Market-Oriented Information Trading in Internet of Things (IoT) for Smart Cities

Market-Oriented Information Trading in Internet of Things (IoT) for Smart Cities

Yang Zhang, Member, IEEE, Zehui Xiong, Student Member, IEEE,

Dusit Niyato, Fellow, IEEE, Ping Wang, Senior Member, IEEE,

and Zhu Han, Fellow, IEEE

Abstract

Internet of Things (IoT) technology is a fundamental infrastructure for information transmission and integration in smart city implementations to achieve effective fine-grained city management, efficient operations, and improved life quality of citizens. Smart-city IoT systems usually involve high volume and variety of information circulation. The information lifecycle also involves many parties, stakeholders, and entities such as individuals, businesses, and government agencies with their own objectives which needed to be incentivized properly. As such, recent studies have modeled smart-city IoT systems as a market, where information is treated as a commodity by market participants. In this work, we first present a general information-centric system architecture to analyze smart-city IoT systems. We then discuss features of market-oriented approaches in IoT, including market incentive, IoT service pattern, information freshness, and social impacts. System design chanlenges and related work are also reviewed. Finally, we optimize information trading in smart-city IoT systems, considering direct and indirect network externalities in a social domain.

I. INTRODUCTION

Smart city [1]–[3] applications aim to improve quality of lives of citizens, increase efficient public and private resource utilization, and reduce pollution, nuisance, and crime. Smart city introduces a unique requirement to Internet of Things (IoT) system designs. Specifically, IoT systems will be used by numerous and diverse smart-city applications and users with large quantities of data generated. This is significantly different from general IoT systems that are designed to support specific single-purpose applications. For example, GPS trace data collected from commuters’ smartphones and video images from city cameras can be jointly used by a government in city planning, public transportation operators in allocating and routing buses and trains, logistic businesses in optimizing package delivery, and the commuters themselves in trip planning. As such, information collected and used in the IoT systems can be treated as a commodity which can be traded among information producers, processors, sellers, and customers/users. Moreover, as smart-city applications mainly aim at ordinary users, e.g., governments and individual users, it is unnecessary to reveal too many IoT device layer details in the applications, such as structures and functionalities. Processed data and services are usually required to be delivered to users. Some practical scenarios and examples of IoT for smart-city applications are shown in Table I.

In this work, we mainly discuss a market-oriented vision of smart-city IoT systems with emerging information exchange and resource allocation techniques. In Section II, we present an information-centric layered architecture of smart-city Iot applications. We identify some unique features of a socio-technical paradigm that makes smart-city IoT systems different from conventional IoT from a “value of information” perspective in Section III. Typical design considerations with related work are also reviewed in Section IV. Next, in Section V, we propose a gametheoretic market model for information trading in smart-city IoT scenarios. Some numerical studies that show the smart-city stakeholders’ behaviors and benefits of the information trading are highlighted.

II. IOT ENABLED SMART CITY

A.                  IoT and Smart City

In conventional IoT applications, e.g., wireless sensor networks (WSNs), technical problems such as physical infrastructure, network architecture and communication techniques have been relatively well studied. However, as human and social factors are heavily involved in smart city applications, in recent studies, it is more suitable to consider smart city as a socio-technical system where the relationships among physical/virtual devices and social participants are driven by values and incentives. As such, market-oriented approaches can be employed as mechanisms to implement practical smart-city IoT systems.

B.                  IoT for Smart City: An Information-centric Architecture

Evidences have shown that smart-city IoT applications are data intensive [3]–[5] with the following features:

•                      Information quantity: A large amount of data and information need to be generated and transported in smart-city IoT applications.

•                      Information heterogeneity: Smart-city IoT systems collect and use various and diverse data from numerous sources with a variety of types, formats, and attributes. Moreover, the role of each participant changes regularly. These data is used jointly to achieve a certain goal.

*                      Information quality: Information exchanged in smart-city IoT systems may have different inherent quality. For example, data collected from an advanced smartphone typically has higher accuracy than that from a small sensor. Additionally, “age of information” is another important quality metric. The information which is promptly retrieved has higher quality than that which is delayed.

•   Consequently, instead of a network-centric architecture focusing on end-to-end information transmissions, an information-centric architecture becomes more promising for the design and deployment of smart-city IoT systems.

An information-centric reference system architecture for an IoT smart city is shown in Fig. 1.

The architecture has three layers with different functionality.

•                      Layer I, information sensing and generating layer: Components working in this layer can sense, collect and generate “raw” data either from the environment or within the network devices, e.g., surveillance records of city video cameras. The layer provides interfaces at the edge of smart-city IoT architecture and physical world.

•                      Layer II, information packaging and processing layer: Raw data from Layer I is processed in this layer and used to provide services. The services can be formed as structured and interpreted information bundles by selectively combining and employing the raw data passed to this layer, e.g., using machine learning algorithms.

•                      Layer III, information application layer: Information services in this layer are applied to serve the needs and demand of city users. For example, traffic forecast information services can be adopted by smart transportation users to predict traffic situation and plan their trips.

The information-centric smart-city IoT architecture allows cross layer implementations, and the key component is an information flow rather than device placement as in conventional IoT systems. This information-centric architecture promotes cross-layer designs and implementations. As many devices in smart-city systems have more computing capability, e.g., smartphones, primitive functions can be provided locally by each device (or a group of devices) to meet immediate information requirements of smart-city users. For example, noise can be removed from the sensed information by sophisticated algorithms running by the smartphones. In this regard, IoT devices defined as micro providers [2] can operate across Layers I and II, as shown in Fig. 1. The information-centric architecture can be employed in market-oriented modeling and analysis for information trading in smart-city IoT systems. With this architecture, entities in smart city can only concentrate on the content and timeliness while acquiring data from the system, instead of low-level communication issues.

III. MARKET-ORIENTED IOT INFORMATION TRADING FOR SMART CITIES

In this section, we present smart-city IoT systems from the “value of information and services” perspective. Information is treated as goods, and information transfer is considered to be market transactions. We propose the definition of market-oriented approaches for information trading in smart-city IoT systems. Some key components and mechanisms in a smart-city information market are discussed.

A. IoT for Smart City: A Value of Information and Services

A smart city is not a simple combination of sensors and sensor users. As a framework of social governance and business solution, technological and societal entities are all adopted in smart city IoT implementations as the core components and information sources. In particular, a smart city involves human participants in forms of individuals and social groups. In this regard, information generated in smart city systems has its inherent values to owners and other participants. Activities to handle and process the collected information can be seen as valueadding or value generation processes. Consequently, information becomes expensive in value when the scale of data sensing, collection and processing as well as the complexity of the smart city applications increase. Therefore, transfers of raw data and processed information should involve value interpretation/proposition and reward implication in terms of either monetary or other types of incentive as in market trading.

One of the incentive schemes in real smart-city IoT applications is the Travel Smart Rewards (TSR) program of Singapore^[1]. In the program, public transit card holders/users, i.e., the information sources, can register and share their travel records with the Land Transport Authority (LTA) of Singapore, i.e., the information processor, for research and public transit management uses, i.e., value-added services. Upon sharing the information, a cash reward is paid by LTA to the registered users. LTA can use such information to schedule trains and buses to reduce fuel consumption, improve waiting time, and driver and operator management, hence increasing operating efficiency and profit.

Evidently, smart-city IoT applications are more concerned about “value of information” than classical performance metrics such as the shortest route, delay and minimum energy consumption of data transmission [6]. Accordingly, we propose a general utility-based value model for marketoriented IoT smart-city information trading. The general form of utility of any system participant can be defined in the form of gross profit, i.e., Utility , Benefit − Cost. Note that in the utility formulation, the benefit can be the gain of the participant in terms of monetary or nominal rewards, depending on the application setting. Moreover, the benefit varies with different system components. Raw data may have a lower value compared with that of higher-level information services provided to users. For example, GPS trace is far cheaper than vehicle routing services for general users. Furthermore, identical raw data can be valued differently by different information processors, depending on their needs and demands. Again, a government agency and commercial map service may value GPS trace differently. Additionally, information received at different time instances leads to considerably diverse values due to timeliness of information. GPS trace which is delayed due to erroneous transmission typically has a low value, especially for real-time traffic prediction.

In this regard, information-centric architecture can well support market-oriented approaches for information trading in smart-city IoT systems. Information value and pricing are affected by the availability, accessibility and quality obtained by different types of users, which are the next topics of discussion.

B.                  Incentive, Pricing and Sensing-as-a-Service

Smart-city applications are complex systems with a variety of participants acting in their own best interests. In the conventional systems, there can be a centralized authority or provider to deploy IoT devices and collect data from them. For smart-city IoT systems, participants are selfinterested such as organizations, competitive service providers and human users with various demand. The participants thus need to be incentivized to cooperate and exchange information.

An efficient market-oriented tool to leverage incentives to manage network communication and data resources is named Smart Data Pricing [7], where dynamic pricing can be applied to network resources. The dynamic pricing approach can both indicate different values of the network resources, e.g., information, as well as exclude or deter some participants which may find the benefit is exceeded by the current price level. Nevertheless, pricing for IoT information exchange in a smart city does not necessarily involve monetary incentives. The participants, especially human users, can be satisfied by different forms of rewards, e.g., priority and social perception. Success stories include crowdsourcing applications such as Foursequare and Pokemon Go, where users are motivated to collect physical world information in exchange of reward points and game achievements.

Based on the information-centric architecture of smart-city IoT, a promising information trading business framework, namely Sensing-as-a-Service (S2aaS), has been introduced [8], [9]. The components in a smart city are categorized into four conceptual layers [9], as shown in Fig. 1, including: sensors to directly collect data, sensor publishers to transfer the data, service providers to use data and produce services, and users. Participants in the same layer of S2aaS can be potential competitors since they provide complementary or substitutable information and services in the “IoT market”. Under the S2aaS framework, information can be circulated as a commodity. Additionally, information with different quality will be priced based on their market values in the trading.

C.                  Information Freshness and Social Impacts

Information has different nature from traditional physical commodities while being priced. Information has merely no marginal cost while being reproduced, i.e., virtually no cost for copying and forwarding information. By contrast, freshness of information and social impacts affect the value of information more significantly.

Age-of-Information (AoI) [10] is a concept to measure the freshness of information, defined as the time interval from the most recent update of the information at the information recipient side. In the context of smart-city IoT, users may operate as “free riders” that wait until the information is disseminated across the IoT network as time elapses. However, as AoI increases with time, the trading value of information decreases immediately after the information is generated. Discriminatory pricing strategies, i.e., based on the time that the information is made available to a participant, can be adopted. As such, the users are required to buy information before the information becomes not fresh or even obsolete.

Equally important, social impacts affect information pricing because of the existence of other participants, defined as externalities, which mainly include the network effect and congestion effect. The network effect represents the situation in which some market participants mutually increase their utility when they are in the system. For example, a commuter benefits more from a higher accuracy of travel time when more of other commuters contribute travel information. On the contrary, the congestion effect decreases the utility due to detrimental performance.

The age-of-information and social impacts directly influence the development of the S2aaS model for smart-city IoT. Specifically, a number of sensor devices increase the impacts on the performance of users. The reason is that service providers may potentially collect information from more sources and with higher precision, which increases the utility of the users in the market.

IV. DESIGN CHALLENGES

To analyze the current progresses, as well as design issues of market-oriented information trading in smart-city IoT applications, typical cases of market-oriented approaches for smart-city IoT applications are presented in this section.

A.                  Raw Data Processing and Trading

In a smart city, raw data can be generated from time to time. Raw data sensed by sensor devices may have little added value as information. As a result, information processing and trading can be a major business for smart-city participants. Placemeter^[2] collects and processes video streams of city cameras, turning the video data into meaningful structures information about street situations, e.g., traffic trends and pedestrian directions. Clearly, the processed information has a much higher value than that of raw video data. In this regard, IoT data owners can become micro service providers [2], as shown in Fig. 1, if they implement some video analytics.

B.                  Mechanism Design for Resource Allocation

In a smart-city IoT application, value of information needs to be revealed to smart city participants. Auctions, as a method for allocating exclusive resources among multiple users, can be employed as a mechanism to valuate information in the smart city for participants by negotiating, asking and bidding deal prices. An auction has been designed for mobile crowdsensing in smart cities [11], where an application task runs a reverse auction to select smartphones for data collection. An important advantage of the auction approach is that there are plenty of auction mechanisms being put into practice with desirable economic properties, e.g., market efficiency and social welfare maximization.

C.                  User Competition and Cooperation

Models developed for competition and cooperation can be adopted into smart-city IoT scenarios for physical resource sharing and multiple access control purposes. Participants in smart city can be motivated to cooperate and compete to achieve optimal strategies of resource utilization and data transmission, such as limited system resources, e.g., communication bandwidth, in a distributed fashion. Game-theoretic approaches are usually employed as a market-oriented approach. For example, the Kolkata paise restaurant (KPR) game is proposed in [12] to optimally allocate multiple resources among users considering the preferences of the users. Likewise, to improve utility of smart-city IoT users in sharing resources, coalition among the users can be formed [13].

D.                 Market-oriented Security in Information Trading

Security issues are important especially for future autonomous information trading processes.

Security in an information market involves data protection and secured trading process. For IoT devices, a specially designed module, such as a security auditing module [14], can be implemented to detect and record threats occurred in IoT operations. Furthermore, blockchain technology [1] is introduced to provide protection and storage for distributed transactions, which is inherently suitable for decentralized information trading accounting in smart-city IoT applications [4].

Existing works in the literature ignore the social impacts which play an important role in market-oriented approaches and designs for smart-city IoT, as well as correlated trading processes among all the different system levels. This motivates us to introduce a novel game-theoretic model for information trading in smart-city IoT. The model is able to capture competition among the participants together with externality.

V. A GAME-THEORETIC MODEL OF INFORMATION TRADING IN SMART-CITY IOT

In this section, we propose a game-theoretic model considering existing externalities in the system. In the model, an IoT service provider offers IoT service to users with a uniform price. Also, the provider pays rewards to IoT information vendors which own sets of IoT devices to generate IoT information.

A. System Model
As shown in Fig. 2, an IoT service market consists of three types of participants, i.e., game
players.
• IoT service provider: An IoT service provider works as a resource-rich agent or institution between IoT users and IoT information vendors. The provider allows users to send a demand to purchase IoT services at a fixed subscription price without knowing the details of IoT vendors. The prices are decided by the provider. The provider serves the user demand by obtaining and processing raw IoT information sensed by IoT devices owned by vendors.
• IoT information vendors: A vendor owns a set of different IoT devices to sense and collect raw data. The raw data, e.g., an image from a video camera, is delivered to the provider for further processing for the IoT services. A certain amount of rewards, decided by the vendor, will be charged to the provider.
• IoT user: An IoT service user requires services from the provider in the market. As
customer-type participants, the behaviour and decisions of IoT service users can be affected
directly and indirectly by other market participants, which is defined as externalities.
In other words, the vendors, provider, and users are considered as the sellers, resellers, and consumers in the market, respectively.
We model the interactions among the three players as the three-stage Stackelberg game, with sub-game perfect equilibrium solved to determine the optimal strategies of all the three types of players in the proposed system.

B. Demand Analysis and Utility of IoT Service User
In the IoT service market, all the three types of participants aim to optimize their payoff. The payoffs of all the three participants are defined as utility functions.
We consider N IoT users in the market. Each user i determines the demand for IoT services, denoted by xi > 0. The utility ui can be affected by the following factors:
• Demand level of the user in the market,
• Number of vendors accessed by the provider, and hence the number of accessible IoT devices in the system, and

• Price charged by the provider.

Note that when there are more IoT devices, the collected service tends to have better quality from richer available data. Thus, it contributes to the utility of users.

In this case, the benefit of each user i includes:
• Direct benefit obtained when user i utilizes the service provided by the provider, denoted as internal effects. The benefit is typically modeled as a concave function fi(xi) of demand that captures the decreasing marginal returns effect, e.g., a linear-quadratic function [15].
• Indirect network effects from the participation level of vendors. Naturally, when more vendors provide sensing data, the quality of IoT service improves, although the user does not directly access the information provided by the vendors.
• Direct network effects caused by all the users in the market. More users in the system
lead to additive utility to an individual user due to physical and psychological reasons. For example, a user may trust and be willing to participate in the market, if the user observes that there are many other users including friends already in the network. In this case, the market appears to be attractive to IoT users. The direct network effect term can be expressed as , where gij is the degree of the network effect that the existence of user j induces the utility of user i and x contains the demands of all users, i.e., x = {x1,x2,...,xN}.

The cost incurred to user i (i.e., negative value) includes:
• Direct congestion effects denoted by ξcon(x) directly affect the user utility negatively. When there are more users with higher demand, congestion happens and service performance is degraded. The congestion effect is typically a convex function in which the marginal cost increases as the total demand increases.
• Price per unit of IoT service charged to all the users is denoted by p.

C. Utilities of IoT Information Vendor and Service Provider
Consider a set of M vendors. Each vendor j decides whether to sell raw sensing data to the provider or not. The utility function of a vendor includes reward (benefit) and cost components. The vendor will participate in the market if reward rj is higher than cost cj. Thus, the utility of vendor j is denoted by Γj = (rj − cj)δj. Here, δj is a dispatch demand function, indicating the amount of user demand that is dispatched/sent directly from the users to vendor j. This dispatched demand can happen when the users find that they can use the sensing data from the vendor directly. For example, the IoT devices of the vendor may implement some sensing data processing, e.g., video analytics, which makes IoT information useful and available to the users. The dispatch demand increases when the reward increases and/or the cost decreases. In other words, the vendor serves more demand when it generates more reward or incurs less cost.

The provider decides the unit IoT service prices p charging to users to maximize its utility, i.e., profit, denoted by Π which is benefit minus cost. The benefit is generated proportional to the demand and price while the cost is the reward paid to the vendors.

D. A Hierarchical Stackelberg Game Formulation

We adopt a three-stage Stackelberg game theoretic model to analyze the proposed IoT market. Figure 2 presents the market-oriented information trading among the provider, vendors, and users as a hierarchical Stackelberg game.

• Stage I (provider): The provider acts as the leader in the game. It decides a price p∗ to maximize its utility Π.

E. Numerical Results
As an example, we evaluate the performance of the IoT market with five users, five vendors, and one provider. We mainly examine the impacts on the utility performance by different system parameters.

We first analyze the impacts of externalities, i.e., direct and indirect network effects, as well as the congestion effect from peer users. As shown in Fig. 3(a), network effect increases the user utility. On the contrary, congestion effect leads to lower user utility. In Fig. 3(b), indirect network effect introduced by the vendors leads to a higher user utility. That is, the existence of vendors in the system encourages the users to generate more demand.

From Figs. 3(a) and (b), direct network effect and congestion effect are both directly correlated with the number of users. When the number of users increases from 1 to 10, the impacts become more significant, as shown by the gap of performance curves in Fig. 3(a) becoming much wider. By contrast, the indirect network effect caused by vendors has relatively mild impacts on the user utility, as shown in Fig. 3(b). When the number of vendors increases, the user utility without indirect network effect is only slightly lower than that when indirect network effect is incorporated. Note that the scenario without indirect network effect happens when the users do not benefit from more vendors participating in the market. For example, the vendors can provide the same sensing data in which more data becomes merely redundant.

The vendor cost cj affects the optimal rewards of vendors, and consequently influences the price and user demand. To examine the impacts of vendor cost, we vary the sensing cost of each IoT device owned by the vendors from 0.5 to 5.0, as shown in Fig. 3(c). This variation can happen, for example, due to different operating IoT device, communication, and processing infrastructure expenses. The vendor utility decreases and approaches 0 because the cost becomes too high, preventing the user demand from being dispatched to the vendor. In this case, as the vendor cost increases, user demand decreases because of the indirect impact, and the amount of demand to the provider increases.

VI. CONCLUSION
In this paper, general approaches for market-oriented IoT smart cities are studied. With the properties of quantity and heterogeneity of information created and circulated in a smart city, an information-centric architecture of IoT for smart cities has been proposed. Based on the information-centric system architecture, we have discussed the value view of smart city systems for information tradings, with the analyses of incentive for trading, service model, information timeliness and social properties. Finally, an IoT system prototype scenario for information trading has been studied employing Stackelberg game theoretic modeling.

REFERENCES

---

[1] https://www.travelsmartrewards.lta.gov.sg/

[2] http://www.placemeter.com/

A Holistic Framework for Open Low-Power Internet of Things Technology Ecosystems

A Holistic Framework for Open Low-Power Internet of Things Technology Ecosystems

Peng Hu1, Member, IEEE

Abstract

The low-power Internet of Things (IoT) has been thriving because of the recent technological advancement and ecosystems meeting the vertical application requirements and market needs. An open IoT technology ecosystem of the low-power IoT has become increasingly important to all the players and stakeholders and to the research community. However, there are several mainstream low-power IoT ecosystems available out of industry consortia or research projects and there are different models implied in them. We need to identify the working framework behind the scene and find out the principle of driving the future trends in the industry and research community. With a close look at these IoT technology ecosystems, four major business models are identified that can lead to the proposed ecosystem framework. The framework considers the technical building blocks, market needs, and business vertical segments, where these parts are making the IoT evolve as a whole for the years to come.

I. Introduction

Internet advances with the openness in mind, so does the Internet of Things (IoT). An IoT system benefit from various kinds of technologies and developmental efforts driven by open IoT technology ecosystems involving open standards, open source tools, and open platforms with key stakeholders. Over the past decade, the innovative sensors, embedded systems, cloud computing, wireless networking technologies have been enriching the openness of IoT systems and fulfilling the needs of IoT system development. As a result, on the one hand, these technologies enable the extremely less power consumption on IoT devices than before, which enables a broad spectrum of zero-battery and battery-powered applications. On the other hand, an IoT system can be effortlessly built with an IoT ecosystem including the off-the-shelf IoT building blocks, ranging from sensors, interface circuits, embedded modules, and connectivity modules, to cloud service and data analytical tools. Low-power wireless networking technologies are indispensable for the low-power Internet of Things (IoT) systems ranging from wearable devices, smart appliances, smart meters, to smart city systems. There are mainstream low-power wireless networking technologies, from zero-battery and low-power wireless personal area network (LPWPAN) open standards, such as EnOcean, Bluetooth LE (BLE), 6LoWPAN, and ZigBee, to low-power wide area network (LPWAN) open standards, such as LTE-M and LoRA. These open-standard-compliant technologies are mostly driven by industry players and create the IoT technology ecosystems. Existing IoT ecosystems result in the fragmented IoT market and the effort of creating a consortium or alliance is the viable way to ensure the monetized return of the IoT businesses. This trend has divided IoT into vertical applications and divided market into vertical business segments. As a result, technology and media competition of the IoT market is no longer between individual firms but “among ecosystems of firms operating in loose alliance” [1]. We need to understand these ecosystems of the low-power IoT business.

However, the current research in the literature mainly discusses the business ecosystem of IoT arising from the supply network [2]–[4] without the consideration of an IoT technology ecosystem involving all IoT players and technical elements with an updated view of technology ecosystems, addressing the IoT research and industry trends, and identifying the framework behind the scene. In this way, we need to have a holistic framework identifying what the key technical building blocks are, how they interacts with business segments, and how to make the IoT business, enablers, stakeholders, and market integrate and co-evolve, in order to generate promising social and economic impact in the future. In this article we provide an updated overview of the IoT ecosystems, discuss the models of open lowpower IoT ecosystems, and propose a holistic framework that allows IoT systems from business and technology perspective to evolve and to integrate. In the following sections we present the concepts of IoT ecosystem, an overview of the existing IoT ecosystems categorized by four business models, as well as a framework that can provide a holistic view of the major low-power IoT ecosystems. Then, we discuss the case studies of the two successful ecosystems, AllJoyn and ARM mbed, with a conclusion.

II. Related Work

There are some related works discussing about the related topics about IoT ecosystems. We firstly introduce the IoT ecosystems and related business models. Then, the classical IoT-based business ecosystems are reviewed in contrast to the low-power IoT technology ecosystems, followed by the common features of IoT ecosystems. In [5], IoT ecosystems can be represented by the three components: enabling technology, IoT viable marketplace, and applications desirable to users or stakeholders. The IoT Architectural Reference Model (ARM) [6] has started an effort on creating an architectural model that allows IoT devices to interoperate in a standard way supported by the use cases introduced in [7]. The players in the IoT market is mentioned in [4] including commercial players, research and academia, governments and utilities, and others. The business ecosystem was introduced in [2] where the key players in addition to the ones in a supply chain are considered such as universities, industry associations and other stakeholders. An IoT-based business ecosystem is introduced in [3], where the and the technical RFID ecosystem is mentioned in [8]. We should note that the IoT ecosystem in this article is not a business ecosystem, but a technology ecosystem in consideration of all elements and business models. A business model is an element in any IoT ecosystems because the ecosystems are created to generate an economic return in the end. The term of business model has different definitions [9] according to contexts. The new IoT-based value creation in the IoT business model is mentioned in [10], where the interoperability of products and services as smart thermostats and light bulbs can generate new services along the data flow, such as processing optimization and forecasting. The business model framework that identifies the components of developing a business model for companies is discussed in [11], where the value proposition is considered the most important element of a business model. In addition to value creation, adoption of IoT can also create the operational process improvements, cost reduction, and risk minimization [3]. There are common features of IoT technology ecosystems. One feature is they need to have technical IoT building blocks such as the hardware platforms, operating systems (OSes), and software frameworks. Technical IoT building blocks play an important role in the IoT technology ecosystem and they mostly meet the open standards. There are rich sets of the technical IoT building blocks from the chip level, infrastructural level, to the end-user application level. These open standards may include open specifications by international consortia or standards made by international standardization bodies. Open standards make sure the interconnectivity and interoperability across IoT systems and de-risk the development for new products and services. This is welcomed by most of the business entities. For example, the IoT-A ARM architecture [6] necessitate different standards for different hardware/software components of a system which they need to be compliant with at different layers following the OSI model. Because of the versatility of IoT applications, various standardization efforts have been made by IEEE, IETF, ITU, and IEC, as well as industry consortia, including AllJoyn Alliance, Bluetooth Special Interest Group (SIG), ZigBee Alliance, Z-Wave Alliance, Open Connectivity Foundation (OCF, formerly called OIC), and EnOcean Alliance. Most standardization efforts made by industry consortia are on top of the underlying component standards and they provide the software frameworks of fast-prototyping an IoT system with inherent interconnectivity and interoperability of other systems with the same software framework.

III. Overview of the Current Low-Power IoT Technology Ecosystems

An IoT ecosystem can be further split into sub ecosystems and a way of identifying these sub ecosystems are through the technical building blocks, although not all the building blocks have their own ecosystem model. Table 1 shows the essential parameters in comparison with different low-power IoT ecosystems from different industry consortia or projects. We name ecosystem after the industry consortium or project name. Where the IoT ecosystem with open standards, proprietary standards such as ANT are not included. The license is listed for each ecosystem to show the openness and the supported OSes are listed as well. The vertical application of the ecosystem shows the business foci of each ecosystem. Most of the ecosystem identifies the home automation, including appliances, lighting, climate control, energy management, access control, safety and security in a home. The energy harvesting based EnOcean IoT ecosystem is shown in Table 1, where it targets at the home automation applications. EnOcean ecosystem is built on top of ISO/IEC 14543-3-10 standard, where it fits a new market, connects the electronics/semiconductor industry and provides development tools. The LPWPAN includes Bluetooth (and BLE defined in the 4.0 specification), ZigBee, DASH7, and ZWave. They originally offered specifications based on the physical-and link-layer open standards such as IEEE 802.15.4, ISO/IEC 18000-7, and IEEE 802.15.1, and now provide a full-fledged development kits and tries to expand the established markets in home automation, to the markets of retail or industrial applications. We can see that they all have similar licensing model where use of software libraries are free but the final products need to go through the licensing or certification process. In addition, WirelessHART is a special ecosystem built on top of IEEE 802.15.4-compliant radios, where it targets the manufacturing automation vertical application that differs itself from other LPWPAN technologies. The rising of low-power wide area network (LPWAN) technologies can extend the communicate coverage to the city wide which fill in the market gap left by the LPWAN technologies. LET-M and LoRa are two examples based on open standards/specifications. LTE-M is alongside the LTE infrastructure so that the LTE carrier providers would welcome this, while LoRa supports the private network without the LTE infrastructure. The licensing model of LTE-M is similar to LTE and it is expected to be bound to the platform providers, while LoRa adopts the LGPL software license. The ecosystems such as IoTivity, Thread, AllJoyn, and ARM mbed can work with the underlying lowpower transport technologies such as BLE, 6LoWPAN, and IEEE 802.15.4. IoTivity and AllJoyn are Linux Foundation software framework projects where they have differences. They have different architectures and AllJoyn focuses on the service framework, device libraries, and applications, while IoTivity focuses on the device discovery, transmission and management. ARM mbed supports low-power radios and focus on the ARM-based hardware platforms with the support from firmware, hardware design, middleware, and cloud services. In addition, it is important to realize the vertical segments that an IoT software framework can address. Although most of them are in the home automation, smart home, and smart city business. However, there is possibility of the IoT software frameworks to be used in other segments such as manufacturing systems, energy systems, and automotive systems.

Table 1. List of major low-power IoT technology ecosystems where the key parameters are listed including the type of business model, software license, supported operating systems (OSes), security, alliance, and vertical applications.

License Supported OSes Connectivity Technology Supported Security Support Firm / Alliance Vertical

IV 3-Clause BSD-based license (OpenThrea d by Nest) Platform independent IEEE 802.15.4/6LoWPAN Yes Thread Group Home automation IoTivity IV Apache License 2.0 Android, Tizen, Arduino, native Linux, and platformindependent library BLE/Bluetooth, WiFi Direct, Ethernet Yes Open Connectivity Foundation (formerly OIC) Generic AllJoyn IV Internet Software Consortium (ISC) Windows, Android, Linux, iOS, and platformindependent library WiFi, Serial, Power Line Communication (PLC), Ethernet, 6LoWPAN, EnOcean Yes AllJoyn Alliance Home automation ARM mbed I, II Apache License 2.0 (and other licenses) Platformindependent library Ethernet, WiFi, 6LoWPAN / Thread, BLE Yes ARM Generic ZigBee I, III Certification process is required Platformindependent IEEE 802.15.4/ZigBee (can be integrated to AllJoyn and OCF) Yes ZigBee Alliance Smart home, lighting, utility, and retail industry Z-Wave I, III Certification process is required Platformindependent ITU-T G.9959/Z-Wave (can be integrated with AllJoyn and OCF, and HomeKit) Yes Z-Wave Alliance Wireless home control and monitoring EnOcean I, III Certification process is required Platformindependent ISO/IEC 14543-3-10:2012 Yes EnOcean Alliance Home automation WirelessHART III Certification process is required Platformindependent IEC 62591 Yes HART Foundation Manufacturing automation LoRa I, III LGPLv2.1 Platformindependent LoRaWAN specification Yes LoRa Alliance Generic LTE-M I N/A Platformindependent LTE Yes 3GPP Automotive, smart grid, smart city, healthcare, smart home, industrial DASH7 I, III OpenTag License OpenTag OS ISO/IEC 18000-7/DASH7 Mode 2 Yes DASH7 Alliance Building automation, smart energy, advertising, automotive, logistics Bluetooth I Licensing process is required Platformindependent IEEE 802.15.1/Bluetooth/BLE Yes Bluetooth SIG Consumer electronics, healthcare RIOT I, IV LGPLv2.1 Linux, platformindependent IEEE 802.15.4/6LoWPAN, WiFi Yes N/A Generic Contiki OS I, IV 3-Clause BSD-based license Contiki OS IEEE 802.15.4/6LoWPAN Yes N/A Generic

IV. Business Models of Low-Power IoT Technology Ecosystems

Here we inherit the definition in [12] to define the business model of IoT ecosystems here as the model that the value of the ecosystem consisting of a consortium or an alliance offers to all the key players including member firms, customers, developers, and researchers. There are four typical business models behind the aforementioned ecosystems as shown in Table 1.

Figure 1. Four common models of low-power IoT ecosystems

A) Model I: Semiconductor-Driven Model

This model is created by the hardware-based ecosystems from semiconductor business models. The semiconductor companies manufacture the hardware components and creates the software libraries and development kits with the partners for customers including developers and system integrators. The research element is the important enabler for design and fabrication process innovations reflected by the Moore’s Law. This model is alongside the traditional vertical segment of embedded systems where the main customers of semiconductor companies are not end users but companies who will buy the hardware components in a large volume to use them in the final products. In recent years, the huge opportunity and availability of IoT technologies have further resulted in the additional competition among the horizontal semiconductor segments, where they are now competing to provide more open, easy-to-use and highquality modules and development kits than before. For example, in 2014, Freescale released the support for MEMS Industry Group by offering loyalty-free open-source sensor fusion library that allows the easy development IoT application based on the essential sensor fusion building blocks. In 2015, Texas Instruments (TI) released the multi-protocol SoC CC2650 chip with a complete software/hardware development kits, such as software libraries, mobile app examples, and cloud service connectors, in addition to the standard offerings such as hardware reference design and application notes. ARM mbed follows a typical semiconductor-driven model where the various hardware platforms and tool chains by ARM mbed are based on ARM microprocessors. Although this model is well validated by all major semiconductor companies such as TI, Atmel, ST, NXP, and Freescale, but ARM mbed has built a fullfeatured technical ecosystem connecting hardware vendors and providing the technical building blocks from chip-level to open circuits, OS, middleware, and cloud services.

Model III: Vertical Application-Driven Model Model IV: Framework-Driven Model Model II: Data Service Driven Model Model I: Semiconductor-Driven Model

B) Model II: Data Service-Driven Model

The data-driven model aims to provide the security, messaging, schemas, processing, and interoperation of generic data required by the IoT applications. It is undeniable to say data is the main part of the value creation in many businesses, and the ecosystem built with this model has been validated by the market in alignment with the market success of the existing platforms such as Amazon IoT platform, IBM Bluemix, Google Cloud Platform, IFTTT, and ThingSpeak. This data service can be provided independently as long as the open data connectors such as REST data connector are available on the hardware platform. In addition, the developers and researchers as the possible customers have already advanced and been benefited from the data service platforms.

C) Model III: Vertical Application-Driven Model

Vertical applications are to denote particular application areas. This vertical application-driven model refers to the IoT application sets that differ from each other, where each would require unique specifications. For example, in addition to the LoRa, ZigBee, Z-Wave, DASH7, and WirelessHART shown in Tabl1, the LORD MicroStrain platform provides another example ecosystem which focuses on the industrial sensing applications. On the sensor device it has specialized protocol for microsecond time synchronization based on IEEE 802.15.4-compliant radios, and one the cloud end it provides the on-line data processing APIs. Another example is the NI wireless monitoring platform where it provides a set of development kits that can work with LabView and data acquisition chases for the industry-grade sensing applications. Shimmer provides a similar IoT solution to the healthcare applications. An IoT system is expected to be a multi-purpose computer system that can support various types of tasks, which generates new business opportunities and values. However, when talking about the IoT ecosystem, we need to conduct an abstraction from this multifaceted use cases of IoT. From a system’s perspective, an IoT system can be split into few main sub systems from sensor and actuator systems, to embedded systems, and distributed systems. From a computer network’s perspective, it can be split into the classical OSI layers from physical layers, network and transport layers, to application layers.

D) Model IV: Framework-Driven Model

The framework-driven model refers to the offering of software frameworks working at a high level of an IoT system with the middleware that makes it hardware platform agnostic. The example of this business model include AllJoyn, Thread, and IoTivity, which are software framework based ecosystems. Today’s IoT system may employ one or more transport technologies and it is important a software framework should support them. For example, a wireless IoT router needs at least a wireless transport and Ethernet connectivity in order to communicate within the local wireless network and to the external IP backbone network. In order to support flexible number of transport technologies, we need to make sure they can be “bridged” to these networks. With the software frameworks, these networks can be integrated into the application protocols in the framework. Figure 1 shows the possible overlaps between aforementioned models which is true from Table 1 that an ecosystem can fit more than models. For example, it is possible that a chip manufacture provides a full set of software development kits that offer some features in the framework-driven model. Moreover, one may wonder the model implied in the IoT ecosystems such as Arduino, Raspberry Pi, or Beagle Bone. They indeed have enabled the proliferation of IoT applications for years, but since they are not focused on the low-power IoT ecosystems, they are out of the scope of this article. In addition to the aforementioned models, there exists a research-driven model that the IoT ecosystem originates from the research community. Contiki OS and RIOT are two examples of it where both of them focus on the low-power IoT solutions. There is a correlation of this model with Model I or IV as some research-driven IoT ecosystems have a well-designed business model with successful market share, support, and platform support from industry. However, the research-driven model of IoT ecosystem needs to plan for the commercialization stage in order to avoid the possible support and maintenance issues that cannot be entirely resolved by the open-source community.

V. A Holistic Framework for Low-Power IoT ecosystems

Based on the aforementioned business models, we introduce a framework for low-power IoT ecosystems as shown in Fig. 2.

Figure 2. A holistic framework for low-power IoT ecosystems. This framework shows the research and market demands as the driving force as well as the vertical business segments and technical IoT building blocks.

Figure 2 contains two parts in the front plane: the technical IoT building blocks part and the common vertical business segments part. The horizontal business segments are briefly shown on the top plane of Fig. 2, where there can be multiple similar business entities of a business block shown on the front plane. The architecture plane indicates the various systems following an architecture based on the common IoT building blocks. On the side plane, the commercial and research market demands are shown as the market driving force of the elements shown on the front plane. In the technical IoT building blocks, the open standard block connects to all other blocks including the sensor/device-level blocks such as drivers, protocol stacks and hardware platforms as well as applicationlevel blocks such as the software framework and IoT apps where the software framework can make all the IoT building blocks work together. There are three categories of open standards, such as sensors standards[1], system integration standards, and interoperation standards, where each category contains a series of standards. The data connectivity block requires physical connection to provide actual wireless connection between IoT devices. These technical IoT building blocks connect the business and research worlds and are mostly open in terms of technology, licensing model, and integration with other building blocks. The blocks representing vertical business segments sit outside the technical building blocks and the dotted arrow showing the relationship between business blocks and technical building blocks. Semiconductor and electronics business provides the essential bedrocks of IoT systems such as sensors, actuators, and electronic components. The system integration business can provide the hardware platforms or out-of-the-box modules based on the electronic components. Manufacturing business includes any manufacturing business regarding the IoT hardware components, such as PCB fabrication and 3D printing business for the custom cases for an IoT device. The data connectivity building block requires the Internet connection services in the “Internet service business” block. Although data connection between local devices in a LPWPAN is free, the data communication between a router and a remote service is not. A company providing LTE-M based LPWAN data connectivity service for smart city applications fits the on-premises or subscription service business model. Data service business includes data storage, data processing, and data operations. These common services are important for many IoT applications. For example, for the smart thermostat application, all thermostat sensing data can be hosted in the data service provider’s premises where the data analysis tool can process the data and show the results to the users instead of showing a large amount of data. Software service business is related to the software related components such as protocol stacks, operating systems, software frameworks, and applications. This software business includes consulting services, outsourcing services, software support services, etc. Each of these services already has a mature business model. The cloud service business closely relates to the modern IoT applications where the distributed services or a remote server can be accessed and hosted. The successful business models based on the service models such as Software-as-a-Service (SaaS), Platform-as-a-Service (PaaS), and Infrastructure-as-a-Service (IaaS) can be used in this business segment. The application solution business relates to the complete IoT system where it can provide service to different kinds of IoT applications such as home automation, smart city, smart grid, and manufacturing automation. These application solutions need to be validated and well-designed and it generate unique value to the customers. The cybersecurity business provides the security solution to the IoT system which is distributed in essence. Although we have seen in Table 1 that security features are provided with each IoT technology but it is just a basic building block which is far less possible to tackle all the cyber threats. The existing cybersecurity industry is the example of this business.

VI. Discussion

AllJoyn and ARM mbed are the recent successful IoT ecosystems that fit the previously proposed Model I and IV and the framework. Both of them are well connected to the business stakeholders, researchers and developers. The AllJoyn software development kits (SDKs) are open to the public and its services are based on the low-power wireless open standards and provides the agent middleware running on the, computers, routers, and mobile/embedded devices. It fits the framework in Fig. 1 in that the commercial market needs a full-fledged and easy-to-use solutions where they are easily built with any possible OSes, including Android, iOS, Windows, Linux, or any generic embedded OS. AllJoyn abstracts the common business logics of the mobile based IoT applications and provides separate software libraries to the mobile phones and embedded devices while keeping the same communication protocol. The applicationlayer communication protocol provides an easy way to enable interoperations between any AllJoyn devices, although it may introduce the latency issue of data transmission for the time-critical application. However, this is not a concern in most of consumer-grade applications. The AllSeen Alliance requires the membership to create a new project or work on the existing projects but it provides the free members for academic users to do so. ARM mbed provides SDKs from the firmware, OS, connector middleware, to the cloud service which can be mapped to the technical building blocks in Fig. 2. It supports a broad range of low-power IoT protocols and application-layer protocols, and it has built partnerships with hardware manufactures where many ARM-based hardware platforms can support ARM mbed. Any users can effortlessly build solutions on top of it with a low cost, although there are multiple licenses involved in the software components. The loosely coupled architecture based on open standards allow the vertical business segments to utilize the ARM mbed solutions. For example, application solution business can build a dedicated virtual application on top of the ARM mbed ecosystem. In addition, the value creation out of AllJoyn and ARM mbed are embodied in the support for device, edge, and cloud computing. The technical building blocks provided by ARM mbed and AllJoyn enable the edge and cloud computing as well as the intrinsic on-node computing. To this end, the intelligence and data processing can be implemented within any architecture.

VII. Conclusion

With the fast-advancing low-power wireless technologies at different scales including LPWPAN and LPWAN, the technical barrier of IoT applications is further removed, resulting in the proliferation of IoT applications and business opportunities in the coming years. The challenge now is how to identify a business model in an IoT technology ecosystem to integrate these technologies. To this end, it is important for us to review the IoT system in consideration of technology evolution, markets, and related businesses. The proposed business models and framework that reveal the principle of various low-power IoT technology ecosystems and help us understand the synergy and convergence of the key elements in the ecosystems.

References:

[1] E. Kelly, “Introduction: Business ecosystems come of age,” in Business Trends Series, Deloitte University Press, 2015.

[2] M. Perkmann and K. Walsh, “University–industry relationships and open innovation: Towards a research agenda,” Int. J. Manag. Rev., vol. 9, no. 4, pp. 259–280, 2007.

[3] K. Rong, G. Hu, Y. Lin, Y. Shi, and L. Guo, “Understanding business ecosystem using a 6C framework in Internet-of-Things-based sectors,” Int. J. Prod. Econ., vol. 159, pp. 41–55, Jan. 2015.

[4] IEEE, “IEEE-SA Internet of Things (IoT) Ecosystem Study,” 2015.

[5] P. F. Ovidiu Vermesan, Internet of Things: Converging Technologies for Smart Environments and Integrated Ecosystems. 2013.

[6] P. G. Martin Bauer, Mathieu Boussard, Nicola Bui, Francois Carrez, A. N. Stephan Haller, Edward Ho, Christine Jardak, Jourik De Loof, Carsten Magerkurth, Stefan Meissner, M. Alexis Olivereau, Alexandru Serbanati, and J. W. W. Thoma, “The IoT Architectural Reference Model,” 2012.

[7] A. Bassi, M. Bauer, M. Fiedler, T. Kramp, R. van Kranenburg, S. Lange, and S. Meissner, Eds., Enabling Things to Talk -- Designing IoT solutions with the IoT Architectural Reference Model. Springer Berlin Heidelberg.

[8] 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,” IEEE Internet Computing, vol. 13, no. 3. pp. 48–55, 2009.

[9] A. Ovans, “What Is a Business Model?,” Harv. Bus. Rev., vol. January, 2015.

[10] G. Hui, “How the Internet of Things Changes Business Models,” Harv. Bus. Rev.

[11] R. M. Dijkman, B. Sprenkels, T. Peeters, and A. Janssen, “Business models for the Internet of Things,” Int. J. Inf. Manage., vol. 35, no. 6, pp. 672–678, Dec. 2015.

[12] A. Osterwalder, Y. Pigneur, and C. L. Tucci, “Clarifying Business Models: Origins, Present, and Future of the Concept,” Commun. Assoc. Inf. Syst., vol. 16, no. 1, pp. 1–25, 2005.

---

[1] Open Standards Protocol Stacks Hardware Components Hardware Abstraction Level Sensors Standards System Integration Standards Interoperations Standards IoT Software Framework Software Service Business System Integration Service Business Internet Service Business (carriers, ISPs, etc.) Data Service Business IoT Apps IoT Operating System Data Connectivity Application Solution Business Semiconductors & Electronics Business Cloud Service Business Cybersecurity Business Manufacturing Business IoT Hardware Platform Vertical Business Segments Technical IoT Building Blocks … Commercial Market Demands Research Market Demands Architecture Horizontal Business Segments