The evolution of computation devices through history was characterized by improvements in their; cost, size, power consumption, computation power, number of deployed devices and connectivity. Emerging generation of devices will be even cheaper, smaller, more energy efficient, more computationally capable and deployed in bigger numbers than the smart mobile devices from previous generation. The emerging devices are connected to Internet and interoperable by default.
The collective existence of new generation of computation devices is called the Internet of Things (IoT). IoT is gaining momentum and positively influences the automation of everyday objects. Technically speaking, IoT is a paradigm that can network any object and allow it to communicate data about its internal states, environment and its usage. IoT devices are currently at the infant stage of development and their mainstream market adoption is still to come, in ever large numbers.
IoT architecture is shown in Figure 1 below. This layered structure is inspired by the network representation partially because IoT can be, indeed, seen as an extension of classical network architectures. In this structure, IoT nodes are the basic building blocks. Nodes are connected to network either directly or through gateways. Network layer represents a network that has ability to cover a geographic area, most commonly LAN or WAN. The network ultimately extends the nodes capabilities all the way to the Cloud. Cloud is composed of three layers; middleware, services and analytics tools and users and 3rd party integration. Middleware is a part of the IoT architecture that enables the management and connectivity for large number of Nodes and serves as a bridge between Nodes, services and users. Services are built on top of the meaningful patterns in data discovered using analytics tools. Users and 3rd party integration layer is where the value is harnessed and/or incorporated to other systems.
Each layer in the given IoT architecture is composed of multiple development technologies and each of them is at different level of tech maturity. Let’s take closer look…
Users and 3rd party integration
The maturity of technologies constituting the users and 3rd party integration layer is very high. Connecting objects to Internet and allowing them to exchange data with other objects create value through careful design of services. Generated value is delivered to users through mobile technologies and web applications. 3rd party integrations are done through well-established web tools/services. Web and mobile technologies used in this layer matured together with the smart phone market in the past decade and now establish themselves as the most mature layer of IoT. The pace of development of technologies contained in this layer is expected to increase in future through new user interface mediums (wearables, voice, mixed reality).
Services and analytics tools
Services layer can be seen as digital abstraction of real world application of IoT. Business value of IoT applications is embedded in service layer. Most importantly, services enable monetization of data. IoT based services are implemented either as enriched version of traditional services or they are designed as completely new services based on the previously untapped data.
According to Altimeter research, companies nowadays leverage IoT mostly to improve customer experiences and to enhance existing, or create new business models. Traditional maintenance and utility companies use IoT to design services that transform nature of their operations from reactive to proactive. At the present state of technology, IoT service provider companies have narrow offerings that solve specific use cases, e.g. predictive maintenance of industry equipment such as pumps, generators or elevators. Currently these early IoT service offerings are in form of monitoring applications. Nevertheless, the expectation of the next generation service providers is to deliver advanced analytics and ability to act based on the data analytics results. For this goal, in order to develop true value-added services and achieve satisfied results, companies will need to dig into domain specific challenges while leveraging connectivity benefits.
Analytics tools used for the benefits of services generate insight from data using various constructed statistical models. Statistical models are based on application data and serve the purpose of structured prediction, anomaly detection and knowledge discovery. Analytics tools used in construction of advanced IoT services are evolving through the evolution of artificial intelligence, machine learning, deep learning and big data technologies. Main questions like how to integrate domain specific knowledge together with analytics tools are currently partially answered for speech recognition and vision based application, for other applications, questions like these will be answered through collection and analysis of substantial amounts of application data.
IoT require Cloud as a medium where aggregation, storage, processing, management and sharing of the data is handled. Middleware is an important part of Cloud that bridges Nodes on one side and services and users on other.
Common elements of middleware include the protocol gateway, data storage, device management and API management. The data storage and API management technologies are the same as used in other Cloud and Web applications. The technology used in these two elements has matured well in the past decade. The protocol gateway technology has partially matured with the conventional web applications (HTTP and TCP protocols). In the aspect of IoT specific protocols such as CoAP and MQTT the developed technologies are new, however, their adoption has been done relatively quickly. Both of these protocols became the standard through the embodiment of their capabilities into well established infrastructure. The device management element of IoT middleware has been adopted from mobile device management and applied to less computationally capable devices. Device management element is technologically well matured, nevertheless, the interoperability challenge in IoT will most probably require the full redesign of this element in the future.
Network layer adds long haul connectivity to the IoT. IoT devices can partially benefit from networks developed for mobile devices. Those networks are fast and reliable and benefit the critical IoT applications. However, ubiquitous IoT devices also have demanding requirements with respect to energy consumption (battery powered), wide area coverage (rural and uninhabited places), easy deployment of new networks, cost of connectivity modules, cost of data transmission and diversity (multi-protocol support).
Speed and reliability requirements of critical IoT applications have been met by 4th generation wireless mobile telecommunication networks (the further improvements are expected by the arrival of 5th generation). These networks operate on licensed spectrum and target high-quality mobile voice, data and video services. The advantages of these networks can be harvested by applications such as traffic and safety control, smart grid, health care and remote manufacturing that can compensate network cost with their need for reliability and speed. On the other hand, ubiquitous and cheap IoT devices used in applications such as smart metering, smart agriculture, logistics, fleet management, retail, smart building and smart homes have different connectivity requirements that cannot be satisfied with the current cellular network installations. Current cellular network installations, EDGE, 2G, 3G and LTE have power hungry protocol implementations, unreasonable prices of communication modules, equipment and data and poor coverage in the rural areas.
New type of networks have been rolled out to meet the IoT connectivity requirements such as, LoRaWAN, SIGFOX, NB-IOT, Weightless or RPMA. All of these networks are basically, low power wide area networks (LPWAN) granting long battery life to the IoT nodes. They are also based in the unlicensed band which lowers the price of communication and due to the narrow band nature the reach of devices running these protocols is remarkable, up to 10 km with very small antennas. Further development of these and similar networks is expected in the near future. With LPWAN the network layer is one step closer to what is really needed in IoT.
Gateways serve primarily as the data bridge between Network and consequently Cloud on one side and IoT Nodes that don’t have direct network connectivity on the other. In that sense, gateways enable the communication between different Nodes and Cloud services by translating different protocols. Besides the protocol translation, gateways also serve the purpose of making sense of data close to the Nodes. They enable so-called smart edge, because Nodes are usually neither able to execute complex computations nor to maintain long term and secure connection to Cloud.
The gateways are deployed in large numbers and constitute a majority of mature embedded IoT elements. One of the main reasons for their maturity is stemmed in well-developed IT legacy. Certain number of gateways were put in place long before the deployment of IoT Nodes to serve as protocol translators. The necessity for gateways will increase in the future due to their important role in making interoperability possible, ability to effectively aggregate and process data while providing remote control and management. Further development in processor technology will enable gateways to become effective computational machines capable of running advanced machine learning algorithms and delivering advanced edge analytics.
The nodes in the context of given IoT architecture (Figure 1) are devices external to — or a part of — real world physical objects. In more general sense, a node is data generating “thing” closest to the physical world. Nodes have the ability to sense their internal states as well as the environment, perform the necessary computations as required by the applications, act on their internal states as well as the environment and communicate data bidirectionally with gateway or Cloud. The synergy of hardware, software, communication and power technologies allows nodes to serve as a front end of IoT in the quest of sense making of physical objects. Among the layers of IoT architecture, the Node layer is characterized with the most stringent security, computation, size, power, cost and communication requirements. These draconian requirements are amplified by the pervasiveness and desired interoperability factors.
Despite the various opposing claims from the community, the Node layer is the most underdeveloped layer of IoT. IoT Nodes are still at their technological infancy due to the challenges to satisfy the necessary design requirements. The tackling of challenges associated with the development of IoT Nodes needs to be done in holistic way by understanding the technological heterogeneity, their place in the big picture and their relation with the application. Currently developed Nodes are partial solution missing one or more components of the necessary holistic solution. Consequentially the developed Nodes are relatively primitive and expensive.
In the layered IoT architecture from Figure 1 the most mature layer is users and 3rd party integration layer. This layer also has very promising future as mobile tech advances through AR/VR and artificial intelligence applications. IoT applications are also gaining another domain of user experience through voice controlled assistants.
Second on the maturity scale is the middleware layer. The middleware layer has adopted most of its technologies from well established web and Cloud applications. Middleware layer is currently employed in all of the existing IoT applications and will play crucial role in making IoT interoperable.
Third most mature layer is the services and analytics layer. Nevertheless, the maturity of tech in this layer can be considered moderate and further developments are necessary. True success and maturity of this layer are strongly related to the successful development and integration of application specific knowledge with the advanced analytics tools.
Fourth place on the maturity scale belongs to network layer. This layer recently went through evolution with the introduction of LPWA networks. These networks satisfy the need of many IoT applications, however their adoption is still in the infancy.
Gateways are on the fifth place. Although large number of gateways is currently deployed in many different IoT applications, most of these gateways are IT legacy that still serves the sole purpose of protocol translation. IoT needs more versatility and there is long road ahead until we reach the scenarios where gateways are at the frontier of security and edge computation.
Nodes layer is by far the least mature layer of IoT. Nodes are the most crowded layer of IoT (deployment number expected in the magnitude of trillion), smallest in size (down to tens of mm²), have smallest power budgets (down to micro Watts) and have cost expectations below 1$. Currently, Nodes are way below the deployment expectations, over-sized, power budgets in the range of Watts and cost is in the range of hundreds of dollars. For Nodes to reach the desired level of maturity there needs to be shift in understanding. Nodes require that holistic design approach which is currently either ignored by companies due to the cost of employing experienced teams or they are not aware of it at all due to the lack of visionary management.
In addition to the discussed challenges in adoption, IoT technologies are currently very fragmented which is a fundamental challenge for the economy of scale and interoperability. To mitigate this challenge some companies have turned to platform technologies. While such approach can work nicely for the Cloud based platforms, there are still no embedded solutions that solve the economy of scale and interoperability challenge on the Node level.