Context-aware Computing in the Internet of Things: A Survey on Internet of Things From Industrial Market PerspectiveNote: IEEE ACCESS Journal Official Title for Citations: ’A Survey on Internet of Things From Industrial Market Perspective’

# Context-aware Computing in the Internet of Things: A Survey on Internet of Things From Industrial Market Perspective Note: IEEE ACCESS Journal Official Title for Citations: ’A Survey on Internet of Things From Industrial Market Perspective’

Charith Perera,  Chi Harold Liu , Srimal Jayawardena,  Min Chen,  Charith Perera, and Srimal Jayawardena, are with the Research School of Computer Science, The Australian National University, Canberra, ACT 0200, Australia. (e-mail: firstname.lastname@ieee.org)Chi Harold Liu is with Beijing Institute of Technology, China. (e-mail: chiliu@bit.edu.cn)Min Chen is with Huazhong University of Science and Technology, China. (e-mail: minchen@ieee.org)Manuscript received xxx xx, xxxx; revised xxx xx, xxxx.
###### Abstract

The Internet of Things (IoT) is a dynamic global information network consisting of Internet-connected objects, such as RFIDs, sensors, and actuators, as well as other instruments and smart appliances that are becoming an integral component of the Internet. Over the last few years, we have seen a plethora of IoT solutions making their way into the industry marketplace. Context-aware communication and computing has played a critical role throughout the last few years of ubiquitous computing and is expected to play a significant role in the IoT paradigm as well. In this article, we examine a variety of popular and innovative IoT solutions in terms of context-aware technology perspectives. More importantly, we evaluate these IoT solutions using a framework that we built around well-known context-aware computing theories. This survey is intended to serve as a guideline and a conceptual framework for context-aware product development and research in the IoT paradigm. It also provides a systematic exploration of existing IoT products in the marketplace and highlights a number of potentially significant research directions and trends.

Internet of things, industry solutions, context-awareness, product review, IoT marketplace

## I Introduction

Over the last few years the Internet of Things (IoT) [1] has gained significant attention from both industry and academia. Since the term was introduced in the late 1990s many solutions have been introduced to the IoT marketplace by different types of organization ranging from start-ups, academic institutions, government organizations and large enterprises [2]. IoT’s popularity is governed by both the value that it promises to create and market growth and predictions [3]. The IoT allows ’people and things to be connected Anytime, Anyplace, with Anything and Anyone, ideally using Any path/network and Any service[4]. Such technology will help to create ’a better world for human beings’, where objects around us know what we like, what we want, and what we need and act accordingly without explicit instructions [2].

Context-aware communication and computing is a key technology that enables intelligent interactions such as those which the IoT paradigm envisions. Let us briefly introduce some of the terms in this domain which will help to better understand the remaining sections. Context can be defined as any information that can be used to characterize the situation of an entity. An entity is a person, place, piece of software, software service or object that is considered relevant to the interaction between a user and an application, including the user and application themselves [5]. Context-awareness can be defined as the ability of a system to provide relevant information or services to users using context information where relevance depends on the user’s task [5]. Context-aware communication and computing has been researched extensively since the early 2000s and several surveys have been conducted in this field. The latest survey on context-aware computing focusing on the IoT was conducted by Perera et al. [2]. Several other important surveys are analysed and listed in [2]. However, all these surveys focus on academic research.

To the best of our knowledge, however, no survey has focused on industrial IoT solutions. All the above-mentioned surveys have reviewed the solutions proposed by the academic and research community and refer to scholarly publications produced by the respective researchers. In this paper, we review IoT solutions that have been proposed, designed, developed, and brought into the market by industrial organizations. These organizations range from start-ups and small and medium enterprises to large corporations. Because of their industrial and market-driven nature, most of the IoT solutions in the market are not published as academic work. Therefore, we collected information about the solutions from their respective websites, demo videos, technical specifications, and consumer reviews. Understanding how context-aware technologies are used in the IoT solutions in the industry’s marketplace is vital for academics, researchers, and industrialists so they can identify trends, industry requirements, demands, and innovation opportunities.

The rest of the article is organized as follows. In Section II, we briefly analyse IoT marketplace trends and growth. The evolution of context-aware technologies and applications are presented in Section III. Then, we introduce the theoretical foundation and our evaluation framework used in this paper in Section IV. Subsequently, in Section V, we review a selected number of IoT solutions from context-aware perspective. Later, we present lessons learned and innovation opportunities based on the evaluation results in Section VI. Finally, we present the conclusion remarks.

## Ii Internet of Things Marketplace

The vision of the IoT has been heavily energised by statistics and predictions. In this section, we discuss some of the statistics and facts related to the IoT which allows us to understand how the IoT has grown over the years and how it is expected to grow in the future. Further, these statistics and facts highlight the future trends in the industry marketplace.

It is estimated that there about 1.5 billion Internet-enabled PCs and over 1 billion Internet-enabled mobile phones today. These two categories will be joined by Internet-enabled smart objects [6, 7] in the future. By 2020, there will be 50 to 100 billion devices connected to the Internet, ranging from smartphones, PCs, and ATMs (Automated Teller Machine) to manufacturing equipment in factories and products in shipping containers [8]. As depicted in Figure 1, the number of things connected to the Internet exceeded the number of people on Earth in 2008. According to CISCO, each individual on earth will have more than six devices connected to the Internet by 2020.

According to BCC Research 2011 market report on sensors, the global market for sensors was around $56.3 billion in 2010. In 2011, it was around$62.8 billion. The global market for sensors is expected to increase to $91.5 billion by 2016, at a compound annual growth rate of 7.8%. One of the techniques for connecting everyday objects into networks is radio frequency identification â RFID technology [9]. In this technology, the data carried by the chip attached to an object is transmitted via wireless links. RFID has the capability to convert dump devices into comparatively smart objects. RFID systems can be used wherever automated labelling, identification, registration, storage, monitoring, or transport is required to increase efficiency and effectiveness. According to Frost & Sullivan (2011), the global RFID market was valued at from$3 billion to $4 billion in 2009. The RFID market will grow by 20% per year through 2016 and reach a volume of approximately from$6.5 billion to almost $9 billion. According to Figure 2, it is expected that five main sectors, education, transportation, industry, healthcare, and retails, will generate 76% of the total RFID market demand by 2016. “Smart city” [10] is a concept aimed at providing a set of new generation services and infrastructure with the help of information and communication technologies (ICT). Smart cities are expected to be composed of many different smart domains. Smart transportation, smart security and smart energy management are some of the most important components for building smart cities [11]. However, in term of market, smart homes, smart grid, smart healthcare, and smart transportation solutions are expected to generate the majority of sales. According to MarketsandMarkets report on Smart Cities Market (2011 - 2016), the global smart city market is expected to cross$1 trillion by 2016, growing at a CAGR of 14.2% as illustrated in Figure 3.

The interconnection and communication between everyday objects, in the IoT paradigm, enables many applications in many domains. Asin and Gascon [12] have listed 54 application domains under 12 categories: smart cities, smart environment, smart water, smart metering, security and emergencies, retail, logistics, industrial control, smart agriculture, smart animal farming, domestic and home automation, and eHealth. After analysing the industry marketplace and careful consideration, we classified the popular existing IoT solutions in the marketplace into five different categories: smart wearable, smart home, smart city, smart environment, smart enterprise. In this paper, we review over 100 different IoT solutions in total. It is important to note that not all the solutions we examined are listed in the technology review in Table II. For the review, we selected a wide range of IoT products which demonstrate different context-aware functionalities.

## Iii Evolution of Context-aware Technology

It is important to understand the evolution of the Internet before discussing the evolution of context-aware technologies. The Internet broadly evolved in five phases as illustrated in Figure 4. The evolution of Internet begins with connecting two computers together and then moved towards creating the World Wide Web by connecting large number of computers together. Mobile-Internet emerged when mobile devices were connected to the Internet. People’s identities were added to the Internet via social networks [13]. Finally, the Internet of Things emerged, comprised of everyday objects added to the Internet. During the course of these phases, the application of context-aware communication and computing changed significantly [2].

In the early phase of computer networking when computers were connected to each other in point-to-point fashion, context-aware functionalities were not widely used. Providing help to users based on the context (of the application currently open) was one of the fundamental context-aware interactions provided in early computer applications and operating systems. Another popular use of context is context-aware menus that help users to perform tasks tailored to each situation in a given application. When the Internet came into being, location information started to become critical context information. Location information (retrieved through IP addresses) were used by services offered over the Internet in order to provide location-aware customization to users. Once the mobile devices (phones and tablets) became a popular and integral part of everyday life, context information collected from sensors built-in to the devices (e.g. accelerometer, gravity, gyroscope, GPS, linear accelerometer, and rotation vector, orientation, geomagnetic field, and proximity, and light, pressure, humidity and temperature) were used to provide context-aware functionality. For example, built-in sensors are used to determine user activities, environmental monitoring, health and well-being, location and so on [14].

Over the last few years social networking [15] has become popular and widely used. Context information gathered through social networking services [16] (e.g. Facebook, Myspace, Twitter, and Foursquare) has been fused with the other context information retrieved through mobile devices to build novel context-aware applications such as activity predictions, recommendations, and personal assistance [17]. For example, a mobile application may offer context-aware functionalities by fusing location information retrieved from mobile phones and recent ‘likes’ retrieved from social media sites to recommend nearby restaurants that a user might like. In the next phase, ‘things’ were connected to the Internet by creating the IoT paradigm. An example of context-aware functionality provided in the IoT paradigm would be an Internet-connected refrigerator telling users what is inside it, what needs to be purchased or what kind of recipes can be prepared for dinner. When the user leaves the office, the application autonomously does the shopping and guides the user to a particular shopping market so s/he can collect the goods it has purchased. In order to perform such tasks, the application must fuse location data, user preferences, activity prediction, user schedules, information retrieved through the refrigerator (i.e. shopping list) and many more. In the light of the above examples, it is evident that the complexity of collecting, processing and fusing information has increased over time. The amount of information collected to aid decision-making has also increased significantly.

## Iv Theoretical Foundation and Evaluation Framework

This section discusses context-aware theories and related historic developments over time. The evaluation framework which we used to review IoT products in the marketplace are built upon the theoretical foundation presented in this section. First, we lay the theoretical foundation and secondly we discuss the evaluation framework.

### Iv-a Context-aware Computing Theories

The term context has been defined by many researchers. Dey et al. [18] have evaluated and highlighted the weaknesses of these definitions. Dey claimed that the definition provided by Schilit and Theimer [19] was based on examples and cannot be used to identify new context. Further, Dey claimed that definitions provided by Brown [20], Franklin and Flachsbart [21], Rodden et al. [22], Hull et al. [23], and Ward et al. [24] used synonyms to refer to context, such as ‘environment’ and ‘situation’. Therefore, these definitions also cannot be used to identify new context. Abowd and Mynatt [25] have identified the five W’s (Who, What, Where, When, Why) as the minimum information that is necessary to understand context. Schilit et al. [26] and Pascoe [27] have also defined the term context.

We accept the definition of context provided by Abowd et al. [5] to be used in this research work, because their definition can be used to identify context from data in general. We presented the definition of context in Section I.

The term context awareness, also called sentient, was first introduced by Schilit and Theimer [19] in 1994. Later, it was defined by Ryan et al. [28]. In both cases, the focus was on computer applications and systems. As stated by Abowd et al. [5], those definitions are too specific and cannot be used to identify whether a given system is a context-aware system or not. We presented the definition provided by Abowd et al. [5] in Section I. After analysing and comparing the two previous efforts conducted by Schilit et al. [26] and Pascoe [27], Abowd et al. [5] identified three features that a context-aware application can support: presentation, execution, and tagging. Even though, the IoT vision was not known at the time these features are identified, they are highly applicable to the IoT paradigm as well. We elaborate these features from an IoT perspective.

• Presentation: Context can be used to decide what information and services need to be presented to the user. Let us consider a smart [29] environment scenario. When a user enters a supermarket and takes their smart phone out, what they want to see is their shopping list. Context-aware mobile applications need to connect to kitchen appliances such as a smart refrigerator [30] in the home to retrieve the shopping list and present it to the user. This provides the idea of presenting information based on context such as location, time, etc. By definition, IoT promises to provide any service anytime, anyplace, with anything and anyone, ideally using any path/network.

• Execution: Automatic execution of services is also a critical feature in the IoT paradigm. Let us consider a smart home [29] environment. When a user starts driving home from their office, the IoT application employed in the house should switch on the air condition system and switch on the coffee machine to be ready to use by the time the user steps into their house. These actions need to be taken automatically based on the context. Machine-to-machine communication is a significant part of the IoT.

• Tagging: In the IoT paradigm, there will be a large number of sensors attached to everyday objects. These objects will produce large volumes of sensor data that has to be collected, analysed, fused and interpreted [31]. Sensor data produced by a single sensor will not provide the necessary information that can be used to fully understand the situation [32]. Therefore, sensor data collected through multiple sensors needs to be fused together [33]. In order to accomplish the sensor data fusion task, context needs to be collected. Context needs to be tagged together with the sensor data to be processed and understood later. Context annotation plays a significant role in context-aware computing research. The tagging operation also identified as annotation.

In Figure 5, we summarise three different context-aware features presented by researchers. It is clear that all these classification methods have similarities. We have considered all these feature sets when developing our evaluation framework.

### Iv-B Evaluation Framework

This section presents the evaluation framework we used to review the IoT products in context-aware perspective. We developed this evaluation framework based on the widely recognized and cited research done by Abowd et al. [5]. In this evaluation, we apply one and half decade old context aware theories into IoT era. Our evaluation is mainly based on three context-aware features in high-level: 1) context-aware selection and presentation, 2) context-aware execution, and 3) context-aware-tagging. However, we have also enriched the evaluation framework by identifying sub-features under above mentioned three features. Our evaluation framework consists of nine (9) features.

The Figure 6 visualizes how data is being collected transferred, processed, context discovered and annotated in typical IoT solutions. It is important to note that not all solutions may use the exact same data flow. Each solution may use part of the architecture in their solution. We will refer to this common data flow architecture during this paper to demonstrate how each solution may design their data flows. Our objective is to identify major strategies that are used by IoT products to offer context-aware functionalities. From here onwards, we explain the taxonomy, the evaluation framework, used to evaluate the IoT products. The results of the evaluation are presented in Table II. Summary of the evaluation framework is presented in Table I.

First we introduce the name of the IoT solution in the column (1) in Table II. We also provide the web page link of the each product / solution. It is important to note that, these products does not have any related academic publication. Therefore, we believe that web page links are the most reliable reference to a given IoT solution. Such links allow readers to follow further reading by using the product name along with web link.

In column (2), we classify each product into five categories. Each category is denoted by a different colour: red   (smart city), yellow   (smart environment), blue   (smart enterprise), green   (smart wearable), and purple   (smart home). Some solutions may belong to multiple categories. We divide the rest of the columns into three section : Context-aware Tagging, Context Selection and Presentation, and Context execution.

#### Iv-B1 Context-aware Tagging Section

Context-aware tagging, which is also called context augmentation and annotation represent the idea of sensing the environment and collecting primary context information. We also believe that secondary context generation is also a part of context-aware tagging feature. Primary context is any information retrieved without using existing context and without performing any kind of sensor data fusion operations [2]. For example, SenseAware (senseaware.com) is a solution developed to support real-time shipment tracking. As illustrated in Figure 7, SenseAware collects and processes context information such as location, temperature, light, relative humidity and biometric pressure in order to enhance the visibility and transparency of the supply chain. SenseAware uses both hardware and software components in their sensor-based logistic solution. such data collection allows different parties engage in supply chain to monitor the movement of goods in real-time and accurately know the quality of the transported goods and plan their processes effectively and efficiently. We list commonly acquired primary context information in column (3) in Table II.

Secondary context is any information that can be computed using primary context. The secondary context can be computed by using sensor data fusion operations or data retrieval operations such as web service calls (e.g. identify the distance between two sensors by applying sensor data fusion operations on two raw GPS sensor values). Further, retrieved context such as phone numbers, addresses, email addresses, birthdays, list of friends from a contact information provider based on a personal identity as the primary context can also be identified as secondary context. For example, Mimo (mimobaby.com) has built a smart nursery system, where parents learn new insights about their baby through connected products like the Mimo Smart Baby Monitor. In this product, turtle is the device that collects all primary context information. Then the data is transferred to an intermediary devices called lilypad. Such responsibility offloading strategy allows to reduce the turtle’s weight at minimum level and to increase the battery life. the communication and processing capabilities are offloaded to the lilypad device which can be easiy recharged when necessary. We can see Mimo Smart Baby Monitor usees some parts of the data flow architecture we presented in Figure II. User interface provided by Mimo and the data flow within the solution is presented in Figure 8. Cloud services [34] performs the additional processing and summarised data is pushed to the mobile devices for context presentation. In the user interface, parents are presented mostly the secondary context information such as baby movement or baby’s sleeping status. Accelerometer sensors are used to discover such secondary context information using pattern recognition techniques. We list secondary context information generated by IoT solutions in column (4) in Table II.

#### Iv-B2 Context Selection and Presentation Section

There are number of commonly used strategies, by most of the IoT solutions in the marketplace, to present context to the users.

Most of the IoT products use some kind of visualization techniques to present context information the users. We call this visual presentation. For example, Fitbit (fitbit.com) is a device that can be worn on multiple body parts in order to tracks steps taken, stairs climbed, calories burned, and hours slept, distance travelled, quality of sleep. This device collects data and present it to the users through mobile devices and web interfaces. Figure 9 illustrates the context presentation of Fitbit. Variety of different charts, graphs, icons and other types of graphical elements are heavily used to summarise and present analysed meaningful actionable data to the users. such visualization strategies are commonly encouraged in human computer interaction domain specially due to the fact that ’a picture is worth a thousand words’. We denote the presence of virtual presentation related to each IoT product using (✓) in column (5) in Table II.

IoT solutions in the market place also employ different commonly used devices to present the context to the users. Typically, an IoT solution offers context presentation and selection via some kind of software application. Some of the commonly used presentation channels are web-based (W), mobile-based (M), desktop-based (D), and objects-based (O). First, three mediums describes themselves. Object-based means that context selection and presentation is done through a custom IoT device itself. Sample IoT solutions that use object-base presentation strategy are presented in Figure 10. We identify the presence of different presentation channels related to each IoT product in column (6) in Table II.

In addition to the context presentation channels, IoT solutions use number of user interaction mechanisms such as voice (V), gesture (G), touch (T). Over last few years, we have seen more and more voice activated IoT solutions are coming to the marketplace. For example, latest technological development such as natural language processing and semantic technologies have enabled the wide use of voice activated IoT solutions. Amazon Echo (amazon.com/oc/echo), Ubi (theubi.com) are two voice activated personnel assistant solutions. Typically, they are capable of answering user queries related to weather, maps, traffic and so on (i.e. commonly asked questions). They are designed to learn from user interactions and customize their services and predictive models based on the user behaviour and preferences. These solutions have gone beyond what typical smart phone assistants such as Google One, Microsoft Cortana, Apple Siri has to offer. For example, Ubi has the cabability to interact with other smart objects in the smart house environment.

More importantly products such as Ivee (helloivee.com), a voice controlled hub for smart homes, facilitates interoperability over the other IoT products in the markets. This means that consumers can use Ivee to control other IoT products Iris (irissmarthome.com), Nest (nest.com), Philips Hue (meethue.com), SmartThings (smartthings.com), and Belkin WeMo (belkin.com). We discuss interoperability matters in details in Section VI. In addition to centralizes home hubs based IoT systems, more and more standalone IoT products also support voice-activated interaction such as executing commands. For example, VOCCA (voccalight.com) is a plug & play voice activated light bulb adapter requires no WiFi, no set-up, no installation.

Gesture has also been used to enable the interactions between IoT products and users. For example Myo (thalmic.com/en/myo/) is a wearable armband that can be used to issues gesture base commands. Myo reads gestures and motion and let hte users to seamlessly control smart phones, presentations, and so on. Nod (hellonod.com) is the a advanced gesture control ring. It allows users to engage objects with user movements. Nod can be considered as a universal controller, allowing effortless communication with all of the smart devices in users connected life, including phones, tablets, Google Glass, watches, home appliances, TVs, computers and more. We identify the presence of different user interaction mechanisms related to each IoT product in column (7) in Table II.

IoT solutions process data in different locations in their data communication flow as shown in Figure 6. Sometimes data is processed within the sensors or the local processing devices. In other circumstances, data is sent to the cloud for processing. Deepening the applications and functionalities each IoT solution tries to provide, data may be processed in real-time (RT) or later (A). Specially, event detection based IoT systems need to act in real-time which requires real-time processing. For example, IoT solutions such as Mimo smart baby monitor performs data processing in real-time as their mission is to increase the health and safety of the toddlers. It is also important to note that not every solution requires data archival. For example, health and fitness related IoT products can be benefited from archiving historic data. Such archives data will allow to produce graphs and charts over time and provide more insights and recommendations t the consumers. More data also facilitates more accurate prediction. However. storing more data cost more and not every solution requires such storage. ShutterEaze (shuttereaze.com) makes it easy for anyone to add remote control functionality and automate their existing interior plantation shutters. For example, IoT product like this will not necessarily be benefited by archiving historic data. Still it can learn user behaviour over time (based on how users use the product), and automate the task without storing data. We identify the usage of real-time and archival techniques in column (8) in Table II.

IoT solutions mainly use three different reaction mechanisms. Most commonly used mechanism is notification (N). This means that when a certain condition is met, IoT solution will release a notification to the users explaining the context. For example, Mimo (mimobaby.com), the baby monitoring product we mentioned earlier, notifies the parents when the baby shows any abnormal movements or breathing patterns. Parent will receive the notification through their smart phone. Some IoT solutions may react by performing actuations (A). For example, Blossom (myblossom.com) ia a smart watering products that can be self-programmed based on real-time weather data and gives the user control over the phone, lowering the water bill up to 30%. In this kind of scenario, the product may autonomously perform the actuations (i.e. open and close sprinklers) based on the context information. Another reaction mechanism used by IoT solutions is providing recommendations (R). For example, MAID (maidoven.com) has a personalization engine that continuously learns about the users. MAID learns what users cook regularly, tracks users activity using data from smart phones and smart watches. Then, it will provide recommendations for a healthy balanced diet. MAID also recommends users to workout or to go for a run based on the calories they consume each day. We identify the usage of reaction mechanisms related to each IoT product in column (9) in Table II.

Another important factor we identified during the product review is the learn-ability. Some products are capable of recording user provided inputs and other autonomously gathered information to predict future behaviours. In computer science, such behaviour is identified as machine learning (ML). For example, Nest (nest.com) thermostat is capable of learning users’ schedules and the temperatures users prefer. It keeps users comfortable and saves energy when they are away. In contrast, products such as Fibaro (fibaro.com) requires users to explicitly defines (UD) event thresholds and triggers as shown in Figure 11. We review the learn-ability of each IoT product in column (10) in Table II.

There are number of different ways that an IoT product would trigger a certain reaction. It is important to note that a single IoT solution may combine multiple triggers together in order to facilitate complex requirements. Some rigger may be spacial (S), temporal (T), or event based (E). Event based triggers are the most commonly used mechanism. For example, the IoT products such as SmartThings (smartthings.com), Ninja Blocks (ninjablocks.com), Fibaro (fibaro.com), Twine (supermechanical.com) allow users to define contextual triggers using sensors, actuators and parameters. Figure 11 and Figure 12 shows how two different products define events.

Low powered bluetooth beacons are commonly used in IoT products, specially in commercial and retail sector for both localization and location-based advertising [35]. For example, XY (xyfindit.com) and Estimote (estimote.com) are two similar products in the IoT marketplace that provide small beacons that can be attached to any location or object. The beacons will broadcast tiny radio signals which smart phones can receive and interpret, unlocking micro-location and contextual awareness. Therefore, IoT products may trigger a reaction when either users entering into or going out from a certain area. There are some other products such as FiLIP (myfilip.com) which users location-aware triggers to make sure children are staying within safe area. FiLIP uses a unique blend of GPS, GSM, and WiFi to allow parents to locate their child using the most accurate location information, both indoors and outdoors. Parents can create a virtual radius around a location, such as home, school or a friend’s house. Further, parents can set up to five such safe zones using the FiLIP app. A notification will be sent to the parent’s smart phone when FiLIP detects that the child has entered or left a safe zone.

In temporal mechanism, trigger is release based on a time schedule. Temporal triggers may refer to time as time of the day (e.g. exactly: 10.30 am or approximately: morning), day of the week (e.g. Monday or weekend), week of the month (e.g. second week), month of the year (e.g January), season (e.g. winter). Figure 11 show how Fibaro system allows to define a trigger by incorporating temporal triggers. IoT products such as Nest thermostat also use temporal triggers to efficiently learn and manager energy consumption.

## V Review of IoT Solutions

In this section we evaluated variety of different IoT solutions in the marketplace based on the evaluation framework presented in the earlier section. Table I, summarises the evaluation framework used and Table II presents the IoT product review results.

## Vi Lessons Learned, Opportunities and Challenges

This section presents some major lessons we learnt during the IoT product review.

### Vi-a Trends and Opportunities

According to our survey on the IoT product marketplace, it is evident that the types of primary context information collected through sensors are mostly limited. However, the ways such collected data is been processed varied significantly based on the application and the required functionalities that the IoT product plan to offer. Therefore, it is important to understand that, in IoT, same data can be used to derive different insights in different domain. In combine, the IoT solutions have used around 30-40 different types of sensors to measure different parameters. The ability to derive different insights using same set of data validates the importance of sensing as a service model [8], which envisions to create a data market that buys and sells data.

Most of the IoT solutions have used some kind of context presentation technique that summarizes and converts the data into a easily understandable format. It is also important to note that, despite the advances in human computer interaction, most of the IoT solutions have only employed traditional computer screen-based technique. Only few IoT solutions really allow voice or object-based direct communications. However, most of the wearable solutions use touch as a common interaction technique. We also see a trend of smart home products also increasingly use touch-based interactions. Hands free voice or gesture based user interaction will help consumers to seamlessly integrate IoT products into their lives. At least, smart watches and glasses may help to reduce the distraction that smart phones may create when interacting with IoT products.

Most of the IoT products ends their services after releasing notification to the consumers. Users will need to perform the appropriate actuation tasks manually. Lack of standards in machine to machine (M2M) communication seems to play a significant role in this matter. We will discuss this issue in Section VI-C. Finally, it is important to note that increasing number of IoT products use data analytic and reasoning in order to embed more intelligence to their products. As a result, there is a need for domain independent, easy to use (e.g. drag and drop configuration without any program coding) analytical frameworks with different characteristics where some may effectively perform on the cloud and the others may work efficiently in resource constrained devices. One solution in this space is Microsoft Azure Machine. Learning. Another generic framework is Wit. Wit (Wit.ai) is a natural language processing API for the IoT which allows developers to easily and quickly add natural language processing functionality to their IoT solutions.

It is important to note that most of the IoT solutions consider families or group of people as a whole, not as individuals. Therefore, most of the IoT solutions are unable to individually and separately identify father, mother or child living in a given house. For example, the temperature that individual family members would like to have can be different. However, most of the modern thermostat only consider context information such as past behaviour, time of the day, presence of a user, and so on. However, it cannot handle individual preferences of the family members. Therefore, embedding such capabilities to the IoT products would be a critical requirement to be successful in future IoT marketplace.

In order to support and encourage the adoption of IoT solution among consumers, it is important to make sure that the usage of products allows to recover the cost of product purchase within a reasonable time period. For example, the Nest thermostat promises that consumers can recover its costs through reducing the energy bill. Auto-Schedule feature in Nest makes it easy to create an energy efficient schedule that help the users to save up to 20% on heating and cooling bills.

### Vi-B Product Prototyping

There are number of do-it-yourself (DIY) prototyping platforms available that allows to create IoT prototypes quickly and easily. Specially these platforms are cheaper and modular in nature. They allow anyone with a new idea to test their initial thoughts with very limited budget, resources, and more importantly less time. Arduino (arduino.cc) (including variations such as Libelium (libelium.com)), .NET Gargeteer (netmf.com/gadgeteer), LittleBits (littlebits.cc) are some well known prototyping platforms. Most of these products are open source in nature. More importantly over the last few years, they have become more interoperable which allows product designers to combine different prototyping platforms together. The programming mechanisms use to program these modules can be varied (e.g. C, C++, C#, Java, Javascript, etc.). Some platforms provide easy and intuitive ways to write program such as mashing-ups and wirings as shown in Figure 13.

There are small computer systems been developed to support IoT prototyping. For example Raspberry Pi (www.raspberrypi.org) is a such product. Raspberry Pi is a credit card-sized single-board computer developed in the UK by the Raspberry Pi Foundation with the intention of promoting the teaching of basic computer science in schools. However, more recently, Raspberry Pis are heavily used in IoT product prototype development. For example, IoT products such as NinjaBlocks (ninjablocks.com) has used Raspberry Pis in their production officially. Further, most of the platforms such as Ardunio can successfully work with Raspberry Pi Computers. Recently, Intel has also produced a small computer (e.g. Intel Galileo and Intel Edison boards) competitive to Raspberry Pi which runs both windows and Linux. The Intel Edison is a tiny computer offered by Intel as a development system for wearable devices.

Programming IDE tools such Microsoft Visual Studio provides significant support for IoT program development by facilitating visual wiring, mash ups and automated code generation. Such ease of programming and prototyping abilities have attracted significant attention from hobbyist, researcher, and even from school children.

These modular based prototyping tools allow to build and test context-aware functionalities efficiently and effectively. Most of these platforms offer large number of sensing modules that allow to collect data from different types sensors. As we mentioned earlier such data can be considered as primary context. Therefore, such primary context can be combined together to generate secondary context information. However, in most of the prototyping platforms, secondary context discovery needs to be done manually or using IF-ELSE statements. However, it would be much useful to develop a standard framework with modularity in mind to address this issue. These modules need to be defined in a standard form despite their differences in real implementations. Further, such context discovery modules should be able to combine together to discover more advance context information [36]. We further explain how such framework should work in real world in Section VI-D.

### Vi-C Interoperability on Product and Services

Interoperability is a critical factor to be successful in IoT domain. Consumers typically do not want to stick into one single manufacturer or service provider. They always go for their preferences and for the factor which are more important to them such as cost, look and feel, customer service, functionality and so on. Interoperability among different IoT products and solutions allows consumers to move from one product to another or combine multiple products and services to build their smart environments as they like in a customize fashion. Further, interoperability [37] is also important to eliminate market domination of large companies that increase the entry barriers for the small IoT product and service providers.

In IoT market place, interoperability is mainly achieved using three methods: 1) partnerships among product and service developers, 2) open and close standards, and 3) adaptors and mediator services. We have seen that major industrial players in the IoT marketplace stablish strategic partnerships with each other in order to enable interoperability among their product and services. However, this is not a scalable strategy to widely enable interoperability among IoT devices. Similarly, large corporations such as Apple (e.g. HomeKit222developer.apple.com/homekit, HealthKit333developer.apple.com/healthkit) and Google (e.g. Fit444developers.google.com/fit) are also attempting to build their own standards and interoperability certifications. This kind of interoperability may lead to corporate domination of IoT marketplace which could also hinder the innovation by small, medium, and start-up companies.

To address the interoperability, there are some alliance have been initiated. For example AllSeen Alliance (allseenalliance.org) has been created to promote some kind of interoperability among IoT consumer brands. AllSeen has developed a standard software platform called AllJoyn. AllJoyn is a system that allows devices to advertise and share their abilities with other devices around them. A simple example would be a motion sensor letting a light bulb know no one is in the room it is lighting. This is the ideal approach the interoperability among IoT products. However, security [38] and privacy in this framework need to be strengthen to avoid using interoperability features to attack IoT products by hackers or evil parties.

Another approach to enable interoperability among different IoT solutions is through adapter services. For example, IFTTT (ifttt.com), If This Then That, is a web based service that allows users to create powerful connections, chains of simple conditional statements. One simple statement is illustrated in Figure 14. Channels are the basic building blocks of IFTTT. Each Channel has its own Triggers and Actions. Some example Channels could be Facebook, Twitter, weather, Android Wear, and so on. Channel could be both hardware or software. Service providers and product manufactures need to register their services with IFTTT once. After that anyone interested ca use that product or service as a channel to compose any recipe. Example list of channels are listed here: ifttt.com/channels. Personal recipes are combinations of a Trigger and an Action from active Channels. Example recipes are shown in Figure 14. For example, first recipe is defined to send a twitter message to a family member when the user reaches home. This kind of recipe can be used to offload responsibility from a child so the system automatically act on behalf of the child and sent a tweet to their parents. Context-aware recommendation can also help users to quickly configure channels in IFTTT. Context could be location, time, family members around, IoT products located near by and so on. Context-aware recommendation [39] can also be done by analysing similar users with similar smart environments.

### Vi-D Resources and Energy Management

Most popular approach of energy management in IoT is through smart plugs. Plugwise (shop.plugwise.com), Thinkecoinc (shop.thinkecoinc.com), Belkin (www.belkin.com) provide similar functionalities and services where they capture energy consumption using smart plugs. These solutions analyse data in many different ways and presented the context information to the users using variety of different charts and graphs. These plugs can also be used to home automation as they can be switched ON and OFF remotely or conditionally. For example, a condition would be temporal (i.e. time-aware behaviour) or spatial (i.e. location-aware behaviour).

There aren’t any IoT solutions that focus on planning or deployment stages of smart environments. Analyse energy consumption is important in both industrial large scale deployments (e.g. waste management solutions discussed in [8]) and in consumer based smart home and office deployments. Lets consider a smart home office planing and deployment scenario. At the moment, IoT marketplace is flooded with large number of IoT smart products that offer different functionalities. However, there aren’t any method for consumers to measure or compare the benefits these products may offer and the associated costs such as cost of purchase, installation and maintains. Further, it is very hard to understand which solutions can work together and complement each other and which work standalone.

It is also difficult to understand where to install certain smart products and how many products are required to cover a certain area. (e.g. what are the ideal locations to install micro-climate sensors within a building which enable to accurately identify the micro-climate behaviour). Another issue would be to determine the coverage of a product. For example, how many motion sensors are required for a given home or office. Currently, to best of our knowledge, there is no such tool that can be used to achieve above planning and installation tasks. As we mentioned before, consumers are always eager to know the costs and benefits of a products. Therefore, it is important to facilitate some tools that can demonstrate cost benefit analysis (e.g. purchase cost, maintenance cost such as energy, energy saving and so on.). Context information will play a significant role in this kind of tools where consumers may need to input the budget, size of the building, their priorities and expectations. The tool will need to make recommendations to the consumers on which product to buy based on the product’s technical specification and other consumers’ reviews and comments.

The planing and installation becomes much more critical in industrial settings. Let considers the agricultural sensing scenario, the Phenonet project, presented in [40]. Phenonet describes the network of sensors collecting information over a field of experimental crops. Researchers at the High Resolution Plant Phenomics Centre [41] needs to monitor plant growth and performance information under different climate conditions over time.

It would be very valuable to have a tool that can help planning large scale sensor deployments. For example, energy predictive models will help the users to decide what kind of energy sources to be used and what kind of battery size to be used in each scenario. The amount of sensor nodes require to cover a curtain geographical area should be able to accurately predicted based on the context information using such tool. For example, in the agricultural sensing scenario, sensors deployments are planned by agricultural scientist who have little knowledge on electronic, communication, or energy consumption. Therefore, it is useful to have a user friendly tool that enables them to plot and visualise a large scale sensor deployment in virtual setting before getting into real world deployments. Perera et al. [40] have present the agriculture scenario in detail.

Context information plays a critical role in sensor configuration in large scale sensor deployments in IoT. The objective of collecting sensor data is to understand the environment better by fusing and reasoning them. In order to accomplish this task, sensor data needs to be collected in a timely and location-sensitive manner. Each sensor needs to be configured by considering context information. Let us consider a scenario related to smart agriculture to understand why context matters in sensor configuration. Severe frosts and heat events can have a devastating effect on crops. Flowering time is critical for cereal crops and a frost event could damage the flowering mechanism of the plant. However, the ideal sampling rate could vary depending on both the season of the year and the time of day. For example, a higher sampling rate is necessary during the winter and the night. In contrast, lower sampling would be sufficient during summer and daytime. On the other hand, some reasoning approaches may require multiple sensor data readings. For example, a frost event can be detected by fusing air temperature, soil temperature, and humidity data. However, if the air temperature sensor stops sensing due to a malfunction, there is no value in sensing humidity, because frost events cannot be detected without temperature. In such circumstances, configuring the humidity sensor to sleep is ideal until the temperature sensor is replaced and starts sensing again. Such intelligent (re-)configuration can save energy by eliminating ineffectual sensing and network communication.

An ideal tool should be able to simulate different types of user scenarios virtually before the real world deployments begin. Once deployed, another set of tools are required to advice and recommend, scientists and non-technical users, on configuring sensor parameters. Configuring sensors in a optimal fashion would lead to longer operation time while maintaining required accuracy. It is important to develop the tools in a modular and standard fashion so the manufacturers of each IoT solution can add their products into a library of product which enables consumers to easily select (may be drag and drop and visualize) the product they prefer for visualization purposes. Further, such tools will need to be able to combine different compatible products together autonomously based on context information such as budget, user preferences, and location information so the users will be offered different combinations to select from.

Resource management is also a critical task that need to be done optimally in IoT domain. Previously, we discussed how data may transferred over the network as well as through different types of data processing devices in Figure 6. It is hard to determine the optimal location555the device that is responsible for processing data to process data. Therefore, it is ideal to have a tool that is capable of evaluating a given software component666A self contained algorithm that may take primary context information as inputs and outputs secondary context information using any kind of data reasoning technique [2]. against a given computational network architecture and deciding which location is optimal to conduct any kind of reasoning based on user preferences, resource availability, context information availability, network communication availability and so on.

### Vi-E Privacy and Data Analytic

IoT marketplace is mainly composed with three parties, namely: device manufacturers, IoT cloud services and platform providers, and third party application developers [15]. All these parities need to consider privacy as a serious requirement and a challenge. In this section, we present some advice on preserving user privacy in IoT domain.

Device Manufacturers: Device manufactures must embed privacy preserving techniques into their devices. Specially, manufactures must implement secure storage, data deletion, and control access mechanisms at the firmware level. Manufactures must also inform consumers about the type of data that are collected by the devices. Moreover, they must also explain what kind of data processing will be employed and how and when data would be extracted out of the devices. Next, the manufactures must also provide the necessary control for the consumers to disable any hardware components. For example, in an IoT security solution, consumers may prefer to disable the outside CCTV cameras when inside the home. However, consumers will prefer to keep both inside and outside cameras active when they leave the premises. Moreover, devices manufactures may also need to provide programming interface for third party developers to acquire data from the devices.

IoT Cloud Services and Platform Providers: It is likely that most of the IoT solutions will have a cloud based service that is responsible for proving advance data analysis support for the local software platforms. It is very critical that such cloud providers use common standards, so that the consumers have a choice to decide which provider to use. Users must be able to seamlessly delete and move data from one provider to another over time. Such a possibility can only be achieved by following a common set of interfaces and data formats. Most of the cloud services will also use local software and hardware gateways such as mobile phones that act as intermediary controllers. Such devices can be used to encrypt data locally to improved security and to process and filter data locally to reduce the amount of data send to the cloud. Such methods will reduce the possibility of user privacy violation that can occur during the data transmission.

Third Part Application Developers: Application developers have the responsibility to certify their apps to ensure that they do not contain any malware. Moreover, it is the developers’ responsibility to ensure that they present clear and accurate information to the users to acquire explicit user consent. Some critical information are: (1) the task that the app performs, (2) the required data to accomplish the tasks, (3) hardware and software sensors employed, (4) kind of aggregation and data analysis techniques that the app will employ, (5) kind of knowledge that the app will derive by data processing. Users need to be presented with a list of features that the application provides, and the authorization that the user needs to give to activate each of those features. The control must be given to the user to decide which feature they want to activate. Moreover, in the IoT, acquiring user consent should be a continuous and ongoing process. Consequently, the application developers must continuously allow the users to withdraw, grant, or change their consent. Moreover, users must be given full access to the data collected by the IoT devices.

### Vi-F Central Hubs

Central hubs are commonly used in IoT solutions. A typical IoT solution may comprises a number of different components. For example, an IoT solution may have sensors, actuators, processing and communication devices. Due to the nature, sensors and actuators may need to deploy in certain location manner (e.g. door sensor must mount on the door). As a result such sensors and actuators need to be small in size. Due to miniature size, it is not possible to enrich them with significant computational capacity. Similarly, most of the time these sensors and actuators would be battery powers (i.e. without having connected to permanent power sources). Therefore, energy management within those sensors and actuators is very critical. As a result, such smaller devices cannot perform significant data processing tasks. On the other hand, these individual devices have only limited knowledge about a given context. For example, a door sensors may only know about the current status of the door. The knowledge that can be derived from such limited amount of data is very constrained. In order to comprehensively understand a given situation, context data from number of sensors and actuators need to be collected, processed, and analysed. To address this issue, most of the IoT solutions have been used a central hubs (sometimes called ‘home hub’) or similar solutions as shown in Figure 15.

Typically, central hubs are larger in size compared to sensors and actuators. Further, they are capable of communicating using multiple wireless protocols such as WiFi, WiFi-direct, Bluetooth ZigBee, Z-wave, and so on. They are also capable of storing data for a significant time period. Typically, only one central hub is required for a large area (e.g. house). These hubs may perform data processing and reasoning tasks (e.g. triggering IF-THEN rules). Further these hubs are typically connected to the cloud services. Dispite the differences in , in high-level, all of these hubs allows to add functionalities over time (i.e. extend the functionalities they may offer), through installing new applications. An app could be a IF-ELSE procedure that explain a certain contextual behaviour as illustrated in Figure 12.

The problem in this approach is that each IoT solution designers are eager to design their own centralized hub. Such design approach significantly reduces the interoperability among different products and services in the IoT marketplace. These hubs are tend to use custom firmware and software framework stacks. Unlike operating systems, they are mostly designed to run under specific hardware platforms and configurations. As a result, it makes harder for other IoT solutions to use or utilize other centralized hubs in the marketplace. Centralized hubs typically does not have any user interface. They are controlled and managed using smart phones, tablets, or computers.

In order to stimulate the adoption of IoT solution among consumers, it is important to design a common software platform using common set of standard. The current mobile app market is an ideal model for IoT domains as well where users may install different applications in order to enhance their existing IoT products. Verification is required to check whether the required hardware devices is available to support the intended software application. This is similar to the some mobile app stores validate the phone specification before pushing the each app to a smart phone. In comparison to mobile phone domain, IoT domain is slightly complex where hardware also play a significant role. A one possible solution is to use hardware adaptors. This means when a IoT product manufacture wants to design a product that is interoperable with a another hub in the IoT marketplace, they need to design a hardware adaptor that may handle the interoperability using two-way conversions.

Finally, it is also important to highlight the necessity of intermediation nodes that can perform multi-protocol communication, bridging short range protocols, and protocol conversions [42]. For example, sensors that may use Bluetooth and ZigBee which can only communicate very short distance. To accommodate such sensors, intermediary nodes may be required. The intermediate nodes may install throughout a given location which may use with log range protocols to communicate with the central hub. The intermediate nodes may use short rage protocols to communicate with sensors and actuators.

### Vi-G Legacy Devices

Most of the IoT products in the marketplace comes with own hardware components and software stacks. However, we have increasingly seen that IoT solutions attempt to enrich legacy devices with smart capabilities. One very popular solution is Nest (nest.com) thermostat. It has the capability to learn from users over time about their behaviour and preferences and control the temperature more efficiently and pro-actively. This thermostat can be installed by replacing the existing non-smart traditional thermostats. Everything else connected to the heating systems would work seamlessly. ShutterEaze (shuttereaze.com) is another example for enriching legacy devices. This example is more into home automation. ShutterEaze makes it easy for anyone to add remote control functionality and automate their existing interior plantation shutters. No shutters changing is required.

A slightly different example is Leeo (leeo.com). As illustrated in Figure 16, Leeo keeps track of smoke alarms, carbon monoxide alarms, and the climate in home. If something is not right, it sends notifications straight to the users phone. It is important to note that, there is no communication between the legacy smoke detection devices / alarms and the Leeo device. They are completely two different systems without any dependencies. Leeo get triggered by the sound that may produce by other traditional alarms. This is a very good examples to demonstrate how to embed smartness to our homes without replacing existing legacy systems. More importantly, any kind of replacing cost a significant amount to the consumers. This kind of solutions eliminates such unnecessary and extra costs that may put consumers away from adopting IoT solutions. The lesson we can learn is that if the legacy devices cannot understand the context it operates and act intelligently, the new devices can be incorporated to embed smartness to the overall system where new devices helps to mitigate the weaknesses in the legacy devices.

## Vii Concluding Remarks

In this survey, we reviewed significant number of IoT solutions in the industry marketplace from context-aware computing perspective. We briefly highlighted the evolution of context-aware technologies and how they have become increasingly popular and critical in today’s applications. First, we reviewed number of IoT products in order to identify context-aware features they support. At the same time, we also categorized the IoT solutions in the market into five different segments: smart wearable, smart home, smart city, smart environment, and smart enterprise. Finally, we identified and discussed seven major lessons learned and opportunities for future research and development in context-aware computing domain. Our ultimate goal is to build a foundation that helps us to understand what has happened in the IoT marketplace in the past so we can plan for the future more efficiently and effectively.

## References

• [1] L. Atzori, A. Iera, and G. Morabito, “The internet of things: A survey,” Comput. Netw., vol. 54, no. 15, pp. 2787–2805, Oct. 2010. [Online]. Available: http://dx.doi.org/10.1016/j.comnet.2010.05.010
• [2] C. Perera, A. Zaslavsky, P. Christen, and D. Georgakopoulos, “Context aware computing for the internet of things: A survey,” Communications Surveys Tutorials, IEEE, vol. 16, no. 1, pp. 414–454, 2013.
• [3] A. Zaslavsky, C. Perera, and D. Georgakopoulos, “Sensing as a service and big data,” in International Conference on Advances in Cloud Computing (ACC-2012), Bangalore, India, July 2012.
• [4] H. Sundmaeker, P. Guillemin, P. Friess, and S. Woelffle, “Vision and challenges for realising the internet of things,” European Commission Information Society and Media, Tech. Rep., March 2010, http://www.internet-of-things-research.eu/pdf/IoT_Clusterbook_March_2010.pdf [Accessed on: 2011-10-10].
• [5] G. D. Abowd, A. K. Dey, P. J. Brown, N. Davies, M. Smith, and P. Steggles, “Towards a better understanding of context and context-awareness,” in Proceedings of the 1st international symposium on Handheld and Ubiquitous Computing, ser. HUC ’99.   London, UK: Springer-Verlag, 1999, pp. 304–307. [Online]. Available: http://dl.acm.org/citation.cfm?id=647985.743843
• [6] G. Kortuem, F. Kawsar, D. Fitton, and V. Sundramoorthy, “Smart objects as building blocks for the internet of things,” Internet Computing, IEEE, vol. 14, no. 1, pp. 44 –51, jan.-feb. 2010. [Online]. Available: http://dx.doi.org/10.1109/MIC.2009.143
• [7] L. Atzori, A. Iera, and G. Morabito, “From smart objects to social objects: The next evolutionary step of the internet of things,” Communications Magazine, IEEE, vol. 52, no. 1, pp. 97–105, January 2014.
• [8] C. Perera, A. Zaslavsky, P. Christen, and D. Georgakopoulos, “Sensing as a service model for smart cities supported by internet of things,” Transactions on Emerging Telecommunications Technologies (ETT), vol. 25, no. 1, pp. 81–93, 2014.
• [9] E. Welbourne, L. Battle, G. Cole, K. Gould, K. Rector, S. Raymer, M. Balazinska, and G. Borriello, “Building the internet of things using rfid: The rfid ecosystem experience,” Internet Computing, IEEE, vol. 13, no. 3, pp. 48–55, 2009.
• [10] A. Caragliu, C. D. Bo, and P. Nijkamp, “Smart cities in europe,” in 3rd Central European Conference in Regional Science-CERS, October 2009, pp. 45–59. [Online]. Available: http://www.cers.tuke.sk/cers2009/PDF/01_03_Nijkamp.pdf
• [11] A. Zanella, N. Bui, A. Castellani, L. Vangelista, and M. Zorzi, “Internet of things for smart cities,” Internet of Things Journal, IEEE, vol. PP, no. 99, pp. 1–1, 2014.
• [12] Libelium Comunicaciones Distribuidas S.L. , “WaspMote,” 2013, http://www.libelium.com/products/plug-sense [accessed on: 2013-08-13].
• [13] A. Weaver and B. Morrison, “Social networking,” Computer, vol. 41, no. 2, pp. 97–100, Feb 2008.
• [14] N. Lane, E. Miluzzo, H. Lu, D. Peebles, T. Choudhury, and A. Campbell, “A survey of mobile phone sensing,” Communications Magazine, IEEE, vol. 48, no. 9, pp. 140 –150, sept. 2010. [Online]. Available: http://dx.doi.org/10.1109/MCOM.2010.5560598
• [15] A. Ortiz, D. Hussein, S. Park, S. Han, and N. Crespi, “The cluster between internet of things and social networks: Review and research challenges,” Internet of Things Journal, IEEE, vol. 1, no. 3, pp. 206–215, June 2014.
• [16] A. Ahmed, T. Qiu, F. Xia, B. Jedari, and S. Abolfazli, “Event-based mobile social networks: Services, technologies, and applications,” Access, IEEE, vol. 2, pp. 500–513, 2014.
• [17] A. Campbell, S. Eisenman, N. Lane, E. Miluzzo, R. Peterson, H. Lu, X. Zheng, M. Musolesi, K. Fodor, and G.-S. Ahn, “The rise of people-centric sensing,” Internet Computing, IEEE, vol. 12, no. 4, pp. 12–21, July 2008.
• [18] A. K. Dey, G. D. Abowd, and D. Salber, “A conceptual framework and a toolkit for supporting the rapid prototyping of context-aware applications,” Hum.-Comput. Interact., vol. 16, pp. 97–166, December 2001. [Online]. Available: http://dx.doi.org/10.1207/S15327051HCI16234_02
• [19] B. Schilit and M. Theimer, “Disseminating active map information to mobile hosts,” Network, IEEE, vol. 8, no. 5, pp. 22 –32, sep/oct 1994. [Online]. Available: http://dx.doi.org/10.1109/65.313011
• [20] P. J. Brown, “The stick-e document: a framework for creating context-aware applications,” Electronic Publishing, vol. 8, no. 2 & 3, pp. 259–272, 1996. [Online]. Available: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.8.7472&rep=rep1&type=pdf
• [21] D. Franklin and J. Flachsbart, “All gadget and no representation makes jack a dull environment,” Proceedings of the AAAI 1998 Spring Symposium on Intelligent Environments, vol. Technical Report SS-98-02, pp. 155–160, 1998. [Online]. Available: http://infolab.northwestern.edu/media/papers/paper10072.pdf
• [22] T. Rodden, K. Chervest, N. Davies, and A. Dix, “Exploiting context in hci design for mobile systems,” in in Workshop on Human Computer Interaction with Mobile Devices, 1998. [Online]. Available: http://eprints.lancs.ac.uk/11619/
• [23] R. Hull, P. Neaves, and J. Bedford-Roberts, “Towards situated computing,” in Wearable Computers, 1997. Digest of Papers., First International Symposium on, oct 1997, pp. 146 –153. [Online]. Available: http://dx.doi.org/10.1109/ISWC.1997.629931
• [24] A. Ward, A. Jones, and A. Hopper, “A new location technique for the active office,” Personal Communications, IEEE, vol. 4, no. 5, pp. 42 –47, oct 1997. [Online]. Available: http://dx.doi.org/10.1109/98.626982
• [25] G. D. Abowd and E. D. Mynatt, “Charting past, present, and future research in ubiquitous computing,” ACM Trans. Comput.-Hum. Interact., vol. 7, pp. 29–58, March 2000. [Online]. Available: http://doi.acm.org/10.1145/344949.344988
• [26] B. Schilit, N. Adams, and R. Want, “Context-aware computing applications,” in Mobile Computing Systems and Applications, 1994. Proceedings., Workshop on, dec 1994, pp. 85 –90. [Online]. Available: http://dx.doi.org/10.1109/MCSA.1994.512740
• [27] J. Pascoe, “Adding generic contextual capabilities to wearable computers,” in Wearable Computers, 1998. Digest of Papers. Second International Symposium on, oct 1998, pp. 92 –99. [Online]. Available: http://dx.doi.org/10.1109/ISWC.1998.729534
• [28] N. S. Ryan, J. Pascoe, and D. R. Morse, “Enhanced reality fieldwork: the context-aware archaeological assistant,” in Computer Applications in Archaeology 1997, ser. British Archaeological Reports, V. Gaffney, M. van Leusen, and S. Exxon, Eds.   Oxford: Tempus Reparatum, October 1998. [Online]. Available: http://www.cs.kent.ac.uk/pubs/1998/616
• [29] Carnot Institutes, “Smart networked objects and internet of things,” Carnot Institutes’ Information Communication Technologies and Micro Nano Technologies alliance, White Paper, January 2011, http://www.internet-of-things-research.eu/pdf/IoT_Clusterbook_March_2010.pdf [Accessed on:2011-11-28].
• [30] A. Moses, “Lg smart fridge tells you what to buy, cook and eat,” The Sydney Morning Herald, January 2012, http://www.smh.com.au/digital-life/hometech/lg-smart-fridge-tells-you-what-to-buy-cook-and-eat-20120110-1ps9z.html [Accessed on: 2012-04-04].
• [31] M. Raskino, J. Fenn, and A. Linden, “Extracting value from the massively connected world of 2015,” Gartner Research, Tech. Rep., April 2005, http://www.gartner.com/resources/125900/125949/extracting_valu.pdf [Accessed on: 2011-08-20].
• [32] X.-W. Chen and X. Lin, “Big data deep learning: Challenges and perspectives,” Access, IEEE, vol. 2, pp. 514–525, 2014.
• [33] H. Hu, Y. Wen, T.-S. Chua, and X. Li, “Toward scalable systems for big data analytics: A technology tutorial,” Access, IEEE, vol. 2, pp. 652–687, 2014.
• [34] S. Cirani, L. Davoli, G. Ferrari, R. Leone, P. Medagliani, M. Picone, and L. Veltri, “A scalable and self-configuring architecture for service discovery in the internet of things,” Internet of Things Journal, IEEE, vol. 1, no. 5, pp. 508–521, Oct 2014.
• [35] E. Ovaska and J. Kuusijarvi, “Piecemeal development of intelligent applications for smart spaces,” Access, IEEE, vol. 2, pp. 199–214, 2014.
• [36] F. Xia, N. Asabere, A. Ahmed, J. Li, and X. Kong, “Mobile multimedia recommendation in smart communities: A survey,” Access, IEEE, vol. 1, pp. 606–624, 2013.
• [37] J. Kiljander, A. D’elia, F. Morandi, P. Hyttinen, J. Takalo-Mattila, A. Ylisaukko-Oja, J.-P. Soininen, and T. Cinotti, “Semantic interoperability architecture for pervasive computing and internet of things,” Access, IEEE, vol. 2, pp. 856–873, 2014.
• [38] S. L. Keoh, S. Kumar, and H. Tschofenig, “Securing the internet of things: A standardization perspective,” Internet of Things Journal, IEEE, vol. 1, no. 3, pp. 265–275, June 2014.
• [39] G. Dror, N. Koenigstein, and Y. Koren, “Web-scale media recommendation systems,” Proceedings of the IEEE, vol. 100, no. 9, pp. 2722–2736, Sept 2012.
• [40] C. Perera, P. P. Jayaraman, A. Zaslavsky, P. Christen, and D. Georgakopoulos, “Mosden: An internet of things middleware for resource constrained mobile devices,” in 47th Hawaii International Conference on System Sciences (HICSS), Kona, Hawaii, USA, January 2014, p. n/a.
• [41] CSIRO, “The high resolution plant phenomics centre,” June 2011, http://www.csiro.au/Outcomes/Food-and-Agriculture/HRPPC.aspx[Accessed on: 2012-09-24].
• [42] M. Palattella, N. Accettura, X. Vilajosana, T. Watteyne, L. Grieco, G. Boggia, and M. Dohler, “Standardized protocol stack for the internet of (important) things,” Communications Surveys Tutorials, IEEE, vol. 15, no. 3, pp. 1389–1406, 2013.
You are adding the first comment!
How to quickly get a good reply:
• Give credit where it’s due by listing out the positive aspects of a paper before getting into which changes should be made.
• Be specific in your critique, and provide supporting evidence with appropriate references to substantiate general statements.
• Your comment should inspire ideas to flow and help the author improves the paper.

The better we are at sharing our knowledge with each other, the faster we move forward.
The feedback must be of minimum 40 characters and the title a minimum of 5 characters