LSM.
When sensors send data, MGSN uses the wrapper in accordance with the data specication mentioned in the sensors manual. The hardware data sent by the sensor needs to be dynamically annotated based on the OpenIoT ontology. So it will enable OpenIoT ontology, which is an extension of SSN ontology, in the Base Stations i.e., MGSN.
Processing could also be done before/after annotating the data. Many classication and clustering algorithms could run in MGSN to do initial ltering or outlier detec- tion. In such case, data will be ltered at the base station right after data is being generated and thus, more eciency will be achieved towards network transmission.
Especially, in case of big data, when a lot of data will be generated by dierent types of sensors (e.g., video, audio, etc.) such pre-processing will be very helpful.
3.5 Summary
This chapter presents the details of MGSN. The components, block diagram and execution ow of MGSN were discussed in detail.
4 IMPLEMENTATION
As this thesis focuses only on the Semantic Sensing component of MGSN, the imple- mentation is only focused in receiving sensor data and annotating them. Mini LSM Storage is not an objective of this thesis. As a result, the annotated data can not be stored in the mobile devices. It has to be sent to the OpenIoT server in cloud. LSM Database is a component of OpenIoT. It is implemented with a semantic database called Virtuoso [51].
Implementation of MGSN has two major parts. First, it needs to have OpenIoT middleware installed and running in a server with all of its components, especially XGSN and LSM. Second, MGSN should be running in a mobile phone which will annotate the data and send it directly to OpenIoT. The annotated data will be received by LSM and will be stored there.
4.1 Installing OpenIoT Middleware
OpenIoT code has been collected from github [51] and installed in the LTU cloud space. The OS has been Ubuntu. JBoss needs to be installed to run OpenIoT, as the components of OpenIoT are mainly developed in Java EE. JBoss is a full Java EE server. For this thesis JBoss 7.1 has been installed and Maven has been congured as well for project dependencies management.
The IP of OpenIoT server is: 130.240.134.132 and XGSN Port is: 22001. The LSM Port is: 8890
OpenIoT has been congured and installed correctly. It has worked ne. Sen- sor registration, virtual sensor denition, wrapper class generation, data collection, publishing, annotation and storing data in LSM - all these sub-tasks have been im- plemented successfully. Even, data from a real-life weather sensor of Sense Smart City Project [56] has been pushed to the XGSN. XGSN collected that data from a JSON API where the weather sensor posted its data. XGSN then has annotated that data and the annotated data has also been stored in LSM.
A separate wrapper class has to be written for this purpose in java called LtuTest- Wrapper. The code is given in Appendix 1. Figure 7 shows the API from where data is pushed to XGSN of the OpenIoT.
Figure 7. JSON API of the Sense Smart City Weather Sensor
This thesis focuses on XGSN and how it works. In the code segments, the working procedure of XGSN has been observed deeply. The XGSN work ow can be better understood from the following two Figures:
Figure 8. Sensor - XGSN Interaction Sequence Diagram (High-level)
Figure 9. Sensor - XGSN Interaction Process Flow (Detail)
XGSN has looked in to the API periodically if there is an update of data through the wrapper class. Once there is an update, the data is collected. It has been observed that data has been pushed in to Virtuoso after the annotaion being done based on SSN Ontology. The annotated data could be seen in Virtuoso; but the GSN only has shown the original data. The query for Virtuoso is given in Appendix 1.
Also, a bug has been noticed regarding storing data in Virtuoso after annotation.
The metadata le that has been generated automatically from OpenIoT platform supports both uppercase and lowercase character as eld name. Where as, XGSN only supports lowercase characters. As a result, though the eld names have been same in metadata le and XGSN, data could not have been pushed in to Virtouso. It
is believed that the problem occurs due to an issue regarding serialization. However, later on, all lowercase characters have been used for eld names to avoid ambiguity.
Detail of the OpenIoT installation is available in Appendix 1
4.2 Development of MGSN
MOSDEN code has been collected from BitBucket. MOSDEN has been developed in Eclipse ADT. Initially, this thesis has tried to convert the code in Android Studio.
However, due to some problems of code migration, later on the development has been made in Eclipse ADT.
At the beginning the objective has been to communicate with OpenIoT from an Android App. It could be done successfully from Android Studio. However, after this thesis has changed the IDE to Eclipse, such communication could not be done.
By communication, simple HTTP Post has been tried to the URL of OpenIoT.
However, the problem has been identied as a proxy related block of the messages from the IDE, as the android emulator could not communicate via the virtual proxy network. After such problem, it has been tried from an android phone in real-time.
This time, the problem has been resolved.
However, when such communication has been tried from the MOSDEN code, a number of problems have been encountered. The MOSDEN code executed perfectly, but during the operation of the app, it keeps crashing repeatedly. Then there is a problem with internal memory management as well which could not be rectied.
Detail of the MGSN development works is available in Appendix 2
4.3 Summary
Chapter 4 presents the implementation process of this thesis. Both the installation of OpenIoT middleware and the development of MGSN from MOSDEN are discussed.
MGSN can not be completely developed due to some implementation issues. The problems are also discussed in this Chapter.
5 CONCLUSIONS AND FUTURE WORK
In this chapter, we discuss the conclusion and future work related to this thesis work. The limitations of the thesis are also discussed. The objective of the thesis is to develop a model that can semantically annotate sensor data in mobile devices. By sensor, both on-board and external sensors are meant. The research challenges are - the choice of the process for annotating data and the choice of ontology for semantic annotation. The summary will demonstrate the key points about this thesis. The future research scopes are also outlined at the end of this chapter.
5.1 Summary
Advanced sensor management includes dierent tasks which require proper under- standing of the sensor data. Thus semantic annotation is always a nice way to handle this requirement. However, on-device sensors or external sensors can send dierent types of data. These data are not uniformed. As a result, without proper structure, semantic annotation would be very dicult and might be unusable by other parties.
Ontology solves these problems. But managing ontology and similarly annotating sensor data can be very dicult without the presence of a proper framework/mid- dleware. OpenIoT helps from that perspective, though the components of OpenIoT are not made to run on mobile platforms. As a result, though OpenIoT works ne, there is a need of separate implementation of similar application for mobile devices.
MOSDEN focuses on opportunistic sensing platform. Thus MOSDEN is one of the rst works in this eld, but lack of semantic annotation causes the need of another research work that will carry out the implementation of a mobile app which will an- notate sensor data based on a specic ontology. MGSN focuses on this requirement and the model can work nicely along with OpenIoT and SSN Ontology.
MGSN can also be helpful for the purpose of energy ecient use of sensors, mobile devices and behavior modication of users through specic suggestions analysing their pattern and need.
5.2 Limitations
The limitations of this work include the development issues that have been discussed in Chapter-4. Moreover, MGSN to be fully eective, and to suggest behavioral change of the users, data analysis is necessary and a proper implementation would be mandatory.
5.3 Future Work
The implementation of all the components of MGSN can be the rst priority for future work. A scenario of saving energy of sensors using annotated energy usage data could be very interesting to implement. Moreover, MGSN enables data stream to be processed at the rst stage after generation and thus, reduce the amount of load in network trac and storage. It could be an interesting point to simulate and check how much trac MGSN can reduce through classication, clustering and ltering of sensor data.
Also, policy could be enabled to ensure which data to process locally, which data to send on the MGSN cloud or which data to send to the OpenIoT in the internet.
At the same time, making MGSN available in mobile devices through ecient but specic apps, OpenIoT will be more distributed and thus will be more compatible for large scale data collection. Research on the direction of Big Stream and the compatibility of MGSN in that domain can be an interesting area.
Another very exciting scope of research could be recommending sustainable practices or behavior for the end-users. Both external and on-board sensor data will be necessary for such requirement. In such case, it may also be analyzed if future behabior of end-users could be predicted.
REFERENCES
[1] M. Collina, M. Bartolucci, A. Vanelli-Coralli, and G. E. Corazza, Internet of things application layer protocol analysis over error and delay prone links, in Advanced Satellite Multimedia Systems Conference and the 13th Signal Pro- cessing for Space Communications Workshop (ASMS/SPSC), 2014 7th. IEEE, 2014, pp. 398404.
[2] L. Skorin-Kapov, K. Pripuzic, M. Marjanovic, A. Antonic, and I. P. Zarko, En- ergy ecient and quality-driven continuous sensor management for mobile iot applications, in Collaborative Computing: Networking, Applications and Work- sharing (CollaborateCom), 2014 International Conference on. IEEE, 2014, pp.
397406.
[3] M. Marjanovi¢, L. Skorin-Kapov, K. Pripuºi¢, A. Antoni¢, and I. Podnar arko, Energy-aware and quality-driven sensor management for green mobile crowd sensing, J. Netw. Comput. Appl., vol. 59, no. C, pp. 95108, Jan. 2016.
[Online]. Available: http://dx.doi.org/10.1016/j.jnca.2015.06.023
[4] A. Z. K. Mitra and C. Ahlund, Context-aware qoe modelling, measurement, and prediction in mobile computing systems, IEEE Transactions on Mobile Computing, vol. 14, no. 5, pp. 920936, May 2015.
[5] C. Perera, A. Zaslavsky, P. Christen, A. Salehi, and D. Georgakopoulos, Cap- turing sensor data from mobile phones using global sensor network middleware, in Personal Indoor and Mobile Radio Communications (PIMRC), 2012 IEEE 23rd International Symposium on, Sept 2012, pp. 2429.
[6] P. P. Jayaraman, A. Sinha, W. Sherchan, S. Krishnaswamy, A. Zaslavsky, P. D.
Haghighi, S. Loke, and M. T. Do, Here-n-now: A framework for context-aware mobile crowdsensing. Citeseer.
[7] Unlocking the promise of a connected world: Using the cloud to enable the internet of things, http://www.oracle.com/us/solutions/internetofthings/iot- and-cloud-wp-2686546.pdf, 2015.
[8] P. P. Jayaraman, D. Palmer, A. Zaslavsky, and D. Georgakopoulos, Do-it- yourself digital agriculture applications with semantically enhanced iot plat- form, in Intelligent Sensors, Sensor Networks and Information Processing (ISSNIP), 2015 IEEE Tenth International Conference on, April 2015, pp. 16.
[9] P. P. Jayaraman, D. Palmer, A. Zaslavsky, A. Salehi, and D. Georgakopoulos, Addressing Information Processing Needs of Digital Agriculture with OpenIoT Platform. Cham: Springer International Publishing, 2015, pp. 137152.
[Online]. Available: http://dx.doi.org/10.1007/978-3-319-16546-2_11
[10] Sustainable and ethical enterprise group, http://www.business.mmu.ac.uk/
seeg/whatis.php, accessed: 2016.07.05.
[11] P. C. S. Pongnumkul and N. Surasvadi, Applications of smartphone-based sensors in agriculture: A systematic review of research, vol. 2015, 2015, pp.
195 308:1195 308:18.
[12] M. S. Hossen, A. F. M. S. Kabir, R. H. Khan, and A. Azfar, Interconnection between 802.15.4 devices and ipv6: Implications and existing approaches, CoRR, vol. abs/1002.1146, 2010. [Online]. Available:
http://arxiv.org/abs/1002.1146
[13] S. Yin, Q. Li, and O. Gnawali, Interconnecting wi devices with ieee 802.15.4 devices without using a gateway, in Distributed Computing in Sensor Systems (DCOSS), 2015 International Conference on, June 2015, pp. 127136.
[14] P. Sthapit, Y.-S. Choi, and G.-R. Kwon, A fast association scheme over ieee 802.15. 4 based mobile sensor network.
[15] A. Salehi and K. Aberer, Demo abstract: Gsn, quick and simple sensor network deployment.
[16] K. Aberer, M. Hauswirth, and A. Salehi, The global sensor networks middle- ware for ecient and exible deployment and interconnection of sensor net- works, Tech. Rep., 2006.
[17] J.-P. Calbimonte, S. Sarni, J. Eberle, and K. Aberer, Xgsn: An open-source semantic sensing middleware for the web of things, in 7th International Work- shop on Semantic Sensor Networks, no. EPFL-CONF-200926, 2014.
[18] I. Podnar arko, K. Pripuºi¢, and M. Serrano, Interoperability and open-source solutions for the internet of things: International workshop, fp7 openiot project, held in conjunction with softcom 2014, split, croatia, september 18, 2014, in- vited papers, Lecture Notes in Computer Science, 2015.
[19] K. Aberer, M. Hauswirth, and A. Salehi, Infrastructure for data processing in large-scale interconnected sensor networks, in Proceedings of the 2007 International Conference on Mobile Data Management, ser. MDM '07.
Washington, DC, USA: IEEE Computer Society, 2007, pp. 198205. [Online].
Available: http://dx.doi.org/10.1109/MDM.2007.36
[20] John Soldatos, Eu openiot project fp7 ict-2011 1.3: Internet-connected ob- jects, http://www.connectedworld.gr/ppts-2014/soldatos.pdf, 2014.
[21] P. P. Jayaraman, C. Perera, D. Georgakopoulos, and A. Zaslavsky, Ecient op- portunistic sensing using mobile collaborative platform mosden, in Collabora- tive Computing: Networking, Applications and Worksharing (Collaboratecom), 2013 9th International Conference Conference on, Oct 2013, pp. 7786.
[22] F. Ciancetta, B. D'Apice, D. Gallo, and C. Landi, Plug-n-play smart sensor based on web service, IEEE Sensors Journal, vol. 7, no. 5, pp. 882889, May 2007.
[23] M. Ruta, F. Scioscia, and E. D. Sciascio, Enabling the semantic web of things:
Framework and architecture, in Semantic Computing (ICSC), 2012 IEEE Sixth International Conference on, Sept 2012, pp. 345347.
[24] S. Dill, N. Eiron, D. Gibson, D. Gruhl, R. Guha, A. Jhingran, T. Kanungo, S. Rajagopalan, A. Tomkins, J. A. Tomlin et al., Semtag and seeker: Boot- strapping the semantic web via automated semantic annotation, in Proceedings of the 12th international conference on World Wide Web. ACM, 2003, pp. 178 186.
[25] V. Uren, P. Cimiano, J. Iria, S. Handschuh, M. Vargas-Vera, E. Motta, and F. Ciravegna, Semantic annotation for knowledge management: Requirements and a survey of the state of the art, Web Semantics: science, services and agents on the World Wide Web, vol. 4, no. 1, pp. 1428, 2006.
[26] L. Reeve and H. Han, Survey of semantic annotation platforms, in Proceedings of the 2005 ACM symposium on Applied computing. ACM, 2005, pp. 1634 1638.
[27] M. Klusch, P. Kapahnke, S. Schulte, F. Lecue, and A. Bernstein, Semantic web service search: a brief survey, KI-Künstliche Intelligenz, pp. 19, 2015.
[28] S. Handschuh and S. Staab, Annotation for the semantic web. IOS Press, 2003, vol. 96.
[29] A. Zaslavsky, C. Perera, and D. Georgakopoulos, Sensing as a service and big data, arXiv preprint arXiv:1301.0159, 2013.
[30] W. S. S. N. I. Group, Semantic sensor network ontology, https://www.w3.
org/2005/Incubator/ssn/ssnx/ssn, accessed: 2016.02.05.
[31] A. Bazzi and A. Zanella, Position based routing in crowd sensing vehicular networks, Ad Hoc Netw., vol. 36, no. P2, pp. 409424, Jan. 2016. [Online].
Available: http://dx.doi.org/10.1016/j.adhoc.2015.06.005
[32] W. Sherchan, P. P. Jayaraman, S. Krishnaswamy, A. Zaslavsky, S. Loke, and A. Sinha, Using on-the-move mining for mobile crowdsensing, in Mobile Data Management (MDM), 2012 IEEE 13th International Conference on. IEEE, 2012, pp. 115124.
[33] T. KINDBERG, People, places, things: Web presence for the real world, Mobile Networks and Applications, vol. 7, no. 5, pp. 365376, 2002.
[34] P. P. Jayaraman, J. B. Gomes, H. L. Nguyen, Z. S. Abdallah, S. Krish- naswamy, and A. Zaslavsky, Cardap: A scalable energy-ecient context aware distributed mobile data analytics platform for the fog, in Advances in Databases and Information Systems: 18th East European Conference, AD- BIS 2014, Ohrid, Macedonia, September 7-10, 2014. Proceedings, vol. 8716.
Springer, 2014, p. 192.
[35] C. Perera, P. P. Jayaraman, A. Zaslavsky, D. Georgakopoulos, and P. Christen, Mosden: An internet of things middleware for resource constrained mobile devices, in 47th Hawaii International Conference on System Sciences (HICSS).
IEEE, 2014, pp. 10531062.
[36] V. Dyo and C. Mascolo, Ecient node discovery in mobile wireless sensor networks, in Distributed Computing in Sensor Systems. Springer, 2008, pp.
478485.
[37] A. Antonic, K. Roankovic, M. Marjanovic, K. Pripuic, and I. P. arko, A mobile crowdsensing ecosystem enabled by a cloud-based publish/subscribe middle- ware, in Future Internet of Things and Cloud (FiCloud), 2014 International Conference on, Aug 2014, pp. 107114.
[38] Ug103.11: Application development fundamentals: Thread, https://www.
silabs.com/Support%20Documents/TechnicalDocs/UG103.11.pdf.
[39] Thread technical overview - berlin, https://threadgroup.org/technology/
downloads, 2015.
[40] J. Eberle, Sustainable and ethical enterprise group, https://github.com/
LSIR/gsn/tree/master/gsn-tiny, accessed: 2016.01.25.
[41] B. Herudek, What is ware? a european internet of things smart city platform, https://www.linkedin.com/pulse/what-ware-european-internet- things-smart-city-platform-herudek, accessed: 2016.09.05.
[42] P. Fernýndez, J. M. Santana, S. Ortega, A. Trujillo, J. P. Suýrez, C. DomÃnguez, J. Santana, and A. Sýnchez, Smartport: A platform for sensor data monitoring in a seaport based on ware, Sensors, vol. 16, no. 3, p. 417, 2016. [Online]. Available: http://www.mdpi.com/1424-8220/16/3/417 [43] S. Di Cola, C. Tran, K.-K. Lau, A. Celesti, and M. Fazio, A
Heterogeneous Approach for Developing Applications with FIWARE GEs.
Cham: Springer International Publishing, 2015, pp. 6579. [Online]. Available:
http://dx.doi.org/10.1007/978-3-319-24072-5_5
[44] F. Ramparany, F. G. Marquez, J. Soriano, and T. Elsaleh, Handling smart environment devices, data and services at the semantic level with the -ware core platform, in Big Data (Big Data), 2014 IEEE International Conference on, Oct 2014, pp. 1420.
[45] Overall ware vision, http://forge.ware.org/plugins/mediawiki/wiki/ware/
index.php/Overall_FIWARE_Vision, accessed: 2016.09.05.
[46] Fiware architecture, http://forge.ware.org/plugins/mediawiki/wiki/ware/
index.php/FI-WARE_Architecture, accessed: 2016.09.05.
[47] T. Zahariadis, A. Papadakis, F. Alvarez, J. Gonzalez, F. Lopez, F. Facca, and Y. Al-Hazmi, Fiware lab: Managing resources and services in a cloud federa- tion supporting future internet applications, in Utility and Cloud Computing (UCC), 2014 IEEE/ACM 7th International Conference on, Dec 2014, pp. 792 799.
[48] M. Fazio, A. Celesti, F. G. Mýrquez, A. Glikson, and M. Villari, Exploiting the ware cloud platform to develop a remote patient monitoring system, in 2015 IEEE Symposium on Computers and Communication (ISCC), July 2015, pp. 264270.
[49] Context management architecture, http://forge.ware.org/plugins/
mediawiki/wiki/ware/index.php/Data/Context_Management_Architecture, accessed: 2016.09.05.
[50] Fiware.openspecication.data.semanticsupport, http://forge.ware.org/
plugins/mediawiki/wiki/ware/index.php/FI-WARE_Architecture, accessed:
2016.09.05.
[51] N. Kefalakis, Openiot - the open source internet of things, https://github.
com/OpenIotOrg/openiot, accessed: 2016.03.18.
[52] R. Hasan, M. M. Hossain, and R. Khan, Aura: An iot based cloud infras- tructure for localized mobile computation outsourcing, in Mobile Cloud Com- puting, Services, and Engineering (MobileCloud), 2015 3rd IEEE International Conference on, March 2015, pp. 183188.
[53] Eugene E. Marinelli, Hyrax: Cloud computing on mobile devices using mapre- duce, Master's thesis, School of Computer Science Carnegie Mellon University, Pittsburgh, PA 15213, September 2009.
[54] Chye Liang Vincent Teo, Hyrax: Crowdsourcing mobile devices to develop proximity-based mobile clouds, Master's thesis, School of Computer Science Carnegie Mellon University, Pittsburgh, PA 15213, August 2012.
[55] K. Mitra, S. Saguna, C. Ahlund, and D. G. Lulea, M2c2: A mobility manage- ment system for mobile cloud computing, in Wireless Communications and Networking Conference (WCNC), 2015 IEEE, March 2015, pp. 16081613.
[56] Additional project information, http://sensesmartcity.org/info2.htm, ac- cessed: 2016.09.12.
MGSN has been developed on Android Studio (1.5.1), Windows 64 Bit. The entire
source code is available on GitHub at https://github.com/m6461/XGSN_MobileAnalytics.
Figure A1.1. OpenIoT Installation
Figure A1.2. OpenIoT Wrapper File Conguration of X-GSN
(continues)
Figure A1.3. OpenIoT Execution Command
Figure A1.4. OpenIoT Query to Show Linked Data from Sensors
(continues)
Figure A1.5. Wrapper Code to Push Sensor Data in OpenIoT from JSON API - 1
(continues)
Figure A1.6. Wrapper Code to Push Sensor Data in OpenIoT from JSON API - 2
(continues)
Figure A1.7. Wrapper Code to Push Sensor Data in OpenIoT from JSON API - 3
(continues)
Figure A1.8. Wrapper Code to Push Sensor Data in OpenIoT from JSON API - 4
(continues)
Figure A1.9. Metadata Information for the JSON API
Figure A2.1. MGSN Code to Mitigate the Internal Resource Problem
