4724 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 14, NO.

11, NOVEMBER 2018

Industrial Internet of Things: Challenges,

Opportunities, and Directions
Emiliano Sisinni , Member, IEEE, Abusayeed Saifullah , Member, IEEE, Song Han , Member, IEEE,
Ulf Jennehag , Member, IEEE, and Mikael Gidlund , Senior Member, IEEE

Abstract—Internet of Things (IoT) is an emerging do-

main that promises ubiquitous connection to the Internet,
turning common objects into connected devices. The IoT
paradigm is changing the way people interact with things
around them. It paves the way for creating pervasively con-
nected infrastructures to support innovative services and
promises better flexibility and efficiency. Such advantages
are attractive not only for consumer applications, but also
for the industrial domain. Over the last few years, we have
been witnessing the IoT paradigm making its way into the
industry marketplace with purposely designed solutions. In
this paper, we clarify the concepts of IoT, Industrial IoT, and
Industry 4.0. We highlight the opportunities brought in by
this paradigm shift as well as the challenges for its realiza-
tion. In particular, we focus on the challenges associated Fig. 1. IoT, CPS, IIoT, and Industry 4.0 in Venn diagram.
with the need of energy efficiency, real-time performance,
coexistence, interoperability, and security and privacy. We
also provide a systematic overview of the state-of-the-art Internet will improve the sustainability and safety of industries
research efforts and potential research directions to solve and society, and enable efficient interaction between the physical
Industrial IoT challenges. world and its digital counterpart, i.e., what is usually addressed
Index Terms—Industrial internet of things (IIoT), real-time as a cyber-physical system (CPS). IoT is usually depicted as the
communication, reliability, security, wireless sensor net- disruptive technology for solving most of present-day society
work (WSN). issues such as smart cities, intelligent transportation, pollution
monitoring, and connected healthcare, to name a few. As a sub-
I. INTRODUCTION set of IoT (see Fig. 1), Industrial IoT (IIoT) covers the domains
of machine-to-machine (M2M) and industrial communication
NTERNET of Things (IoT) is a computing concept describ-
I ing ubiquitous connection to the Internet, turning common
objects into connected devices. The key idea behind the IoT
technologies with automation applications. IIoT paves the way
for better understanding of the manufacturing process, thereby
enabling efficient and sustainable production.
concept is to deploy billions or even trillions of smart objects Flexibility and scalability required by IoT communications
capable of sensing the surrounding environment, transmit and are typically addressed using wireless links. In the past, wire-
process acquired data, and then feedback to the environment. It less technologies in industrial applications were mostly based
is expected that by the year 2021, there will be around 28 billion on ad hoc solutions, e.g., individually developed for connecting
connected devices [1]. Connecting unconventional objects to the moving parts or hard-to-reach devices. Only recently, standards
purposely designed for the industry (e.g., WirelessHART [2]
Manuscript received January 8, 2018; revised May 13, 2018; accepted and ISA100.11a [3]) were released. However, they face limita-
June 28, 2018. Date of publication July 2, 2018; date of current ver- tions in terms of scalability and coverage when very large areas
sion November 1, 2018. Paper no. TII-18-0060. (Corresponding author:
Mikael Gidlund.) need to be covered. While cellular technologies, such as 3/4/5G
E. Sisinni is with the Department of Information Engineering, University technologies, promise to connect massive devices over long dis-
of Brescia, Brescia 25123, Italy (e-mail: [email protected]). tances, they require infrastructure support and licensed band [4].
A. Saifullah is with the Department of Computer Science, Wayne State
University, Detroit, MI 48202 USA (e-mail: [email protected]) IIoT applications typically require relatively small throughput
S. Han is with the Department of Computer Science and Engineer- per node and the capacity is not a main concern. Instead, the need
ing, University of Connecticut, Storrs, CT 06269 USA (e-mail: song. for connecting a very large number of devices to the Internet
[email protected]).
U. Jennehag and M. Gidlund are with the Department of Information at low cost, with limited hardware capabilities and energy re-
Systems and Technology, Mid Sweden University, Sundsvall SE-851 70, sources (e.g., small batteries) makes latency, energy efficiency,
Sweden (e-mail: [email protected]; [email protected]). cost, reliability, and security/privacy more desired features [5].
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org. Meeting the above mentioned requirements poses a number
Digital Object Identifier 10.1109/TII.2018.2852491 of key challenges on the evolution of IIoT. Addressing these
1551-3203 © 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
SISINNI et al.: INDUSTRIAL IOT: CHALLENGES, OPPORTUNITIES, AND DIRECTIONS 4725

challenges is critical in order to ensure a massive rollout of IIoT TABLE I

COMPARISON BETWEEN CONSUMER IOT AND IIOT
technologies. In this paper, we clarify the concepts of IoT, IIoT,
and the current trend of automation and data exchange in man-
ufacturing technologies called Industry 4.0. We highlight the
opportunities brought in by IIoT as well as the challenges for
its realization. In particular, we focus on the challenges associ-
ated with the need of energy efficiency, real-time performance,
coexistence, interoperability, and with the security and privacy
issues. We also provide a systematic overview of the state-of-
the-art research efforts and potential future research directions
to address IIoT challenges.
The rest of this paper is organized as follows. Section II
compares IoT, IIoT, and Industry 4.0. Section III provides an
overview of the recent activities on the definition of the IIoT ar-
chitecture, protocol stack, as well as the standardization efforts.
Section IV describes opportunities that IIoT will offer and chal-
lenges that have to be solved. Finally, we give some concluding production plants) and innovative approaches for self-
remarks in Section V. organizing systems (e.g., autonomic industrial plant that re-
quires little, if any, human intervention) [9].
While the most general communication requirements of IoT
II. IOT, IIOT, AND INDUSTRY 4.0
and IIoT are similar, e.g., support for the Internet ecosystem
IoT, IIoT, and Industry 4.0 are closely related concepts but using low-cost, resource-constrained devices and network scal-
cannot be interchangeably used. In this section, we provide a ability, many communication requirements are specific to each
rough classification of these terms. Regarding the IoT, several domain and can be very different, e.g., quality of service (QoS)
definitions exist, each one trying to capture one of its funda- (in terms of determinism, latency, throughput, etc.), the avail-
mental characteristics. It is often considered as a sort of web ability and reliability, and the security and privacy. IoT focuses
for the machines, highlighting the aim of allowing things to more on the design of new communication standards which can
exchange data. However, application fields are so diverse that connect novel devices into the Internet ecosystem in a flexible
some requirements (especially those related to communication and user-friendly way. By contrast, the current design of IIoT
aspects) can be very different, depending on the intended goals emphasizes on possible integration and interconnection of once
and end users, the underlying business models and the adopted isolated plants and working islands or even machineries, thus
technological solutions. What is usually addressed as IoT, could offering a more efficient production and new services [9]. For
be better named as consumer IoT, as opposed to IIoT [6], [7]. this reason, compared with IoT, IIoT can be considered more
Consumer IoT is human centered; the “things” are smart con- an evolution rather than a revolution. Table I gives a qualitative
sumer electronic devices interconnected with each other in order comparison of these technologies.
to improve human awareness of the surrounding environment, Regarding the connectivity and criticality, IoT is more flexi-
saving time and money. In general, consumer IoT communica- ble, allowing ad hoc and mobile network structures, and having
tions can be classified as machine-to-user and in the form of less stringent timing and reliability requirements (except for
client–server interactions. medical applications). On the other hand, IIoT typically em-
On the other hand, in the industrial world, we are assisting to ploys fixed and infrastructure-based network solutions that are
the advent of the digital and smart manufacturing, which aims at well designed to match communication and coexistence needs.
integrating operational technology with information technology In IIoT, communications are in the form of M2M links that
(IT) domains [8]. In very few words, the IIoT (the basic pillar have to satisfy stringent requirements in terms of timeliness and
of digital manufacturing) is about connecting all the industrial reliability. In process automation domain, process monitoring
assets, including machines and control systems, with the infor- and control application can be grouped into three subcategories:
mation systems and the business processes. As a consequence, monitoring/supervision, closed-loop control, and interlocking
the large amount of data collected can feed analytic solutions ad control. While monitoring and supervision applications are
and lead to optimal industrial operations. On the other hand, less sensitive to packet loss and jitter, and can tolerate trans-
smart manufacturing obviously focuses on the manufacturing mission delay at second level, closed-loop control and inter-
stage of (smart) product life cycle, with the goal of quickly locking, and control applications require bounded delay at mil-
and dynamically responding to demand changes. Therefore, the lisecond level (10−100 ms) and a transmission reliability of
IIoT affects all the industrial value chain and is a requirement 99.99% [5].
for smart manufacturing. Comparing the data volume, the generated data from IoT is
As underlined in the following, communication in IIoT is heavily application dependent, while IIoT currently targets at
machine oriented, and can range across a large variety of differ- analytics, e.g., for predictive maintenance and improved logis-
ent market sectors and activities. The IIoT scenarios include tics. This implies that very large amount of data are exchanged
legacy monitoring applications (e.g., process monitoring in in IIoT. For example, it is reported that the Rio Tinto mine
4726 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 14, NO. 11, NOVEMBER 2018

generates up to 2.4 TB of data per minute, according to Cisco A. IIoT Architecture

Global Cloud Index.
A reference architecture is a higher level of abstraction de-
The concept of Industry 4.0 (where 4.0 represents the fourth scription that helps identify issues and challenges for different
industrial revolution) arises when the IoT paradigm is merged
application scenarios. The design of an IIoT architecture needs
with the CPSs idea [10]. Originally defined in Germany, the
to highlight extensibility, scalability, modularity, and interoper-
Industry 4.0 concept has gained a global visibility and it is ability among heterogeneous devices using different technolo-
nowadays universally adopted for addressing the use of Internet
gies. Several reference architecture frameworks originated in
technologies to improve production efficiency by means of smart
the past in different application contexts for both IoT and IIoT
services in smart factories. CPSs extend real-world, physical [17]. The typically adopted approach is a multilayer description
objects by interconnecting them altogether and providing their
organized around the services offered at each level, depending
digital descriptions. Such information, stored in models and on the selected technologies, business needs, and technical re-
data objects that can be updated in real time, represents a second quirements. For instance, the International Telecommunication
identity of the object itself and constitutes a sort of “digital twin.”
Union supports an IoT architecture made of five layers: sens-
Thanks to the dynamic nature of these digital twins, innovative ing, accessing, networking, middleware, and application layers.
services, that were not possible in the past, can be implemented Jia et al. [18], Domingo [19], and Atzori et al. [20] suggested the
across the whole product life cycle, from inception to disposal
identification of three major layers for IoT: perception layer (or
of manufactured products. In summary, IIoT is a subset of IoT sensing layer), network layer, and service layer (or application
which is specific to industrial applications. The manufacturing layer). Liu et al. [21] designed an IoT application infrastruc-
phase of the product life cycle is where the IoT and Industry 4.0
ture that contains the physical layer, transport layer, middleware
meet, originating to the IIoT. Fig. 1 shows intersections of IoT, layer, and applications layer. In [22], a four-layered architecture
CPS, IIoT, and Industry 4.0.
is derived from the perspective of offered functionalities, that in-
As a concluding remark, it has to be highlighted that the IIoT
cludes the sensing layer, the networking layer, the service layer,
paradigm is not intended for substituting traditional automa- and the interface layer. The Reference Architectural Model In-
tion applications, but aims at increasing the knowledge about
dustrie 4.0 (RAMI 4.0) [23] focuses on next generation indus-
the physical system of interest. As a consequence, the IIoT (at
trial manufacturing systems; it identifies a three-dimensional
least today) is not related to control applications at the field model whose axes are the life cycle and value stream, related to
level, where bounded reaction time (i.e., determinism) must be
products’ life cycle, and the hierarchy levels, related to the dif-
ensured. On the contrary, as previously stated, IIoT applica-
ferent component functionalities. The hierarchy axis describes
tions, including supervision, optimization, and prediction activ- the IT representative and includes a communication layer.
ities, are typically grouped into the so-called digital or cloud
Recently, the Industrial Internet Consortium released the
manufacturing (CM). The growing interest toward this topic “Reference Architecture” document [24]. In particular, it fo-
is confirmed by the wide range of literature. A survey about cuses on different viewpoints (formally business, usage, func-
CM is reported in [11]. In the past, the supervision activities
tional, and implementation views) and provides models for each
were dominated by man, but efficient M2M communications one. The implementation viewpoint is focused on the technolo-
make human intervention superfluous and extend the operat- gies and the system components that are required for imple-
ing range to geographical scale. For instance, the availability of
menting the functionalities prescribed by the usage and func-
reliable, short-latency connections on such a large scale may tional viewpoints. Thus, it provides not only the description of
increase the revenue [12]. The work in [13] highlights the the IIoT system general architecture (i.e., its structure and the
importance of real-time, large-scale approach for equipment-
distribution of components, and the topology by which they
maintenance applications. An IIoT-based dynamic production are interconnected), but includes a description of interfaces and
logistics architecture is presented in [14] for real-time synchro-
protocols as well. Roughly speaking, two different kinds of in-
nization of internal and public production logistic resources. In
formation are transferred in IIoT systems, depending on if the
[15], the optimization of production scheduling is based on IIoT data have to be processed yet (data flow) or they are the results
decentralized energy prediction algorithms fed by the current
of some elaborations (control flow).
state of the machines. As a concluding remark, the progressive
Some architectural patterns are also emerging and provid-
reduction of latency and jitter of Internet-based connectivity ing coherent system implementations and paving the way to
will increase the range of possible applications, as reported
innovative business models and services, usually in a multiple-
in [16]. tier arrangement, dictated by the very heterogeneous devices
and networks. In the widely accepted three-tier pattern [25],
III. STATE OF THE ART
edge, platform, and enterprise tiers are connected by proxim-
As IIoT interconnects a large number of component leverag- ity, access, and service networks. The edge defines the domain
ing sensing, communication, and data processing technologies, in which IIoT components interact with each other. Thus, it
it is not possible to have a comprehensive description of all consists of sensors, controllers, and actuators interconnected
the recent advancements in such a diverse field. However, some by independent local area networks (the proximity networks,
foundational aspects can be highlighted, i.e., the architecture, usually in the form of fieldbuses) to an edge gateway, which
the connectivity, and the standardization. in turn connects to larger networks (access network) of the
SISINNI et al.: INDUSTRIAL IOT: CHALLENGES, OPPORTUNITIES, AND DIRECTIONS 4727

Fig. 2. Three-tier IIoT architecture.

platform tier, providing global coverage. Finally, the platform

tier leverages on the service network to establish links with the
enterprise tier that implements domain-specific applications and
provides end-user interfaces. Fig. 2 tries to graphically show the Fig. 3. Hourglass-shaped IIoT protocol stack.
complexity of the IIoT hybrid architecture; in particular, the in-
creased latency and data aggregation of the different tiers are
highlighted. ecosystem, thus providing poor interoperability. Fieldbuses
are vertical solutions covering most of the functionalities of
the communication stack. Fortunately, latest technologies (e.g.,
B. IIoT Connectivity the many different flavors of the real-time Ethernet solutions)
The connectivity of today’s IIoT varies depending on which natively adopt Ethernet and IP protocols, thus making it eas-
the combination of backbone and edge architecture is useful in ier to provide technical interoperability, i.e., the ability to share
a given situation, and a combination of wireless and/or wired packets in a common format [26]. Due to its full IP compat-
technologies is adopted. A key goal is to avoid isolated systems ibility and incorporation of the Common Industrial Protocol
based on proprietary solutions and enable data sharing and in- (CIP), and reliance on standard Internet and Ethernet tech-
teroperability among these closed subsystems (brownfield) and nology (IEEE802.3 combined with the TCP/IP Suite), Ether-
the yet-to-come applications (greenfield), within and across in- Net/IP makes itself particularly suitable for IIoT. As an exam-
dustries. Neither the seven-layer Open Systems Interconnect nor ple, myriads of motion applications in industries feature a bevy
the five-layer Internet model is adequate to take into account the of connected components—from I/O blocks and vision sensors
distributed nature of sensors, controllers, gateways, and other to servo and variable frequency drives. EtherNet/IP can unite all
components involved in IIoT and different layering is required. of these moving parts via CIP communications running on Eth-
The IIoT initiative feasibility requires communication protocols ernet [27]. Since it is built on the IP suite, EtherNet/IP is gaining
that are able to support efficient, timely, and ubiquitous informa- momentum from the development and refinement of associated
tion aggregation and availability. Lower levels of the stack must protocols. In addition to TCP/UDP at the transport layer, it can
adequately respond to scalability and flexibility requirements. access higher-level functionality through HTTP. Connectivity
Upper levels must allow so-called “smart devices” (i.e., offering between industrial equipment, Ethernet networks, and the In-
both computation and communication capabilities) to transport ternet can enable time-sensitive communications to streamline
“smart data”, not limited to the information of interest but also plant operations, thereby enabling real-time manufacturing for
providing awareness of the users they are intended to and all enterprises with global supply chains.
the semantic rules to be correctly understood at destinations 1) Stack Lower Layers: In IIoT stack, the lowest layer is the
as well. Three macro layers can be identified, i.e., networking physical one, which refers to the exchange of physical signals on
(dealing with frames and packets), connectivity (dealing with media linking the participants. Above it lies the link layer, which
messages), and information (dealing with end-user data struc- connects adjacent participants allowing to exchange frames by
tures). The protocol heterogeneity of the IIoT is mirrored in a means of signaling protocols. It has to be noticed that the al-
hourglass-shaped stack (see Fig. 3). The neck is represented by ready available solutions explicitly designed for the industrial
the network layer, i.e., the Internet (and its different flavors, as market have some limits. Well-accepted standards defined in
IPv4, IPv6, 6LowPAN, RPL, etc.), but above and below sub- the IEC62591 and IEC62743 (commercially known as Wire-
layers are not yet clearly defined, despite they are of critical lessHART and ISA100.11a) are based on IEEE802.15.4 com-
importance for ensuring interoperability at different levels. pliant radio and are not designed to connect a large number
Additionally, it is worth mentioning that most of current in- of devices, as in typical IIoT applications. Consequently, sev-
dustrial applications exploit fieldbuses, each having its own eral independent networks must be deployed, each one with
4728 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 14, NO. 11, NOVEMBER 2018

its own IIoT gateway. On the contrary, low-power wide-area the initial doubts for BLE was due to its range limitations since it
network (LPWAN) solutions are gaining momentum in recent only supports star network and limited number of devices [43].
years for occupying the lower two levels of the protocol stack, To overcome those limitations, BLE mesh networking standard
with multiple competing technologies being offered or under was recently released and initially considered for home automa-
development [28]. tion. The main challenge with BLE mesh networking targeting
LPWANs allow to communicate over long distances (several real-time communication is that the connection establishment
kilometers) at very low transmission power. SigFox [29] and procedure introduces a long delay (e.g., several hundred ms).
LoRaWAN [30] are two of the most interesting proposals To overcome this problem, many upper layer protocols, such as
[31]–[33]. However, SigFox, based on ultra narrowband mesh and beacon, try to leverage on the connectionless scheme
technology (i.e., communication channels with a bandwidth since there is no need to establish connections before sending
on the order of 100 Hz), is mainly intended for smart city data. However, this does not ensure reliable communication due
applications, e.g., smart metering, since a device can send at to lack of a good medium access control. Besides, the through-
most 140 messages per day, each one typically having 3 s air put is much lower than 1 Mb/s since there is a limitation of
time. Thus, it is not suitable for many industrial applications that sending packet in this bearer, i.e., at least 20 ms interval is re-
require real-time performance or frequent sampling. LoRaWAN quired in order to reduce intra-interference and avoid collisions.
(maintained by the LoRa alliance) leverages on proprietary Recently, there has been some interesting work about using BLE
LoRa radios and offers 125 kHz or 250 kHz wide channels and mesh networking for real-time communication targeting low la-
low data rate (from about 10 kb/s down to less than 400 b/s). tency applications in industrial automation. In [44], the authors
It has been demonstrated that by mimicking the time-slotted presented a real-time protocol aimed to overcome the problem
channel hopping of typical wireless industrial communications, with range limitations of mesh technology and support bounded
thousands of communication opportunities per second are real-time traffic. Their protocol exploits time division multiple
affordable [34]. As a final remark, it has to be highlighted that access with an optimized transmission allocation to provide data
LPWANs generally operate in the sub-GHz region, that ensures packets with real-time support. It works on standard BLE de-
good coverage, but is often limited by duty-cycled transmission vices. In [45], the authors presented a bandwidth reservation
of 1% or 0.1%, or “Listen Before Talk” (LBT) medium access mechanism for partitioning the radio transceiver between two
strategy. Also, both SigFox and LoRaWAN are primarily uplink protocols, namely the BLE and a real-time custom protocol.
only. LoRaWAN can enable bidirectional communication, but 2) Stack Upper Layers: The aim of upper layers of the IIoT
it has to rely on time-synchronized beacons and schedules, stack is to facilitate/ensure so called syntactic interoperability,
which is an overhead. The recently developed Sensor Network i.e., the capability to use a common data structure and set of
Over White space (SNOW) [35]–[38] is an LPWAN that rules for information exchanges [46], [47]. It is the actual ap-
enables concurrent bidirectional communications, thus making plication that finally provides the semantic interoperability, i.e.,
it suitable for control applications. However, SNOW operates the capability to interpret exchanged data unambiguously [26].
over the TV white spaces and thus its performance depends on In light of this requirement, the Industrial Internet Consortium
the availability of white spaces. proposed to separate upper layer protocols into just two levels;
The use of unlicensed spectrum has raised certain reliabil- the lower is occupied by the transport layer, that is in charge
ity issues, since there is no guarantee of service availability, of exchanging variable length messages among the involved
in addition to the aforementioned duty cycle and LBT regula- applications; the upper constitutes the framework layer, which
tions. For this reason, fifth generation cellular access (5G) is manages the transfer of structured data having higher abstraction
often envisioned as a viable IIoT solution, in addition to reg- (e.g., state, events, streams, etc.). According to this classifica-
ular telecommunication applications using the cellular infras- tion, the transport layer is loosely related to the transport layer of
tructure. Currently there is no finalized standard for 5G (which OSI (and Internet) model; indeed UDP and TCP are foundations
actually is an umbrella for many specifications). However, the for other transport protocols. However, some functionalities of
cost of technical solutions to be applied at the physical layer the session, presentation, and application layers are included as
to satisfy industrial needs can be an important issue. Only a well.
sound business model and a strong argument for using licensed A well accepted and widely used solution for implementing
frequency bands (both missing today) could bring market ac- horizontal integration relies on messaging protocols (often im-
ceptance within industrial automation for 5G [4]. Narrowband plemented by message-oriented middleware). These protocols
LPWAN technology standard to operate on cellular infrastruc- support the publisher/subscriber paradigm, where both sides of
ture and bands as NB-IoT received attention recently, but despite the actual data exchange are, in general, not directly connected.
its potential, there are some issues regarding scalability and The application that wants to publish a message connects to a
network resource slicing between IoT applications and other so-called message queue broker for placing it in a queue; subse-
broadband services that need further studies [39]. In licensed quently, subscribers automatically receive the message as a push
cellular spectrum, EC-GSM-IoT [40] and LTE Cat M1 (LTE- notification. The delivering modality is said to be persistent if it
Advanced Pro) [41] are also under development. A key require- survives a broker failure. Messaging solutions ensure scalability
ment of all these technological solutions is that they need cellular since the applications do not have to know each other. Today,
infrastructure. a prevailing messaging protocol is Message Queue Telemetry
Bluetooth Low Energy (BLE) [42] is another interesting alter- Transport, standardized by the Advancing Open Standards for
native for IIoT since it offers ultra-low power consumption, but the Information Society (OASIS) alliance. A different approach
SISINNI et al.: INDUSTRIAL IOT: CHALLENGES, OPPORTUNITIES, AND DIRECTIONS 4729

relies on request/response data delivery, and synchronous or in URLLC is to provide a reliable transmission time interval
asynchronous data exchanges are permitted. In the synchronous operation.
data exchanges, the requestor waits for replies before issuing Considering that a relevant part of IIoT communications will
the next request. In an asynchronous case, the reply is returned probably be implemented as wireless links, coexistence issues
at some unknown later time to the requestor. A well-known ex- arise as well. The IEC62657 provides a sort of glossary of in-
ample of request/response protocol is Constrained Application dustrial automation requirements for harmonizing concepts and
Protocol defined by the the IETF Constrained RESTful Envi- terms of the telecommunication world and defines coexistence
ronments working group [47]. parameters (in the form of templates) and guidelines for ensuring
The framework layer provides services to the above applica- wireless coexistence within industrial automation applications
tion and manages the life cycle of any piece of data from the along the whole life cycle of the plant.
creation to the deletion. Protocols at this level offer the ability
to discover and identify data objects and can understand the
transported data meaning (i.e., are not opaque). This aware- IV. OPPORTUNITIES AND CHALLENGES
ness is exploited for optimally delivering the information at the A key reason for adopting IIoT by manufacturers, utility
destination. The open platform communications-unified archi- companies, agriculture producers, and healthcare providers is to
tecture (a multipart document set managed by the OPC founda- increase productivity and efficiency through smart and remote
tion, formally known as the IEC62541) is an example of such management. As an example, Thames Water [48], the largest
a framework. It describes a service-oriented architecture based provider of drinking and waste water services in the U.K.,
on client/server architecture in which the server models data, in- is using sensors, and real-time data acquisition and analytics
formation, processes, and systems as objects that are presented to anticipate equipment failures and provide fast response to
to clients together with services that the client can use. critical situations, such as leaks or adverse weather events. The
utility firm has already installed more than 100 000 smart meters
in London, and it aims to cover all customers with smart meters
by 2030. With more than 4200 leaks detected on customer pipes
C. Standardization of IIoT
so far, this program has already saved an estimated 930 000 L of
Standardization is an important step for a technology to be water per day across London. As another example, the deploy-
widely supported and well accepted. Interesting to note, most ment of 800 HART devices for real-time process management at
of the past standardization activities focused on very specific Mitsubishi chemical plant in Kashima, Japan, has been increas-
domains, thus resulting in disjoint and somewhat redundant ing the production performance by saving US$20–30 000 per
development. The standardization process has to face several day that also averted a $3-million shutdown [49].
challenges; currently, there is a plethora of competing standard- Precision agriculture powered by IIoT can help farmers to
ization bodies and consortia initiatives at every layer of the IIoT better measure agricultural variables, such as soil nutrients,
stack referring to a variety of fragmented, often inconsistent, fertilizer used, seeds planted, soil water, and temperature of
and opponent requirements. Obviously, such an approach is stored produce, allowing to monitor down to the square foot
detrimental to IIoT, whose fundamental aim is to bring together through a dense sensor deployment, thereby almost doubling
and share information coming from very heterogeneous things. the productivity [50]–[52]. Companies like Microsoft (Farm-
The actual fragmentation is effectively highlighted by the ETSI Beats project [53], [54]), Climate Corp [55], AT&T [56], and
technical report ETSI TR 103375, whose aim is to provide the Monsanto [57] are promoting agricultural IoT. IIoT can also
roadmaps of the IoT standards. Generally speaking, the ongoing significantly impact the healthcare field. In hospitals, human or
standardization activities include horizontal standards, aiming technological errors caused by false alarms, slow response, and
at ensuring interoperability; vertical standards, aiming at iden- inaccurate information are still a major reason of preventable
tifying requirements of individual applications and use cases; death and patient suffering. By connecting distributed medi-
and promotional activities, supported by industrial consortia and cal devices using IIoT technologies, hospitals can significantly
government groups. overcome such limitations, thereby improving patient safety and
Focusing on industrial applications, the most significant and experiences, and more efficiently using the resources.
important efforts are those carried out by the International Elec- IIoT also provides opportunities to enhance efficiency, safety,
trotechnical Commission (IEC), which created many different and working conditions for workers. For example, using un-
study groups and technical committees on the subject and pub- manned aerial vehicles allows inspecting oil pipelines, moni-
lished a couple of white papers about IIoT and the smart factory toring food safety using sensors, and minimizing workers’ ex-
with the aim of assessing potential global needs, benefits, con- posure to noise and hazardous gases or chemicals in industrial
cepts, and preconditions for the factory of the future. It is worth environments. Schlumberger [58], for example, is now monitor-
noting that, regarding the connectivity issues, the aforemen- ing subsea conditions using unmanned marine vehicles, which
tioned IEC62541 is the only standard originated in the industrial can travel across oceans collecting data for up to a year without
vertical context. fuel or crew, moving under power generated from wave en-
Standardization activities for 5G targeting IIoT and critical ergy. Through remote monitoring and sensing powered by IIoT,
communication is ongoing in 3GPP and falls under the umbrella mining industries can dramatically decrease safety-related inci-
of Ultra Reliable Low Latency Communications (URLLC) with dents, while making mining in harsh locations more economical
the aim of providing 1 ms latency. One way to reduce the latency and productive. For example, Rio Tinto, a leading mining com-
4730 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 14, NO. 11, NOVEMBER 2018

pany, intends its automated operations in Australia to preview a how well it satisfies the end-to-end (e2e) deadlines of the real-
more efficient future for all of its mines to reduce the need for time sensing and control tasks executed in the system [62], [63].
human miners[59]. Time-slotted packet scheduling in IIoT plays a critical role in
Despite the great promise, there are many challenges in re- achieving the desired QoS. For example, many industrial wire-
alizing the opportunities offered by IIoT, which should be ad- less networks perform network resource management via static
dressed in the future research. The key challenges stem from data link layer scheduling [64]–[71] to achieve deterministic
the requirements in energy-efficient operation, real-time perfor- e2e real-time communication. Such approaches typically take a
mance in dynamic environments, the need for coexistence and periodic approach to gather the network health status, and then
interoperability, and maintaining the security of the applications recompute and distribute the updated network schedule infor-
and users’ privacy as described below. mation. This process however is slow, not scalable, and incurs
considerable network overhead. The explosive growth of IIoT
applications, especially in terms of their scale and complexity,
A. Energy Efficiency has dramatically increased the level of difficulty in ensuring the
Many IIoT applications need to run for years on batteries. desired real-time performance. The fact that most IIoT must deal
This calls for the design of low-power sensors, which do not with unexpected disturbances further aggravate the problem.
need battery replacement over their lifetimes. This creates a Unexpected disturbances can be classified into external dis-
demand for energy-efficient designs. To complement such de- turbances from the environment being monitored and controlled
signs, upper layer approaches can play important roles through (e.g., detection of an emergency, sudden pressure, or temper-
energy-efficient operation. Many energy-efficient schemes for ature changes), and internal disturbances within the network
wireless sensor network (WSN) have been proposed in recent infrastructure (e.g., link failure due to multiuser interference
years [60], but those approaches are not immediately applicable or weather related changes in channel signal-to-noise ratio).
to IIoT. IIoT applications typically need a dense deployment of In response to various internal disturbances, many centralized
numerous devices. Sensed data can be sent in queried form or in scheduling approaches [72]–[77] have been proposed. There are
a continuous form, which in a dense deployment, can consume a also a few works on adapting to external disturbances in crit-
significant amount of energy. Green networking is thus crucial in ical control systems. For example, rate-adaptive and rhythmic
IIoT to reduce power consumption and operational costs. It will task models are introduced in [78] and [79], respectively, which
lessen pollution and emissions, and make the most of surveil- allow tasks to change periods and relative deadlines in some
lance and environmental conservation. LPWAN IoT technolo- control systems, such as automotive systems.
gies achieve low-power operation using several energy-efficient Given the requirement of meeting e2e deadlines, the afore-
design approaches. First, they usually form a star topology, mentioned approaches for handling unexpected disturbances are
which eliminates the energy consumed through packet routing almost all built on a centralized architecture. Hence, most of
in multihop networks. Second, they keep the node design simple them have limited scalability [80]. The concept of distributed
by offloading the complexities to the gateway. Third, they use resource management is not new. In fact, distributed approaches
narrowband channels, thereby decreasing the noise level and have been investigated fairly well in the wireless network com-
extending the transmission range [35], [61]. munity (e.g., [81]–[85]). However, these studies typically are
Although there are numerous methods to achieve energy ef- not concerned with real-time e2e constraints. A few, which con-
ficiency, such as using lightweight communication protocols sider real-time constraints, mainly focus on soft real-time re-
or adopting low-power radio transceivers, as described above, quirements and do not consider external disturbances that IIoT
the recent technology trend in energy harvesting provides an- must have to deal with. Only recently, we have started to see
other fundamental method to prolong battery life. Thus, en- some hybrid and fully distributed resource management ap-
ergy harvesting is a promising approach for the emerging proaches for IIoT [86], [87]. However, how to ensure bounded
IIoT. Practically, energy can be harvested from environmental response time to handle concurrent disturbances is still an open
sources, namely, thermal, solar, vibration, and wireless radio- problem.
frequency (RF) energy sources. Harvesting from such environ-
mental sources is dependent on the presence of the correspond-
ing energy source. However, RF energy harvesting may provide C. Coexistence and Interoperability
benefits in terms of being wireless, readily available in the form With the rapid growth of IIoT connectivity, there will be
of transmitted energy (TV/radio broadcasters, mobile base sta- many coexisting devices deployed in close proximity in the
tions, and hand-held radios), low cost, and in terms of small limited spectrum. This brings forth the imminent challenge
form factor of devices. of coexistence in the crowded Industrial, Scientific and Med-
ical (ISM) bands. Thus, interference between devices must
be handled to keep them operational. Existing and near-future
B. Real-Time Performance
IIoT devices will most likely have limited memory and intelli-
IIoT devices are typically deployed in noisy environments gence to combat interference or keep it to a minimum. While
for supporting mission- and safety-critical applications, and there exists much work on wireless coexistence considering
have stringent timing and reliability requirements on timely WiFi, IEEE802.15.4 networks, and Bluetooth (see surveys [88]–
collection of environmental data and proper delivery of control [91]), they will not work well for IIoT. Due to their dense
decisions. The QoS offered by IIoT is thus often measured by and large-scale deployments, these devices can be subject to
SISINNI et al.: INDUSTRIAL IOT: CHALLENGES, OPPORTUNITIES, AND DIRECTIONS 4731

an unprecedented number of interferers. Technology-specific communication among lightweight devices regarding CPU,
features of each IIoT technology may introduce additional memory, and energy consumption [102]. For this reason, tra-
challenges. ditional protection mechanisms are not sufficient to secure
To ensure good coexistence, it will become important that fu- the complex IIoT systems, such as secure protocols [103],
ture IIoT devices can detect, classify, and mitigate external inter- lightweight cryptography [104] and privacy assurance [105]. To
ference. Recently, some work regarding classifying interference secure the IIoT infrastructure, existing encryption techniques
via spectrum sensing [92] on IIoT devices has been presented but from industrial WSNs may be reviewed before they are applied
most of the existing work fails since a very long sampling win- to build IIoT secure protocols. For instance, scarce computing
dow is needed and the proposed spectrum sensing methods need and memory resources prevent the use of resource-demanding
much more memory than what is available in existing commer- crypto primitives, e.g., public-key cryptography (PKC). This
cial IIoT devices. Hence, in [93], a promising method was pre- challenge is more critical in the applications of massive data
sented and implemented in Crossbow’s TelosB mote CA2400, exchanged with real-time requirements. To address privacy and
which is equipped with Texas Instrument CC2420 transceiver. security threats in IIoT, one can argue for a holistic approach
That method manages to classify external interference by using as pointed out in [106]. This means that aspects, such as plat-
support vector machines with a sensing duration below 300 ms. form security, secure engineering, security management, iden-
Moreover, existing devices based on IEEE802.15.4 standards do tity management, and industrial rights management, must be
not have any forward error correcting (FEC) capabilities to im- taken into account, throughout the whole life cycle of the sys-
prove the reliability of the data packet. There exists some work tems and products.
that investigated error control codes for industrial WSNs and There exist several security properties to consider when de-
the results clearly show that FEC will improve reliability and signing secure IIoT infrastructure [107].
the coexistence [94]–[96]. However, most of the available FEC 1) IIoT devices need to be tamper resistant against potential
methods are optimized for long packets. Given that IIoT com- physical attacks, such as unauthorized re-programming
munication will mainly consist of short packets (50–70 bytes) and passive secret stealing, while allowing the authorized
and many applications are time critical, more research is needed users to update the security firmware on the device.
to find good error correcting codes for IIoT communication [97]. 2) The storage of IIoT device should be protected against
If the research of error correcting codes for IIoT devices should adversary by keeping the data encrypted to keep the con-
be successful, it is also important that more emphasis be given on fidentiality.
investigating and understanding the complex radio environment 3) The communication network among the IIoT devices
where many of these IIoT devices will be deployed [98], [99]. should be secured to keep confidentiality and integrity.
The rapid growth of IIoT technologies also brings forth the 4) The IIoT infrastructure needs efficient identification and
requirements of interoperability. Namely, in the future, a fully authorization mechanisms, so that only authorized enti-
functional digital ecosystem will require seamless data sharing ties can access the IIoT resource.
between machines and other physical systems from different 5) The system should be available within normal operation,
manufacturers. The lack of interoperability among IIoT devices even with the physical damage to the devices by malicious
will significantly increase the complexity and cost of IIoT de- users. This guarantees the robustness of IIoT.
ployment and integration. The drive towards seamless interoper- Typically, symmetric-key cryptography can provide a
ability will be further complicated by the long life span of typical lightweight solution for IIoT devices. However, both the key
industrial equipment, which would require costly retrofitting or storage and the key management are big issues if using
replacement to work with the latest technologies. symmetric-key encryption, especially when considering low-
The challenges of device diversity in IIoT can be addressed capacity devices.
along three dimensions: multimode radios, software flexibility, Additionally, if one device in IIoT is compromised, it may
and cross-technology communication [100]. Multimode radios leak all other keys. PKC generally provides more secure fea-
allow diverse IIoT devices to talk to each other. Software flexibil- tures, and low storage requirements, but suffers from high com-
ity enables support for multiple protocols, connectivity frame- putational overhead due to complex encryption. Thus, reducing
works and cloud services. Recently, cross-technology commu- the overhead of complex security protocols for public-key cryp-
nication [101], without the assistance of additional hardware, tosystems remains a major challenge for IIoT security. In PKC,
has been studied for communication across WiFi, ZigBee, and elliptic-curve cryptography provides a lightweight solution re-
Bluetooth devices. Such approaches are specific to technologies, garding computational resources. It provides a smaller key size,
and thus, future research is needed to enable cross-technology reducing storage, and transmission requirements.
communication in IIoT devices. In IIoT systems, it is important to provide the identification to
get the legal access. The secure IIoT infrastructure must ensure
the object identification regarding the integrity of records used
D. Security and Privacy in the naming systems, such as domain name system (DNS). The
Besides the requirements of energy efficiency and real-time DNS can provide name translation services to the Internet user,
performance, security is another critical concern in IIoT. In however, it is in an insecure way which remains vulnerable to
general, IIoT is a resource-constrained communication net- various attacks by deliberated adversary [108]. This challenge
work which largely relies on low-bandwidth channels for stays valid even for a bounded and closed environment. Thus,
4732 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 14, NO. 11, NOVEMBER 2018

without the integrity protection of the identification, the whole managed information. Furthermore, typical IIoT applications
naming system is still insecure. Security extensions to DNS, like, have to deal with constrained resources (both power and
Domain Name Service Security Extension (DNSSEC) increases computing) and must be operative for extended periods of time,
security and is documented in IETF RFC4033 [109]. However, ensuring availability and reliability. We have described the
due to its high computation and communication overhead, it is state-of-the-art research and standardization efforts, and future
challenging to directly apply DNSSEC to the IIoT infrastructure. research directions to address IIoT challenges.
IIoT devices should follow specific schemes and rules for au-
thentication to exchange/publish their data. Due to the resource REFERENCES
constraints of the IIoT devices, low-cost authentication schemes
[1] Ericsson. (2016, January). Cellular networks for massive IoT. [Online].
have not been provided as much as needed [110]. Although PKC Available: https://www.ericsson.com/assets/local/publications/white-
systems provide the methods for constructing authentication and papers/wp_iot.pdf
authorization schemes, it fails to provide a global root certifica- [2] F. Group. (2007). WirelessHART specification. [Online]. Available:
http://www.hartcomm2.org
tion authority (CA), which largely hinders many theoretically [3] International Society of Automation (ISA). ISA100: Wireless systems
feasible schemes from actually being deployed. Without provid- for automation. [Online]. Available: https://www.isa.org/store/ansi/
ing the global root CA, it becomes very challenging to design isa-10011a-2011-wireless-systems-for-industrial-automation-process-
control-and-related-applications/118261
a secure authentication system in IIoT. Thus, currently, if we [4] M. Gidlund, T. Lennvall, and J. Akerberg, “Will 5G become yet another
intend to provide the secure authentication for IIoT devices, we wireless technology for industrial automation?” in Proc. IEEE Int. Conf.
have to use the high-cost solutions which is a conflict with the Ind. Technol., 2017, pp. 1319–1324.
[5] J. Akerberg, M. Gidlund, and M. Bjorkman, “Future research challenges
main goal of the lightweight principle of IIoT [111]. Further- in wireless sensor and actuator networks targeting industrial automation,”
more, it is a big challenge to issue a certification to each object in Proc. 9th IEEE Int. Conf. Ind. Inform., 2011, pp. 410–415.
in IIoT since the total number of objects could be huge. [6] M. R. Palattella et al.,“Internet of things in the 5G era: Enablers, ar-
chitecture, and business models,” IEEE J. Sel. Areas Commun., vol. 34,
Privacy is a very broad and diverse concept. Many definitions no. 3, pp. 510–527, Mar. 2016.
and perspectives have been provided in the literature. Gener- [7] D. Bandyopadhyay and J. Sen, “Internet of things: Applications and
ally speaking, privacy in IIoT is the threefold guarantee [112] challenges in technology and standardization,” Wireless Pers. Commun.,
vol. 58, no. 1, pp. 49–69, 2011.
for: 1) awareness of privacy risks imposed by things and ser- [8] M. R. Palattella, P. Thubert, X. Vilajosana, T. Watteyne, Q. Wang, and
vices; 2) individual control over the collection and processing T. Engel, Internet of Things. IoT Infrastructures: Second International
of information; and 3) awareness and control of subsequent use Summit. Cham, Switzerland: Springer, 2016.
[9] L. D. Xu, W. He, and S. Li, “Internet of things in industries: A survey,”
and dissemination to any outside entity. The major challenges IEEE Trans. Ind. Inform., vol. 10, no. 4, pp. 2233–2243, Nov. 2014.
for privacy lie in two aspects: data collection process and data [10] M. Wollschlaeger, T. Sauter, and J. Jasperneite, “The future of industrial
anonymization process. Typically, data collection process deals communication: Automation networks in the era of the internet of things
and industry 4.0,” IEEE Ind. Electron. Mag., vol. 11, no. 1, pp. 17–27,
with the collectible data and the access control to these data dur- Mar. 2017.
ing the data collection from smart things; data anonymization is [11] W. He, and L. Xu, “A state-of-the-art survey of cloud manufacturing,”
a process to ensure data anonymity through both cryptographic Int. J. Comput. Integr. Manuf., vol. 28, no. 3, pp. 239–250, 2015. [Online].
Available: https://doi.org/10.1080/0951192X.2013.874595
protection and concealment of data relations. Due to the re- [12] I. Lee, “An exploratory study of the impact of the internet of things (IoT)
strictions on the collection and storage of private information, on business model innovation: Building smart enterprises at fortune 500
privacy preservation can be ensured during the data collection. companies,” Int. J. Inf. Syst. Soc. Chang., vol. 7, no. 3, pp. 1–15, Jul. 2016.
[Online]. Available: http://dx.doi.org/10.4018/IJISSC.2016070101
However, given the diversity of the things in data anonymiza- [13] P. O’Donovan, K. Leahy, K. Bruton, and D. T. J. O’Sullivan, “An indus-
tion, different cryptographic schemes may be adopted, which trial big data pipeline for data-driven analytics maintenance applications
is a challenge to privacy preserving. Meanwhile, the collected in large-scale smart manufacturing facilities,” J. Big Data, vol. 2, no. 1,
p. 25, Nov. 2015. [Online]. Available: https://doi.org/10.1186/s40537-
information needs to be shared among the IIoT devices, and 015-0034-z
the computation on encrypted data is another challenge for data [14] T. Qu, S. P. Lei, Z. Z. Wang, D. X. Nie, X. Chen, and G. Q. Huang, “IoT-
anonymization. based real-time production logistics synchronization system under smart
cloud manufacturing,” Int. J. Adv. Manuf. Technol., vol. 84, no. 1, pp. 147–
164, Apr. 2016. [Online]. Available: https://doi.org/10.1007/s00170-015-
7220-1
V. CONCLUSION [15] S. G. Pease et al., “An intelligent real-time cyber-physical toolset
for energy and process prediction and optimisation in the future in-
This paper presented an overview of the emerging IIoT dustrial internet of things,” Future Gener. Comput. Syst., vol. 79,
solutions. What is proposed as a revolution for the consumer pp. 815–829, 2018. [Online]. Available: http://www.sciencedirect.com/
market can be another step of the ever evolving industrial science/article/pii/S0167739X1630382X
[16] T. H. Szymanski, “Supporting consumer services in a deterministic in-
communication world. Several technologies are involved and dustrial internet core network,” IEEE Commun. Mag., vol. 54, no. 6,
terms, such as IoT, IIoT and Industry 4.0, are often misused. pp. 110–117, Jun. 2016.
In this paper, we have provided a systematic overview of IIoT, [17] M. Weyrich and C. Ebert, “Reference architectures for the internet of
things,” IEEE Softw., vol. 33, no. 1, pp. 112–116, Jan./Feb. 2016.
focusing on the definition of its architecture and describing the [18] X. Jia, Q. Feng, T. Fan, and Q. Lei, “Rfid technology and its applications
protocol ecosystem which is emerging from standardization in internet of things (IoT),” in Proc. 2nd Int. Conf. Consum. Electron.,
efforts. We have also discussed the challenges for its realiza- Commun. Netw., 2012, pp. 1282–1285.
[19] M. C. Domingo, “An overview of the internet of things for people with
tion. Besides the QoS requirements that characterize industrial disabilities,” J. Netw. Comput. Appl., vol. 35, no. 2, pp. 584–596, 2012.
communications, IIoT suffers from yet to be considered [20] L. Atzori, A. Iera, and G. Morabito, “The internet of things: A survey,”
security challenges that stem from the high sensitivity of the Comput. Netw., vol. 54, no. 15, pp. 2787–2805, 2010.
SISINNI et al.: INDUSTRIAL IOT: CHALLENGES, OPPORTUNITIES, AND DIRECTIONS 4733

[21] C. H. Liu, B. Yang, and T. Liu, “Efficient naming, addressing and pro- [46] A. Al-Fuqaha, A. Khreishah, M. Guizani, A. Rayes, and M. Moham-
file services in internet-of-things sensory environments,” Ad Hoc Netw., madi, “Toward better horizontal integration among IoT services,” IEEE
vol. 18, pp. 85–101, 2014. Commun. Mag., vol. 53, no. 9, pp. 72–79, Sep. 2015.
[22] L. Da Xu, W. He, and S. Li, “Internet of things in industries: A sur- [47] A. Al-Fuqaha, M. Guizani, M. Mohammadi, M. Aledhari, and M.
vey,” IEEE Trans. Ind. Inform., vol. 10, no. 4, pp. 2233–2243, Nov. Ayyash, “Internet of things: A survey on enabling technologies, pro-
2014. tocols, and applications,” IEEE Commun. Surv. Tut., vol. 17, no. 4,
[23] H. Flatt, S. Schriegel, J. Jasperneite, H. Trsek, and H. Adamczyk, “Anal- pp. 2347–2376, Fourthquarter 2015.
ysis of the cyber-security of industry 4.0 technologies based on RAMI [48] J. P. Tomas (2017). Thames water rolls out smart meter project in Lon-
4.0 and identification of requirements,” in Proc. IEEE 21st Int. Conf. don. [Online]. Available: https://wiprodigital.com/cases/progressive-
Emerg. Tech. Factory Autom., 2016, pp. 1–4. metering-a-utilitys-strategi c-move-into-predictive-planning/
[24] Industrial Internet Consortium. Industrial internet reference architecture. [49] [Online]. Available: https://fieldcommgroup.org/sites/default/files/plant_
[Online]. Available: http://www.iiconsortium.org/IIRA.htm of_the_year/PotY_2009_Mitsubishi_Chem.pdf
[25] International Electrotechnical Commission (IEC), IoT 2020: Smart and [50] M. H. Almarshadi and S. M. Ismail, “Effects of precision irrigation on
Secure IoT Platform. IEC, Switzerland, 2016. productivity and water use efficiency of alfalfa under different irrigation
[26] J. Kiljander et al., “Semantic interoperability architecture for pervasive methods in arid climates,” J. Appl. Sci. Res., vol. 7, no. 3, pp. 299–308,
computing and internet of things,” IEEE Access, vol. 2, pp. 856–873, 2011.
2014. [51] H.-J. Kim, K. A. Sudduth, and J. W. Hummel, “Soil macronutrient sens-
[27] EtherNet/IP Automation meets the Internet of Things. [On- ing for precision agriculture,” J. Environmental Monit., vol. 11, no. 10,
line]. Available: http://www.industrial-ip.org/en/industrial-ip/ethernet- pp. 1810–1824, 2009.
ip/ethernetip-info graphic [52] N. D. Mueller, J. S. Gerber, M. Johnston, D. K. Ray, N. Ramankutty, and
[28] D. Ismail, M. Rahman, and A. Saifullah, “Low-power wide-area net- J. A. Foley, “Closing yield gaps through nutrient and water management,”
works: Opportunities, challenges, and directions,” in Proc. Workshop Nature, vol. 490, no. 7419, pp. 254–257, 2012.
Program 19th Int. Conf. Distrib. Comput. Netw., 2018, pp. 8:1–8:6. [53] D. Vasisht et al., “Farmbeats: An IoT platform for data-driven agri-
[29] Sigfox. Sigfox - the global communications service provider for the culture,” in Proc. 14th USENIX Symp. Netw. Syst. Des. Implementation,
internet of things (iot). [Online]. Available: http://sigfox.com 2017, pp. 515–529.
[30] LoRa Alliance. LoRaWAN. [Online]. Available: https://www.lora- [54] Microsoft. FarmBeats: IoT for agriculture. [Online]. Avail-
alliance.org able: https://www.microsoft.com/en-us/research/project/farmbeats-iot-
[31] W. Yang et al., “Narrowband wireless access for low-power massive agriculture/
internet of things: A bandwidth perspective,” IEEE Wireless Commun., [55] C. Corporation. Data-driven agricultural decisions and insights to maxi-
vol. 24, no. 3, pp. 138–145, Jun. 2017. mize every acre. [Online]. Available: https://www.climate.com
[32] P. Ferrari, A. Flammini, M. Rizzi, E. Sisinni, and M. Gidlund, “On [56] AT&T M2X. Agriculture IoT software as a service (SaaS).
the evaluation of LoRaWAN virtual channels orthogonality for dense [Online]. Available: https://m2x.att.com/iot/industry-solutions/iot-data/
distributed systems,” in Proc. IEEE Int. Workshop Meas. Netw., 2017, agriculture/
pp. 1–6. [57] J. Hawn, “Agricultural IoT promises to reshape farming.,” RCR
[33] M. Rizzi, P. Ferrari, A. Flammini, and E. Sisinni, “Evaluation of Wireless News, Nov. 2015. [Online]. Available: https://www.internet-of-
the IoT LoRaWAN solution for distributed measurement applica- things-research.eu/pdf/IERC_Position_Paper_IoT_Governance_Privacy_
tions,” IEEE Trans. Instrum. Meas., vol. 66, no. 12, pp. 3340–3349, Security_Final.pdf
Dec. 2017. [58] Schlumberger Robotics Services. [Online]. Available: https://www.slb.
[34] M. Rizzi, P. Ferrari, A. Flammini, E. Sisinni, and M. Gidlund, “Using com/∼/media/Files/production/brochures/flow_assurance/subsea_
LoRa for industrial wireless networks,” in Proc. IEEE 13th Int. Workshop monitoring_control_br.pdf
Factory Commun. Syst., 2017, pp. 1–4. [59] T. Simonite, “Mining 24 hours a day with robots,” MIT Tech-
[35] A. Saifullah, M. Rahman, D. Ismail, C. Lu, R. Chandra, and J. Liu, nol. Rev., Dec. 2016. [Online]. Available: https://www.internet-of-
“SNOW: Sensor network over white spaces,” in Proc. 14th ACM Conf. things-research.eu/pdf/IERC_Position_Paper_IoT_Governance_Privacy_
Embedded Netw. Sensor Syst., 2016, pp. 272–285. Security_Final.pdf
[36] A. Saifullah, M. Rahman, D. Ismail, C. Lu, J. Liu, and R. Chandra, [60] T. Rault, A. Bouabdallah, and Y. Challal, “Energy efficiency in wireless
“Enabling reliable, asynchronous, and bidirectional communication in sensor networks: A top-down survey,” vol. 67, pp. 104–122, 2014.
sensor networks over white spaces,” in Proc. 15th ACM Conf. Embedded [61] 3GPP. (2016, Jun.), Standardization of NB-IOT completed [On-
Netw. Sensor Syst., 2017, pp. 1–14. line]. Available: http://www.3gpp.org/news-events/3gpp-news/1785-nb_
[37] M. Rahman and A. Saifullah, “Integrating low-power wide-area net- iot_complete
works in white spaces,” in Proc. ACM/IEEE Conf. Internet-of-Things [62] P. Ferrari, A. Flammini, E. Sisinni, S. Rinaldi, D. Brandão, and M.S.
Des. Implementation, 2018, pp. 255–260. Rocha, “Delay estimation of industrial IoT applications based on mes-
[38] A. Saifullah, M. Rahman, D. Ismail, C. Lu, J. Liu, and R. Chandra, saging protocols,” IEEE Trans. Instrum. Meas., to be published, doi:
“Low-power wide-area networks over white spaces,” ACM/IEEE Trans. 10.1109/TIM.2018.2813798.
Netw., pp. 1–14, 2018. [63] T. Zheng, M. Gidlund, and J. Akerberg, “WirArb: A new MAC protocol
[39] Y. D. Beyene et al., “Nb-iot technology overview and experience from for time critical industrial wireless sensor network applications,” IEEE
cloud-ran implementation,” IEEE Wireless Commun., vol. 24, no. 3, Sensors J., vol. 16, no. 7, pp. 2127–2139, Apr. 2016.
pp. 26–32, Jun. 2017. [64] S. Han, X. Zhu, D. Chen, A. K. Mok, and M. Nixon, “Reliable and
[40] GSMA. (2016, Oct.). 3gpp low power wide area technologies. real-time communication in industrial wireless mesh networks,” in Proc.
[Online]. Available: https://www.gsma.com/iot/wp-content/uploads/ IEEE Real-Time Embedded Technol. Appl. Symp., 2011, pp. 3–12.
2016/10/3GPP-Low-Power-Wide-Area-Technologies-GSMA-White- [65] Q. Leng, Y.-H. Wei, S. Han, A. Mok, W. Zhang, and M. Tomizuka, “Im-
Paper.pdf proving control performance by minimizing jitter in RT-WiFi networks,”
[41] u-blox. Lte cat m1.’ [Online]. Available: https://www.u-blox.com/en/lte- in Proc. IEEE Real-Time Syst. Symp., 2014, pp. 63–73.
cat-m1 [66] A. Saifullah, C. Lu, Y. Xu, and Y. Chen, “Real-time scheduling for
[42] Bluetooth Core Specification Version 5.0, Bluetooth SIG, Kirkland, WA, WirelessHART networks,” in Proc. IEEE Real-Time Syst. Symp., 2010,
USA, 2016. pp. 150–159.
[43] R. Rondón, M. Gidlund, and K. Landernäs, “Evaluating bluetooth low [67] J. Song et al., “WirelessHART: Applying wireless technology in real-
energy suitability for time-critical industrial iot applications,” Int. J. Wire- time industrial process control,” in Proc. IEEE Real-Time Embedded
less Inf. Netw., vol. 24, no. 3, pp. 278–290, Sep. 2017. Technol. Appl. Symp., 2008, pp. 377–386.
[44] G. Patti, L. Leonardi, and L. L. Bello, “A bluetooth low energy real-time [68] Y.-H. Wei, Q. Leng, S. Han, A. K. Mok, W. Zhang, and M. Tomizuka,
protocol for industrial wireless mesh networks,” in Proc. IECON-42nd “RT-WiFi: Real-time high-speed communication protocol for wireless
Annu. Conf. IEEE Ind. Electron. Soc., 2016, pp. 4627–4632. cyber-physical control applications,” in Proc. IEEE Real-Time Syst.
[45] M. Marinoni, A. Biondi, P. Buonocunto, G. Franchino, D. Cesarini, and Symp., 2013, pp. 140–149.
G. Buttazzo, “Real-time analysis and design of a dual protocol support [69] A. Saifullah, Y. Xu, C. Lu, and Y. Chen, “End-to-end communication
for bluetooth LE devices,” IEEE Trans. Ind. Inform., vol. 13, no. 1, delay analysis in industrial wireless networks,” IEEE Trans. Comput.,
pp. 80–91, Feb. 2017. vol. 64, no. 5, pp. 1361–1374, May 2015.
4734 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 14, NO. 11, NOVEMBER 2018

[70] A. Saifullah et al., “Schedulability analysis under graph routing in [93] S. Grimaldi, A. Mahmood, and M. Gidlund, “An SVM-based method
WirelessHART networks,” in Proc. IEEE Real-Time Syst. Symp., 2015, for classification of external interference in industrial wireless sensor
pp. 165–174. and actuator networks,” J. Sensor Actuator Netw., vol. 6, no. 2, p. 9,
[71] A. Saifullah, S. Sankar, J. Liu, C. Lu, B. Priyantha, and R. Chandra, “Cap- 2017.
Net: A real-time wireless management network for data center power [94] F. Barac, M. Gidlund, and T. Zhang, “Scrutinizing bit- and symbol-
capping,” in Proc. IEEE Real-Time Syst. Symp., 2014, pp. 334–345. errors of IEEE802.15.4 communication in industrial environments,”
[72] O. Chipara, C. Lu, and G.-C. Roman, “Real-time query scheduling for IEEE Trans. Instrum. Meas., vol. 63, no. 7, pp. 1783–1794, Jul. 2014.
wireless sensor networks,” IEEE Trans. Comput., vol. 62, no. 9, pp. 1850– [95] Y. H. Yitbarek, K. Yu, J. Akerberg, M. Gidlund, and M. Bjorkman,
1865, Sep. 2013. “Implementation and evaluation of error control schemes in industrial
[73] O. Chipara, C. Wu, C. Lu, and W. Griswold, “Interference-aware real- wireless sensor networks,” in Proc. IEEE Int. Conf. Ind. Technol., 2014,
time flow scheduling for wireless sensor networks,” in Proc. 23rd Eu- pp. 730–735.
romicro Conf. Real-Time Syst., 2011, pp. 67–77. [96] F. Barac, M. Gidlund, and T. Zhang, “Ubiquitous, yet deceptive:
[74] T. L. Crenshaw, S. Hoke, A. Tirumala, and M. Caccamo, “Robust implicit Hardware-based channel metrics on interfered WSN links,” IEEE Trans.
EDF: A wireless MAC protocol for collaborative real-time systems,” Veh. Technol., vol. 64, no. 5, pp. 1766–1778, May 2015.
ACM Trans. Embedded Comput. Syst., vol. 6, no. 4, p. 28, 2007. [97] F. Barac, S. Caiola, M. Gidlund, E. Sisinni, and T. Zhang, “Channel di-
[75] A. Saifullah et al., “Near optimal rate selection for wireless control agnostics for wireless sensor networks in harsh industrial environments,”
systems,” ACM Trans. Embedded Comput. Syst., vol. 13, no. 4s, pp. 1– IEEE Sensors J., vol. 14, no. 11, pp. 3983–3995, Nov. 2014.
25, 2014. [98] P. Agrawal, A. Ahlen, T. Olofsson, and M. Gidlund, “Long term channel
[76] W. Shen, T. Zhang, M. Gidlund, and F. Dobslaw, “SAS-TDMA: A source characterization for energy efficient transmission in industrial environ-
aware scheduling algorithm for real-time communication in industrial ments,” IEEE Trans. Commun., vol. 62, no. 8, pp. 3004–3014, Aug.
wireless sensor networks,” Wireless Netw., vol. 19, no. 6, pp. 1155–1170, 2014.
2013. [99] T. Olofsson, A. Ahlen, and M. Gidlund, “Modeling of the fading statistics
[77] F. Dobslaw, T. Zhang, and M. Gidlund, “End-to-end reliability-aware of wireless sensor network channels in industrial environments,” IEEE
scheduling for wireless sensor networks,” IEEE Trans. Ind. Inform., Trans. Signal Process., vol. 64, no. 12, pp. 3021–3034, Jun. 2016.
vol. 12, no. 2, pp. 758–767, Apr. 2016. [100] L. Ascorti, S. Savazzi, G. Soatti, M. Nicoli, E. Sisinni, and S. Galimberti,
[78] G. C. Buttazzo, E. Bini, and D. Buttle, “Rate-adaptive tasks: Model, “A wireless cloud network platform for industrial process automation:
analysis, and design issues,” in Proc. IEEE Des., Autom. Test Europe Critical data publishing and distributed sensing,” IEEE Trans. Instrum.
Conf. Exhib., 2014, pp. 1–6. Meas., vol. 66, no. 4, pp. 592–603, Apr. 2017.
[79] J. Kim, K. Lakshmanan, and R. Rajkumar, “Rhythmic tasks: A new task [101] S. M. Kim and T. He, “Freebee: Cross-technology communication via
model with continually varying periods for cyber-physical systems,” in free side-channel,” in Proc. 21st Annu. Int. Conf. Mobile Comput. Netw.,
Proc. IEEE/ACM 3rd Int. Conf. Cyber-Physical Syst., 2012, pp. 55–64. 2015, pp. 317–330.
[80] C. Lu et al., “Real-time wireless sensor-actuator networks for industrial [102] T. Heer, O. Garcia-Morchon, R. Hummen, S. L. Keoh, S. S. Kumar,
cyber-physical systems,” Proc. IEEE, vol. 104, no. 5, pp. 1013–1024, and K. Wehrle, “Security challenges in the IP-based internet of things,”
May 2016. Wireless Pers. Commun., vol. 61, no. 3, pp. 527–542, 2011.
[81] A. Gupta, X. Lin, and R. Srikant,“Low-complexity distributed scheduling [103] J. Gubbi, R. Buyya, S. Marusic, and M. Palaniswami, “Internet of things
algorithms for wireless networks,” IEEE/ACM Trans. Netw., vol. 17, (IoT): A vision, architectural elements, and future directions,” Future
no. 6, pp. 1846–1859, Dec. 2009. Gener. Comp. Syst., vol. 29, no. 7, pp. 1645–1660, 2013.
[82] X. Lin and S. B. Rasool, “Constant-time distributed scheduling policies [104] P. H. Cole and D. C. Ranasinghe, Networked RFID Systems &
for ad hoc wireless networks,” IEEE Trans. Autom. Control, vol. 54, Lightweight Cryptography. Springer, Switzerland, 2007.
no. 2, pp. 231–242, Feb. 2009. [105] H. C. Pöhls et al., “Rerum: Building a reliable IoT upon privacy-and
[83] N. Vaidya, A. Dugar, S. Gupta, and P. Bahl, “Distributed fair scheduling security-enabled smart objects,” in Proc. IEEE Wireless Commun. Netw.
in a wireless LAN,” IEEE Trans. Mobile Comput., vol. 4, no. 6, pp. 616– Conf. Workshops. 2014, pp. 122–127.
629, Nov./Dec. 2005. [106] A.-R. Sadeghi, C. Wachsmann, and M. Waidner, “Security and privacy
[84] K. S. Vijayalayan, A. Harwood, and S. Karunasekera, “Distributed challenges in industrial internet of things,” in Proc. 52nd Annu. Des.
scheduling schemes for wireless mesh networks: A survey,” ACM Com- Autom. Conf., 2015, p. 54.
put. Surv., vol. 46, no. 1, p. 14, 2013. [107] A. W. Atamli and A. Martin, “Threat-based security analysis for the
[85] X. Wu, R. Srikant, and J. R. Perkins, “Scheduling efficiency of distributed internet of things,” in Proc. IEEE Int. Workshop Secure Internet Things,
greedy scheduling algorithms in wireless networks,” IEEE Trans. Mobile 2014, pp. 35–43.
Comput., vol. 6, no. 6, pp. 595–605, Jun. 2007. [108] Z.-K. Zhang, M. C. Y. Cho, C.-W. Wang, C.-W. Hsu, C.-K. Chen, and
[86] T. Zhang et al., “Distributed dynamic packet scheduling for handling S. Shieh, “IoT security: Ongoing challenges and research opportunities,”
disturbances in real-time wireless networks,” in Proc. IEEE Real-Time in Proc. IEEE 7th Int. Conf. Service-Oriented Comput. Appl., 2014,
Embed. Tech. App. Symp., 2017, pp. 261–272. pp. 230–234.
[87] T. Zhang, T. Gong, Z. Yun, S. Han, Q. Deng, and X. S. Hu, “FD-PaS: A [109] R. Arends, R. Austein, M. Larson, D. Massey, and S. Rose, “DNS security
fully distributed packet scheduling framework for handling disturbances introduction and requirements,” IETF, Fremont, CA, USA, Rep. RFC
in real-time wireless networks,” in Proc. IEEE Real-Time Embedded 4033, Tech. Rep., 2005.
Technol. App. Symp., 2018, pp. 1–12. [110] G. Baldini et al., “IoT governance, privacy and security issues,” Position
[88] D. Yang, Y. Xu, and M. Gidlund, “Coexistence of IEEE 802.15.4 based Paper, Eur. Res. Cluster Internet Things. 2015, http://www.internet-of-
networks: A survey,” in Proc. 36th Annu. Conf. IEEE Ind. Electron. Soc., things-research.eu/pdf/IERC_Position_Paper_IoT_Governance_Privacy_
2010, pp. 2107–2113. Security_Final.pdf.
[89] D. Yang, Y. Xu, and M. Gidlund, “Wireless coexistence between [111] S. Raza, “Lightweight security solutions for the internet of things,” Ph.D.
IEEE802.11- and IEEE802.15.4-based networks: A survey,” Int. J. Dis- dissertation, MDH, Västerås, Sch. Innovation, Design and Engineering,
trib. Sensor Netw., vol. 7, no. 1, p. 912152, 2011. Mälardalen University, Sweden, 2013.
[90] A. Sikora and V. F. Groza, “Coexistence of IEEE802.15.4 with other sys- [112] J. H. Ziegeldorf, O. G. Morchon, and K. Wehrle, “Privacy in the internet
tems in the 2.4 GHz-ISM-band,” in Proc. IEEE Instrum. Meas. Technol. of things: Threats and challenges,” Secur. Commun. Netw., vol. 7, no. 12,
Conf., 2005, pp. 1786–1791, vol. 3. pp. 2728–2742, 2014.
[91] L. L. Bello and E. Toscano, “Coexistence issues of multiple co-located
IEEE802.15.4/zigbee networks running on adjacent radio channels in
industrial environments,” IEEE Trans. Ind. Inform., vol. 5, no. 2, pp. 157–
167, May 2009.
[92] T. M. Chiwewe, C. F. Mbuya, and G. P. Hancke, “Using cognitive ra-
dio for interference-resistant industrial wireless sensor networks: An
overview,” IEEE Trans. Ind. Inform., vol. 11, no. 6, pp. 1466–1481, Dec. Authors’ photograph and biography not available at the time of
2015. publication.