ArticlePDF Available

Abstract and Figures

Functioning of the Internet is persistently transforming from the Internet of computers (IoC) to the ‘Internet of things (IoT)’. Furthermore, massively interconnected systems, also known as cyber-physical systems (CPSs), are emerging from the assimilation of many facets like infrastructure, embedded devices, smart objects, humans, and physical environments. What the authors are heading to is a huge ‘Internet of Everything in a Smart Cyber Physical Earth’. IoT and CPS conjugated with ‘data science’ may emerge as the next ‘smart revolution’. The concern that arises then is to handle the huge data generated with the much weaker existing computation power. The research in data science and artificial intelligence (AI) has been striving to give an answer to this problem. Thus, IoT with AI can become a huge breakthrough. This is not just about saving money, smart things, reducing human effort, or any trending hype. This is much more than that – easing human life. There are, however, some serious issues like the security concerns and ethical issues which will go on plaguing IoT. The big picture is not how fascinating IoT with AI seems, but how the common people perceive it – a boon, a burden, or a threat.
Content may be subject to copyright.
IET Research Journals
Artificial Intelligence in Internet of Things ISSN 1751-8644
doi: 0000000000
Ashish Ghosh1, Debasrita Chakraborty2, Anwesha Law3
1,2,3Machine Intelligence Unit, Indian Statistical Institute, 203 B.T. Road, Kolkata-700108, West Bengal, India
* E-mail:
Abstract: Functioning of the Internet is persistently transforming from the Internet of Computers (IoC) to the “Internet of Things
(IoT).” Furthermore, massively interconnected systems, also known as Cyber Physical Systems (CPS) are emerging from the
assimilation of many facets like infrastructure, embedded devices, smart objects, humans and physical environments. What we
are heading to is a huge “Internet of Everything in a Smart Cyber Physical Earth.” IoT and CPS conjugated with “data science” may
emerge as the next “smart revolution”. The concern that arises then is to handle the huge data generated with the much weaker
existing computation power. The research in data science and artificial intelligence (AI) has been striving to give an answer to this
problem. Thus, IoT with AI can become a huge breakthrough. This is not just about saving money, smart things, reducing human
effort or any trending hype. This is much more than that - easing human life. There are, however, some serious issues like the
security concerns and ethical issues which will go on plaguing IoT. The big picture is not how fascinating IoT with AI seems, but
how the common people perceive it - a boon, a burden or a threat.
Keywords: Artificial Intelligence, Internet of Things, Intelligent Systems
1 Introduction
We are quite fascinated by the word “smart”. But, what we have
today is still far from being smart like a human. Let us consider
the example of a smartphone, although it is “smart”, it cannot do
much automatically. For example, it is not able to put notifications
or message alerts in ‘silent mode’ automatically when the owner is
driving. It would be smarter if it could at least reduce distractions
caused by the alerts when the owner is driving. This requires some
kind of wireless connection between the person, his/her smartphone
and the car. In another situation, if the owner falls sick, the smart
phone should make an emergency call to a family member or a hos-
pital nearby. It will again need certain connections and information
(about the family members and hospitals) to facilitate this. If we go
on giving examples like this, we will see nearly everything present in
the physical world need to be connected to everything else to meet
some requirements or the other. To make these things “smart”, we
will need artificial intelligence.
AI is a technology that targets at making computers do human
like reasoning. This development will accelerate the digital trans-
formation of industries. Be it humans, animals, plants, machines,
appliances, soil, stones, lakes, buildings or anything one can think of,
connecting them together and making “smart decisions” can make
the world an autonomous place. To make the world and its physical
objects actually autonomous, we need a machine learning (ML) [1]
emulating human learning as well as a data analysis (DA) [2] mod-
ule in the system. ML would create techniques to facilitate learning
in various components/devices of the network to make them auto-
matic and self standing, whereas DA would evaluate/analyse all the
data that is generated over time to find out the past trends and be
more efficient/effective in future. This trend has been growing and
now efforts are being made to incorporate ML and DA into sensors
[3] and embedded systems [4] of the smart systems. The technology
behind AI is really intriguing and what it will turn into forces us to
rethink everything we know about the meaning and purpose of life
and work. The pace at which ML and DA are driving AI, calls for
a good need to discuss trends, challenges and threats that will grow
One of the greatest ideas behind this trend is the Internet of
Things (IoT) [5] which anticipates a world saturated with installed
intelligent gadgets, frequently called “smart objects” [6], inter-
connected through the Internet or other communication mediums
like Bluetooth, infrared, etc. These connections will be human-
human, human-physical things, and physical things-physical things.
Internet of Everything [7] is also a similar idea that suggests that
every living, non-living or virtual object is connected to each other
through some communication medium. When these concepts are
deployed to the physical world, what we get is a Cyber Physical
System (CPS) [8]. Such a world would be data wealthy, using which
knowledge could be extracted. Various disciplines like Database
Management System (DBMS) [9], Pattern Recognition (PR) [10],
Data Mining (DM) [2], Machine Learning (ML) [1], Big Data Ana-
lytics (BD) [11] will need improvised methods to deal with the
data, overlapping largely in their scope. This article mainly revolves
around intuitions, challenges and applications of artificial intelli-
gence in the concepts of Internet of Things, Cyber Physical Systems
and Internet of Everything.
2 Artificial Intelligence
Artificial Intelligence (AI) is the science of instilling intelligence in
machines so that they are capable of doing tasks that traditionally
required the human mind. AI based systems are evolving rapidly
in terms of application, adaptation, processing speed and capabil-
ities. Machines are increasingly becoming capable of taking on
less-routine tasks. While, humans intelligence is actually ‘taking’ a
perfect decision at the appropriate time, AI is merely about ‘choos-
ing’ a right decision at the appropriate time. To put it plainly, the
creativity in decision that humans can take is lacking in AI. It may
be argued that human ingenuity will always change the role of pro-
ductive work, but, AI based systems have quite elegantly reduced
repetition of human efforts and could give results in comparatively
low time. Most of the ongoing works in AI can be termed as ‘Narrow
AI’. This means that only certain tasks are enhanced by technology.
However, we are aiming for something much more than that. Hence,
many fields have conjugated to drive the AI development.
Various domains like philosophy, computer science, mathematics,
statistics, biology, physics, sociology, psychology and many more
have come up together to boost the interdisciplinary nature of AI.
Intelligence, comes from all the data generated in each of these
domains. Analysis of this data is important to bring out the prin-
ciples behind it. Human brain is capable of doing it easily, but it
takes a long time. This is because, the data in real world, has some
unwelcome properties:
IET Research Journals, pp. 1–11
The Institution of Engineering and Technology 2015 1
ReView by River Valley Technologies CAAI Transactions on Intelligence Technology
2018/09/27 14:56:24 IET Review Copy Only 2
This article has been accepted for publication in a future issue of this journal, but has not been fully edited.
Content may change prior to final publication in an issue of the journal. To cite the paper please use the doi provided on the Digital Library page.
Huge volume
Unstructured nature
Varied data sources
Needs real-time processing
Changes continuously
There are other properties too like volatility, virility, etc. AI can
be regarded as a technique to use the data in an efficient manner
so that it is understandable to the people who provide it, modifiable
(in case of errors), holds usefulness in the present scenario and is
AI, therefore, relies heavily on data science techniques. To state
in a broader way, data science is the science of developing tools
and methods to analyze large volumes of data and gain informa-
tion from it. The discipline is therefore, an amalgamation of many
other research areas. For developing tools, the ideas mainly come
from computer science which are primarily concerned with algo-
rithmic efficiency and storage scalability. For analysis, the ideas
come from much more varied sources. Methodologies are borrowed
from both the basic sciences (like physics, statistics, graph theory)
as well as the social sciences (like economics, sociology, political
science). Specific techniques which are naturally interdisciplinary
are also very popular in data science, such as pattern recognition
[10], machine learning [2], data mining [12], database management
systems [9] and big data analytics [13, 14].
One of the main tools to achieve artificial intelligence is machine
learning (ML). Human brain can solve certain types of learning prob-
lems. For example, there are plenty of optical neurons in the visual
system which make object recognition easy for humans. Learning is
not only restricted to humans, it is diversified to animals, plants, etc.
A bird learns to fly, a child learns to speak, plants learn to adapt to the
environment and so on. Our very survival depends on the ability to
learn and adjust to the environment. Machines can be equivalently
made to learn and modify itself for better performance imitating
the natural process of learning, to be termed as ‘machine learning’.
Learning (including machine learning) mainly takes place in three
ways: Supervised [2], Reinforcement and Unsupervised [15]. Other
methods [1] like semi-supervised learning [16], active learning [17],
inductive learning [18], deductive learning [1], transfer learning [19],
etc. also exist. Some are even inspired by the biological sciences
to mimic the evolution process of living beings [20]. The goal of
ML is not just instilling consciousness in a machine, but to design
algorithms that allow the machine to learn.
Learning can be defined as the act of acquiring or improving
behaviors, skills, values, preferences, thereby increasing the knowl-
edge. It may also include synthesizing various types of information.
Basically, learning is the mechanism by which a system modifies its
parameters such that its future performance can be improved. This
process of learning can be imitated by machines with the help of
‘machine learning’ [2]. Machine learning is an emerging field in
computer science research which gives inanimate systems an abil-
ity to learn [21] without actually having to program them explicitly.
In contrast to more traditional uses of computers, the IoT scenario
where the volume, variety, velocity and complexity of the data are
overwhelming, it is impossible for a human programmer to provide
an explicit, fine detailed specifications to execute the task. Thus, the
concept of machine learning is made to be concerned with implicit
learning skills, which would make a computer/system eventually
teach themselves to adapt to the current environment and make inde-
pendent decisions. This is how machine learning makes up for the
smart concept in CPS or IoT [22].
Machine learning is an approach to achieve artificial intelligence
[23] which is based around the concept that machines should be
given access to data so that they can learn for themselves. The way
that we will eventually create human-like AI has frequently been
talked about as a certainty by researchers. Surely, we are moving
towards that objective with expanding speed. A significant part of
the advancement that we have found in recent years is all because
of the fundamental changes in how we view AI working, which
have been brought about mainly by ML. Therefore, it would not
be inappropriate to give ML the credit of instilling smartness in
2.1 Smartness or Intelligence
‘Smartness’ or intelligence is at both microscopic and macroscopic
levels of IoT. These sentences may sound like a far-futuristic wave
of talking refrigerators and self-driving taxis, but it means much
more than that. Now, smart objects are mostly concerned about data,
devices, and connectivity. The data needs to be analysed to bring
out the hidden insights; this can be done with the help of Big Data
Analytics (BDA). Eventually, it is the analysis of this big data with
machine learning that makes the whole system smart.
Table 1 should makethe idea clear about the extent machine learn-
ing has spread into the idea of ‘smartness’. It shows few examples of
animals whose smartness have been replicated by several man-made
AI machines. Such machines are or will be capable of performing
certain functions like the corresponding animal or will have some
similar characteristics. Although complete replication of all the char-
acteristics of the living being has not been achieved, but research
is progressing gradually towards making these AI machines behave
more like its living counterpart.
Level Animal Machine Year
example example
Adaptive learning Earthworm Smart 2011
of new responses thermostat onwards
Learning by trial Fish CRONOS 2005
and error robot onwards
Learning by setting Octopus Cog 1999-2003
a goal, acting to
achieve it and
then assessing itself
Self consciousness Chimpanzee Siri 2011
and higher
order thoughts
Has emotions 1-6 year Cozmo 2016
like frustration old child onwards
and happiness
Has full theory 7-11 year Pepper 2014
of mind, old child onwards
interpret human
emotions, and
responds back
Passes Turing 12+year human MIT’s AI 2014
test program onwards
Table1 Smart animal to smart machine analogy
It is seen that certain characteristics and behaviour are yet to be
instilled into machines to make them somewhat “intelligent”. The
philosophy that drives machine learning is – to automate the ana-
lytical models and enable algorithms to continuously learn from the
available data. This data should be stored or tracked, in order to be
processed on time. There may be a lot of available data generated
each moment, but all of it may not be useful. The key idea, is to
collect relevant data and analyse it efficiently.
3 Internet of Things
Even a few decades back, nobody could have imagined having a
video chat with their families in a different continent. Nowadays, it
is a common thing. All of these is due to technology getting cheaper,
and devices emerging with new and improved capabilities. People
can get things done with a click on their smartphone, be it sending
emails, paying bills, transferring money or booking a cab.
IET Research Journals, pp. 1–11
The Institution of Engineering and Technology 2015
ReView by River Valley Technologies CAAI Transactions on Intelligence Technology
2018/09/27 14:56:24 IET Review Copy Only 3
This article has been accepted for publication in a future issue of this journal, but has not been fully edited.
Content may change prior to final publication in an issue of the journal. To cite the paper please use the doi provided on the Digital Library page.
Fig. 1: Different fields merging into IoT
What we had since 1991 was “Internet of Computers (IoC)” and
it gradually grew in size as more and more people started using it.
With the advent of pocket phones and connected devices, the Inter-
net of Devices started and eventually grew larger as mobile phones,
computers, laptops and tablets became cheaper and more accessi-
ble to the common man. Gartner, Inc. forecasted that 6.4 billion
connected things will be in use worldwide in 2016, up 30 percent
from 2015, and will reach 20.8 billion by 2020 [24]. In 2016, more
than 5.5 million new things got connected every day, thus, emerg-
ing the huge scope for Internet of Things. Since various things are
continuously connecting to form an IoT, there are various disciplines
that get associated with IoT. Therefore, IoT can also be thought of
as a combination of various domains. Figure 1 gives a representa-
tive list of some domains (most of these overlap with each other
in terms of concepts and techniques) constituting the IoT. Internet
of things is just a connected system of physical things (like appli-
ances, crop fields, plants, animals, etc.) and humans. Humans are
connected to these devices using some smart objects attached to both
which are capable of sending, receiving and analysing data. These
smart objects represent the entity (a human or a physical thing), it is
attached to, in the network.
3.1 Internet of Everything
Usually people get confused about the concepts of Internet of Things
and Internet of Everything. According to Cisco [7], “the Inter-
net of Everything is the intelligent connection of people, process,
data and things.” The IoE connects up the physical things to the
cyber things into one cohesive whole. It is not just about allowing
devices to talk to each other; it is about allowing everything (liv-
ing, non-living or any virtual object) to talk about each other. This
virtual object part is missing in IoT. IoT may have smart objects
(attached to physical things and humans) and an Internet infras-
tructure, but does not include a smart non-physical entity (kind of
a ‘cyber thing’ analogous to any physical thing). In IoE, connec-
tions can be human-human, physical thing-physical thing, cyber
thing-cyber thing, human-physical thing, physical thing-cyber thing,
human-cyber thing. The concepts IoT and IoE are very overlapping.
To get a better view of the concepts, we illustrate a venn diagram in
Figure 2.
We will describe Figure 2 in terms of sets. The following relations
Things Intelligence = Smart Objects (Devices)
Network Intelligence = Smart Network
Things Network = Networked Devices
Services Intelligence = Smart Services
Services Network = Internet Services
Things Intelligence Network = Internet of Things (IoT)
Internet Services Intelligence = Internet of Services (IoS)
Internet of Things Internet of Services = Internet of Everything
IoE has turned into a catchphrase to depict the integration of con-
nectivity and intelligence to pretty much everything (physical or
virtual) with a specific end goal to give them special functionali-
ties. For example, a smart website that may have some embedded
intelligence to identify when a person is getting annoyed by an
unnecessary advertisement or getting excited by an offer flashed on
the screen. Let us imagine a user-specific website; different users
see different layout/representation of the same website. In future, we
might also be able to develop web-based facilities so that even the
disabled could use the Internet for their benefits. Then only the true
purpose of the Internet would be served. Internet is for everyone and
IET Research Journals, pp. 1–11
The Institution of Engineering and Technology 2015 3
ReView by River Valley Technologies CAAI Transactions on Intelligence Technology
2018/09/27 14:56:24 IET Review Copy Only 4
This article has been accepted for publication in a future issue of this journal, but has not been fully edited.
Content may change prior to final publication in an issue of the journal. To cite the paper please use the doi provided on the Digital Library page.
Fig. 2: A Venn diagram for the concept of Internet of Things (IoT), Internet of Services (IoS) and Internet of Everything (IoE)
everything. Thus, comes the need to understand the key concepts that
build these IoE and IoT.
3.2 Things and Everything
When we are talking about IoT and IoE, we must be very clear about
the concept of “things” and “everything”. One straightforward con-
cept that may come to mind is anything that can be connected may be
the “thing” in IoT. However, we define it other way round. There can
be more features in making a physical object a “thing”. The “thing”
(living or non-living) should have:
1. a way to generate or collect data,
2. a way to process data,
3. a way to send or receive data,
4. a way to identify itself.
The main concept to consider, when thinking of IoT, is that
“Things” are physical objects, i.e., anything that has a real life pres-
ence. The Internet as we know it is not just made of physical devices.
For instance, a website cannot be thought to be a physical entity; it
exists somewhere virtually. This is true for services that we might
use every day, such as online shopping sites, social media sites, etc.
These “intelligent services” along with the “things” make the “every-
thing”. Thus, inter-connections as well as intra-connections between
“things” from physical world and “intelligent services” from the
cyber world make the IoE.
3.3 AI enabled IoT
IoT is a vast concept encompassing too many sensors, actuators,
data storage and data processing capabilities interconnected by the
Internet. Thus any IoT enabled device can sense its surroundings,
transmit, store and process the data gathered and act accordingly.
The last step of acting accordingly is entirely dependent on the pro-
cessing step. The true smartness of an IoT service is determined by
the level of processing or acting that it can perform. A non-smart IoT
system will have limited capability and will be unable to evolve with
the data. However, a smarter IoT system will have artificial intelli-
gence and may serve the actual goal of automation and adaptation. In
this context, few examples of existing IoT services with the working
of AI behind them are discussed here.
3.3.1 Voice assistants: These are cloud-based voice services
which act as table-top personal assistants for users. They per-
form various tasks through third-party applications and other smart
devices in their proximity. They are capable of answering queries,
calling cabs, making restaurant reservations, playing music, switch-
ing smart lights on/off, and many more tasks based on the user’s
voice commands. Few of the well known voice assistants are:
Alexa is the voice assistant from Amazon, which is used in prod-
ucts like Amazon Echo, Amazon Tap, etc. There is a specific set of
skills known as the Alexa Skills Kit (ASK) that can be modified and
updated to personalize or improve certain skills.
Siri from Apple Inc. is used in Apple Homepod which serves a
similar purpose.
Google Assistant used in Google Home has additional fea-
tures where it can recognize upto six different users and pull their
respective details to converse with them.
These voice assistants are capable of performing multiple tasks
mostly due to the application of various subfields of AI. Auto-
matic far-field voice recognition, wake word detection, speech to
text conversion, natural language processing and understanding,
contextual reasoning, dialogue management, question answering,
conversational AI, etc. are performed continuously to make the voice
assistants perform functions real time.
3.3.2 Robots: Recent advancements in this field of robotics has
led to the creation of robots who have increased human likeness and
are able to interact with humans while understanding, reciprocating
and expressing certain human emotions. Robots are IoTs in them-
selves since they contain multiple sensors and actuators along with
AI that helps them continuously learn and adapt themselves over
Pepper from SoftBank Robotics is a human-shaped robot which
is termed as a humanoid companion which can interact with humans.
It is able to understand a human’s emotion through his/her facial
expression, body movement, tone of voice, words used, etc. It
IET Research Journals, pp. 1–11
The Institution of Engineering and Technology 2015
ReView by River Valley Technologies CAAI Transactions on Intelligence Technology
2018/09/27 14:56:24 IET Review Copy Only 5
This article has been accepted for publication in a future issue of this journal, but has not been fully edited.
Content may change prior to final publication in an issue of the journal. To cite the paper please use the doi provided on the Digital Library page.
can identify four human emotions, namely joy, sadness, anger and
surprise, and reciprocates appropriately through movement, touch,
words and display on its screen. It is able to move around and interact
with humans and other devices in its vicinity.Pepper is commercially
used in various stores to interact with customers.
Sophia from Hanson Robotics is a social humanoid robot which is
incredibly human-like and can express emotions through more than
50 facial expressions. During a conversation, it is able to maintain
eye contact with the human while conversing. Sophia is world’s first
robot to get a country’s full citizenship. She has even given multiple
interviews and sang in a concert.
Robotic Kitchen from Moley Robotics is an advanced fully func-
tional robot integrated into a kitchen. It has robotic arms, oven, hob
and a touchscreen unit for human interaction and is able to prepare
food of expert quality from its recipe library.
Application of natural language processing, computer vision,
shape recognition, object recognition, detection & tracking, block-
chain technology to analyse inputs & responses, facial recognition,
voice recognition, speech-to-text technology, obstacle recognition,
haptics, etc. have been widely used in these robots to enable them
function effectively.
3.3.3 Smart Devices: In an IoT apart from the voice assis-
tants and robots, there are smart objects/devices that are present
which make the task simpler for humans. Smart objects that are AI
enabled use applications of object identification, facial recognition,
voice recognition, speech and expression identification, deep neural
networks, transfer learning, computer vision, etc.
Smart Oven by June aims to cook food perfectly everytime. It has
an HD camera and food thermometer which helps to automatically
monitor the food being cooked inside the oven, and can switch cook-
ing modes if necessary. This oven can be operated through Alexa and
can recommend and configure automatic cook program by analysing
the likings of the user.
SkyBell is an HD WiFi doorbell from Honeywell that allows the
user to answer the door through a smartphone or a voice assistant.
The video camera at the doorbell sends an alert and live feed to the
home owner’s phone to alert him/her about the person at the door.
The owner can converse with the person through SkyBell even from
a remote location. This has helped keep trespassers and burglars at
Smart Lights by Deako can be controlled remotely through
smartphones and Alexa or Google assistant. They are connected via
the Internet and can receive software upgrades from time to time.
Automotive AI by Affectiva is an in-cabin sensing AI that can
be used in robo-taxis and highly automated vehicles. It detects emo-
tional and cognitive state of the occupants in the vehicle from their
face and voice through in-cabin cameras and microphones.
3.3.4 Industrial IoT: Apart from being used inside smart
homes, IoT has a huge application area in the various industrial sec-
tors. These solutions perform statistical and financial analysis of a
company as a whole and finally provide predictions using some AI
and machine learning algorithms.
Primer is a product from Alluvium which provides industrial
solutions. Primer creates a real time Stability Score analysis based
on the data collected, the sensors in the system and assets. It aims to
detect potential issues well beforehand and helps operators identify
the anomalies and make necessary changes from something as small
as a sensor to the entire facility.
Plutoshift is another industrial IoT based solution. It enables
industrial companies to continuously track the performance of their
assets, measures financial impact and provides support for informed
decision making.
Thus, the opportunities and potential of both AI and IoT can
advance when they are combined. As IoT generates data, ML and
BDA carry the potential to find insights of huge value in the data.
Without the AI, the data that the IoT generates remains useless. IoT
needs to depend on AI as it is impossible for any human to find infor-
mation in the data that IoT generates. Moreover, if a new pattern in
data is detected, the machine will be capable enough to learn by itself
which will be impossible for a non-AI IoT system to do.
4 Cyber Physical Systems
The term “cyber-physical systems (CPS)” emerged around 2006,
when it was coined by Helen Gill at the National Science Foun-
dation in the United States [8]. CPS according to National Science
Foundation (NSF) are “engineered systems that are built from, and
depend upon, the seamless integration of computational algorithms
and physical components”. Today it is thought of as a system that
works on and is monitored by computer-based mechanisms (built
inside each component as well as in the complete system), strongly
connected via the Internet and is easily accessible to its users.
During World War II, Wiener established a technology in anti-
aircraft guns that can automatically aim and fire. In spite of the fact
that the components he utilized did not include advanced computers,
his control logic was effectively a computation, but one did with
analog circuits and mechanical parts. This thought has emerged as a
necessity driven concept. Few decades back, people only imagined
automatic cars; today, people are already creating automated cars
with advanced abilities to help reduce accident rates. To enhance this
plan further, future road networks may also be connected with the
cars via the Internet and communicate the information to help reduce
traffic congestion, accidents, etc. These may also be connected to
police station, hospitals, and so on.
In today’s context, CPSs are emerging from the integration
of infrastructure, smart objects, embedded computational devices,
humans and physical environments, which are normally tied by
a communication framework. These incorporate frameworks like
Smart Cities, Smart Grids, Smart Factories, Smart Buildings, Smart
Homes and Smart Cars where everything is associated with each
other protest. They are expected to give a resilient, flexible, profi-
cient and cost-effective scenario. Let us imagine that a road-accident
patient is rushed to a hospital only to be asked to make a police com-
plaint first or wait for the police to arrive. If somehow these systems
could be interconnected, the information about the accident would
be immediately sent to the police. All the necessary actions would
be taken instantly and the chances of delay in the treatment would be
reduced. However, such connections between objects should mani-
fest as a valid relationship in the physical world too. For example, a
traffic monitoring system should be connected to police stations and
hospitals, but in no way it should be linked to person’s home security
system. Connecting these two may lead to security issues, and also
overburden the data store as well as the network. So, the connections
made between devices and systems should be carefully planned,
keeping in mind all the pros and cons of each connection. To make
these connections and systems work perfectly and efficiently in har-
mony, an independent platform needs to be created which work for
the individual objects as well as the system on the whole.
4.1 CPS - a combination of disciplines
The CPS technology is prevalently from the industrial sector and
serves as the engine of innovation for a new era of end user products.
The infrastructure of CPS is, therefore, a combination of numerous
disciplines (most of which overlap).
1. Machine Learning: It is a platform to learn the trends of the
system from data generated in the past, to make an informed decision
in future, without manual monitoring.
2. Big Data Analytics (Data Science): All the data that is generated
in the huge interconnected system will constitute a massive content
that will be processed and analysed over time to make the system bet-
ter. Usually, machine learning algorithms are modified and adapted
to handle the ‘Big Data’ scenario.
3. Design: The overall system needs a robust, tolerant and efficient
design, which connects all the components as required.
IET Research Journals, pp. 1–11
The Institution of Engineering and Technology 2015 5
ReView by River Valley Technologies CAAI Transactions on Intelligence Technology
2018/09/27 14:56:24 IET Review Copy Only 6
This article has been accepted for publication in a future issue of this journal, but has not been fully edited.
Content may change prior to final publication in an issue of the journal. To cite the paper please use the doi provided on the Digital Library page.
4. Process Science: Different industrial manufacturing processes
are demanding the use of automation in their production lines.
5. Wireless Sensor Networks (Communication): The whole sys-
tem would depend on communication; wireless connections between
each component of the system would help pass information from one
object/ system to another.
6. Software: All the working devices and system need a software to
work. These softwares would be system and task specific.
7. Embedded Systems: The gadgets/devices that constitute a CPS
would contain embedded systems, like camera, temperature mea-
sure, etc. Each device would have different embedded systems or
sensors as per their requirements.
8. Cybernetics: It is relevant to mechanical, physical, biological,
cognitive and social systems. To enable any device attached to an
entity for storing, processing, sending/receiving data, this field is
highly in demand.
9. Mechatronics and Robotics: These are fields which seek
human-like actions for different tasks. These will not be manu-
ally handled or given some fixed instructions, rather they will be
intelligent enough to know what needs to be done at the right time.
10. High Performance/Cloud Computing: Typically, the issues
considered cannot be solved on a single commodity computer within
a sensible amount of time (excessively complex operations are
required) or the execution is incomprehensible, because of restricted
accessible assets (a lot of training data is required). High Per-
formance/Cloud computing is the way to beat these impediments
by utilising specialized or high-end hardware or by accumulating
computational power from several units. The corresponding data dis-
tribution and operations across several units simultaneously requires
the concept of parallelism.
11. Cognitive Science: Cognitive science predominantly involves
concepts from psychology, philosophy, neuroscience, anthropology,
computer science and linguistics. It is the study of mind and the level
of intelligence. The goal is to understand the nature of knowledge in
various living beings and how that knowledge is acquired, processed
and used.
This is not an exhaustive list of the various domains that may merge
up to make CPS. Due to the interdisciplinary nature of the research,
most of the concepts overlap. Moreover, other domains may also
collaborate in future to improve the CPS scenario in some form.
5 Components of IoT-CPS
Now that we have established a clear relationship between IoT, CPS
and the related terms around it, the ecosystem of these technolo-
gies matter most. Since CPS is a combination of subsystems, we can
concentrate on the structure and components of IoT initially. If we
dismantle the various parts of IoT, we would be left with something
as shown in Figure 3.
From Figure 3, we can see that there are various components
in an IoT system. Apart from network infrastructure and security,
a major portion of IoT requires data storage and processing on a
macroscopic (i.e., in the overall system) as well as on a microscopic
level (i.e., in each smart object locally). The smart objects them-
selves should have some data processing, intelligence and decision
making capability in them. For this, they need to have built-in data
processing tools to analyse the sensor data and make some smart
decision. Machine learning and data analytics [12] are the best can-
didates for such smart data analysis. On a macroscopic level too,
more than billions of things will generate data independently and it
would be transmitted over the network to some remote data storage
locations for further data analysis and all this in real-time. It sounds
like a big data task. A lot of data will be generated, stored, and pro-
cessed continuously. So, big data analytic (BDA) and ML together
will weave out the intelligence in IoT.
Any smart object can also have a limited data storage capability
and limited data processing capability as well. For example, a smart
watch signals to walk whenever the user is detected stationary (sit-
ting or lying down) for a long period of time. However, it does not
alert when the user is sleeping. It can clearly differentiate when the
user is sleeping and sitting. To do so, it does not need to transmit
the data to any server and perform a remote processing. It collects
data and does some small analysis inside itself to actuate the alarm.
These short term decision making capabilities are embedded in a
smart device. For long term decision making or for finding insights,
remote storage and processing may be needed.
IoT will mean too many connected devices. When such “every-
thing to everything” connection is established, physical world will
be full of sensors/actuators and the virtual world will be full of data.
The network will be highly complicated, and data would be gener-
ated throughout the CPS all the time. Different analysing systems
will be handling different parts of the IoT-CPS. All of the data is not
needed to be handled at one place or at one time. So, smaller relevant
portions of the data are extracted and dealt with as and when needed.
The data should be analysed reasonably in real-time in order to make
useful decisions. The data that is generated and handled by IoT is
essentially done by the individual parts of IoT which together form
a whole system. These parts of IoT are discussed in the following
5.1 Smart Objects
To catch up with such a substantial concept, we will require at least
millions (or more) of data generating smart objects. These will act
like the building blocks of such a major system. We have two ele-
ments to consider in the physical world; a physical entity (PE) and
smart object (SO).
A PE can be anything from people, creatures, plants which may
not be able to directly interface with the IoT but is an integral part
of the system. Such physical entities will have smart objects (SOs)
attached to them. These SOs are the AI elements that have the ability
to communicate via network. They can be anything like implanted
chips, wearables or smartphone that is somehow attached to the PE.
So, an SO becomes that device which helps a PE connect to the
‘Internet’ of things. The Internet however is virtual and both the PE
and SO are physical objects. Hence, they must need a digital rep-
resentation. The digital representation of the PE by the SO is the
digital entity (DE). For example, if we are the PE, our smartphone
becomes the SO and our social media app becomes the DE. Some-
times, the PE itself maybe integrated to an SO, like an automated
car. SOs are physical world portrayal of DEs in digital world, hav-
ing the capability to sense, store, process (locally) and communicate
via networking. SOs may act as intelligent agents with some level
of self rule, cooperate with other entities, and exchange information
with human clients and other computing devices within the inter-
connected Cyber Physical Systems. DEs are virtual programming
elements which have autonomous objectives. They can be either
services or simple coherent data entries.
A Physical Entity (PE) or thing can be represented in the cyber
world by a DE by its Digital Proxy (DP). DPs can be viewed as
users in cyber world, just like our social media profiles (our DP)
are viewed as being us (where we are the PE). Every PE has a DP,
which is used to represent it in the digital world. There are numer-
ous sorts of digital portraits (which are known as DE) of PE that
we can imagine: avatars, 3D models, objects (or instances of a class
in an object-oriented programming language) and even a social net-
work account could be viewed as such. However, in the IoT context,
Digital Proxies have two fundamental properties:
1. Each Digital Proxy must have just a single ID that distinguishes
it from others. The association between the Digital Proxy and the
Physical Entity must be established automatically.
2. Relevant digital parameters pertaining to the characteristics of the
Physical Entity can be refreshed upon any difference in the former.
Similarly, changes that influence the Digital Proxy ‘might’ be shown
on the Physical Entity in the physical world through actuators.
Data generated by these smart objects need to be transmitted
through wireless technologies, and the objects themselves should
be clearly identifiable. All transmitted data can be collected in a
distributed database and then monitored, analysed, processed. The
IET Research Journals, pp. 1–11
The Institution of Engineering and Technology 2015
ReView by River Valley Technologies CAAI Transactions on Intelligence Technology
2018/09/27 14:56:24 IET Review Copy Only 7
This article has been accepted for publication in a future issue of this journal, but has not been fully edited.
Content may change prior to final publication in an issue of the journal. To cite the paper please use the doi provided on the Digital Library page.
Fig. 3: IoT architecture tree
Fig. 4: An example of real world to virtual world mapping
Internet of Things will push the advancement in digital storage too.
The collection, transmission and processing this massive data to
mine out valuable insights in real-time bring us back to the topic
of data analytics and machine learning again.
5.2 Data Storage and Data Processing
As we understand, the main motive of IoT and CPS is to create
an autonomous system that can handle different situations across
the globe, which would eventually assist humans to lead a bet-
ter life. The basic IoT-CPS framework consists of smart objects
(which seem like the nodes in a graph) and the connections between
them. Assuming all the nodes and connections are made, the data is
being generated and communicated from one node to another every
moment. But the SOs do not know what to do with it. Neither can
they store it, nor do they know how to process it; this would make the
whole system useless. The objective of being autonomous, making
decisions and taking actions would never be fulfilled without having
proper data storage and processing units. This is an essential fea-
ture, needed both locally in smart objects, and also globally in the
complete system. The SOs will handle small sets of data continu-
ously flowing into the system; this can be stored temporarily in the
SO till a task is done, and then it can be moved to the global data
store. The data store of the whole system may not receive streaming
data, but will mostly get large chunks of collected data from time to
time. To handle both these types of data in real time, and utilize them
effectively, the role of big data analytics is crucial.
All these data are to be stored, but what exactly needs to be done
in the processing phase is unknown. We expect a smart system, IoT-
CPS, to work autonomously, i.e., observe its surroundings (through
different parameters), learn from experience, understand the need of
the hour, and make a useful decision/action. For an object/system
to imitate humans, it needs the ability to learn from the data. Since
human intervention may not be available or desired most of the time,
the system needs to learn independently. All these can be effectively
done with the help of artificial intelligence.
5.3 Communication Networks
Continuous analysis of big data over these platforms demand an
efficient and reliable network structure. Virtualisation of nearly
every physical thing imposes big challenges on the network service
providers. There should be advanced wireless technologies that can
handle such enormous eruption of devices. Smart devices need a
smart network infrastructure. Connecting machines and devices to
telecommunication networks is not new. What makes IoT an inno-
vation is the incorporation of smartness in the devices as well as
the network. This will ensure networks which will automatically
detect the necessity of connections between two objects and thereby
increasing or decreasing the connection strength. Moreover, smart
IET Research Journals, pp. 1–11
The Institution of Engineering and Technology 2015 7
ReView by River Valley Technologies CAAI Transactions on Intelligence Technology
2018/09/27 14:56:24 IET Review Copy Only 8
This article has been accepted for publication in a future issue of this journal, but has not been fully edited.
Content may change prior to final publication in an issue of the journal. To cite the paper please use the doi provided on the Digital Library page.
networks may also be smartly secure. They will easily identify the
intrusion or theft situations and take necessary steps for that. There
are numerous such capabilities that are still waiting to be harnessed.
Some people believe that for long distance operations, 5G net-
works can meet all the requirements of IoT devices. These 5G
networks will be faster and smarter, but, sometimes it is hard to
imagine why we would need such high bandwidth of data. For exam-
ple, a smart water-heater with 5G connection will seem unnecessary.
That is because, much of the technology that will connect to the
network during this IoT surge are not yet invented. For example,
our personal assistants might not be inside our phones. They can be
holographic projections roaming with us and also connected to the
Internet. 5G is projected to be deployed by 2020. But, it is yet to be
witnessed whether a 5G network can also handle the IoT or we need
some more Gs.
Connection through the Internet may not always be necessary. For
short range communication services, we might use Bluetooth tech-
nology. The new Bluetooth Low-Energy (BLE) has been designed to
operate on very low power consumption. The devices may be con-
nected to the smartphone via BLE and may be used to send/recieve
small chunks of data only. For smaller devices that are not being
moved around too far and which may not need to be connected
throughout the day, Bluetooth connection seems to be a good choice.
Although, BLE may not be suitable for transferring bigger files or to
hold a good connection at a large distance. In those cases, WiFi con-
nectivity is often an obvious choice for people who want wireless
connectivity in a local area. Typically it has a data transfer rate of
150-200 Mbps and can be maximum upto 600 Mbps. It can therefore
be beneficial for file transfers, but is too power consuming under the
scenario of IoT.
There are several other communication protocols that may
become the skeleton for IoT. Zigbee [25], Z-Wave [26], 6Low-
PAN [27], Near Field Communication (NFC) [28], Sigfox [29],
LoRaWAN [30], etc. are used for various use cases.
5.4 Security
All these things may sound fascinating, but as Sarah Jeong has
pointed out in her book [31], IoT means “Internet of Garbage”. It
says that “if the Internet was a city, its streets would be piled so
high with garbage that driving to the grocery store would be almost
impossible”. Internet contains garbage like harassment and intimi-
dation, crimes, copyright abuse, malwares, spams, etc. But we can
develop better interactions and better discourse, through insightful
architecture, strict moderation and efficient community manage-
ment. We just need to filter the useful content from the garbage first
and then attempt to find value in it.
As IoT will be rapidly adopted across the globe, it will generate
new demands. The biggest concern, after putting together everything
in IoT like the smart objects, big data analytics and communication
capabilities, is to ensure the security in such a large scale scenario.
Securing the IoT devices means much more than just securing the
devices themselves. The software applications and network connec-
tions that link to those devices should also be secure. Users of smart
objects and IoT will be highly vulnerable since their data is available
on a network. There are three key issues of IoT devices and services -
data confidentiality, privacy and trust. In IoT, the user along with the
authorised smart objects may access the data. The IoT device needs
to be able to verify that the entity (person or other device) is autho-
rized to access the service. Therefore, authentication and identity
management is needed.
The act of protecting the interconnected systems and their com-
ponents has come to be known as ‘cyber security’. Cyber security is
of utmost importance while dealing with smart devices, IoT and CPS
to avoid hackers from accessing users data. Cyber Security aims to
1. protect both IoT devices and services of unauthorized access
from within the devices and externally.
2. protect the services, hardware resources, information and data,
both in transition and storage.
There are various technologies for cyber security like crypto-
graphic systems, firewall, intrusion detection systems, anti-malware
softwares and scanners, secure socket layers.
Moreover, there are always some ethical issues. Suppose a small
wearable gadget records the health and fitness information of a user.
This information is available to the gadget service providers, since
the gadget is connected to their global database. Now, the service
providers may sell this user data to other companies without the
user’s consent. Depending on the fitness tracker information of the
user, he/she may start receiving offers or advertisement via mes-
sages/emails about some new fitness gear. In this case, the IoT is
anticipating what might interest the user might buy. Some users
may not want their personal information sold in this way, while
others do not mind promotional offers. In another case, the user’s
personal information may be used against him/her leading to some
unwanted situation. Most of the time, selling of these user data with-
out user consent are not beneficial to the user. Data sharing should
be an option for the user to choose. Selling or distribution of user’s
personal data should be done only with user’s consent.
6 Artificial Intelligence and IoT-CPS
The first industrial revolution during 1760-1840 gave rise to a rapid
growth of machines. With the advent of the second industrial rev-
olutions (1870-1914) people became richer and urban. Currently, a
“smart” or “cyber” revolution is under way. A number of interdis-
ciplinary technologies and sciences are converging and giving rise
to smarter softwares, new materials, dexterous robots, ground break-
ing inventions (like 3D printers) and a whole range of personalised
web services. As compared to the previous two industrial revolu-
tion phases, this smart revolution is evolving at an exponential pace.
Growing interest in the study and development of artificial intelli-
gence (AI) [23] are pushing the product vendors to introduce AI
into almost every strategy they make. Almost every organisation has
plenty of data at hand, and therefore needs AI to use it efficiently for
their own benefit.
6.1 AI enabled IoT-CPS
Talking of data, we have plenty of that in the IoT-CPS scenario.
Data - big or small - is an invariably integral part of the IoT world
of connected devices. The smart objects should themselves be able
to do a small scale local processing and have some inherent intelli-
gence. However, for a data dependant decision, more data should
be utilised. Storing these data for analytics inside a smart object
may not always be feasible. Here, the macroscopic version comes
to play; the data are sent to remote locations in a distributed fash-
ion and are analysed. The analysis results are then integrated and
finally the decision, in some necessary cases, may be sent back to
the smart object where the actuator can then perform its task. The
time between sending the data and actuating the decision should
be practically less or else it would not be meaningful. Traditional
analytic tools are not capable of capturing the entire essence of this
massive data in real-time. On one hand, the volume, velocity and
variety is too large for comprehensive analysis; whereas on the other
hand, the range of possible relationships and correlations between
different data sources are very vast for any analyst to manually com-
prehend. A good machine learning system, in order to deal such big
data, requires
1. data preparation capabilities,
2. learning algorithms - basic and advanced,
3. automation and adaptive processes,
4. scalability,
5. ensemble modelling and
6. real-time decision making.
This means the system should be able to make most of the deci-
sions and take required actions quickly. ML already has some good
capability [1] of letting computers do some thinking for us. But, we
are striving for more when we are trying to deal with big data. That
IET Research Journals, pp. 1–11
The Institution of Engineering and Technology 2015
ReView by River Valley Technologies CAAI Transactions on Intelligence Technology
2018/09/27 14:56:24 IET Review Copy Only 9
This article has been accepted for publication in a future issue of this journal, but has not been fully edited.
Content may change prior to final publication in an issue of the journal. To cite the paper please use the doi provided on the Digital Library page.
is why we need to adapt the ML methods to handle big data and also
build some new ideas.
The emergence of CPS and IoT are inspired totally by the idea
of social, economic and human benefit. Therefore, CPS and IoT
can be thought to be almost anything like personalised healthcare,
smart grids, smart industries, smart transportation, etc. For exam-
ple, a smart industry can improve its manufacturing processes by
sharing real-time information among the various industrial equip-
ments, supply chains, distributors, business systems, and customers.
A healthcare CPS like a smart hospital may monitor the physi-
cal conditions of patients remotely to serve the far to reach areas.
Whenever a road accident occurs, the nearest hospital, police station
and family member may be notified. An ambulance is immediately
sent to the accident location, the on-duty doctor is alerted, and the
police arrives at the spot without wasting any time doing things
manually. Similar emergency situations should benefit the most from
these interconnected autonomous systems. This ‘smartness’ is what
artificial intelligence will bring to such IoT-CPS infrastructure.
The IoT-CPS applications involve components that interact
through a complex physical environment. Such a connected envi-
ronment is therefore a challenging innovation that can possibly
change existing ventures. For example, manufacturing industries,
energy systems, healthcare, transportation facilities, buildings, criti-
cal infrastructure, emergency response systems, defense, agriculture,
etc. will undergo an upgradation to their smarter and interconnected
versions. Such organisations should have system aware assets which
would automatically asses the forthcoming faults or failures in the
system. By being system aware, we mean that a device implanted
in any part of a machine should be able to sense itself along with its
environment. The advances in AI applied to such connected IoT-CPS
scenario will help us understand the vision of not only a smarter but
a “brilliant planet”.
6.2 Cognitive AI and IoT-CPS
IoT not only means a combination of wireless sensor networks, data
storage, embedded systems or security issues, it means much more
than that. It is a vision of a world connected by intelligence. This
seems to be science fiction, but this is what makes IoT the buzz word
today. The conventional way to deal with programmable computing
is to filter the information through a progression of a fixed set of rules
and then arrive at an outcome. However, it will not be that efficient
to fulfill all that IoT promises to deliver. This is because inflexi-
bility restrains their convenience in tending to numerous parts of a
mind boggling, quick paced world, where the information processing
capability declines exponentially and it goes unused. Psychological
figuring has no such constraints. Instead of having a set of rules,
psychological frameworks learn from the hidden relations within the
connections with people, things, environment and their encounters
with each other. So, instead of being deterministic, they are proba-
bilistic. This empowers them to keep pace with the volume, variety,
variability and unusualness of data produced by the IoT.
These psychological frameworks are called cognitive computa-
tion models in formal terms. These frameworks form an essential
part of the artificial intelligence in IoT-CPS. They are capable
of comprehending the 80 percent of the world’s information that
researchers call “unstructured”. Examples of such unstructured data
include recordings, sound, even online journals, images, mails and
tweets. This implies that almost all the organizations are presently
ready to light up parts of the IoT that were already imperceptible. At
the point when such cognitive understanding is connected to the IoT,
the outcome is what we call Cognitive IoT, which we characterize
as frameworks that mix knowledge into, and gain from, the physical
“Cognition” is the process of acquiring knowledge and under-
standing through thoughts, experience, and senses. So, intuitively,
cognitive IoT can be thought of as an extension of IoT which is capa-
ble of understanding, reasoning and learning. These three aspects of
cognition vary in meaning while comparing human cognition and
cognitive IoT. For IoT, to ‘understand’ would mean to be able to col-
lect large volumes of data from the network and find the underlying
meaning of the data. It would be able to create concepts, identify
various entities and define relationships between them. To have the
ability to ‘reason’ for an IoT means it should be capable of finding
appropriate answers to queries or solve relevant problems without
being explicitly programmed. Lastly, a cognitive IoT should be able
to ‘learn’, i.e., independently infer new information from the data at
hand using the past knowledge it has obtained.
6.3 Example Cases of AI enabled IoT-CPS
Although, machines are not made to completely replace humans,
they are just there to help humans reduce the task load. Obviously,
humans need to maintain supremacy over machines. AI is most
effective when it is conjoined with human intelligence, rather than
replacing it. It highlights the idea that computers and humans have
different strengths in the vast field of excellence: computers are
much more efficient at doing arithmetic jobs and counting, while
humans show a remarkable performance in logic and reasoning.
These differing forms of intelligence are complimentary, not diamet-
rically opposites. Thus, AI is the technology that can fulfil our dream
to have ‘things’ that can ‘think’ [32]. Some examples of cases where
artificial intelligence have been incorporated and used in IoT-CPS
scenario are as follows.
Energy Utilization: Algorithms have been developed on a small
scale, to reduce energy consumption in a coffee machine (ARIIMA).
It can be modified and implemented in other scenarios to reduce
energy consumption; for example, in temperature control systems of
houses, which can make them efficient and reduce wastage. Different
houses will have different temperature settings adjusted according to
the residents, and the system will learn that efficiently.
Routing/Traffic: Traffic management or routing is a field of appli-
cation in ML. Depending on different parameters like traffic, road
condition, weather, etc. best routes are suggested.
Cost Savings: Predictive abilities are amazingly helpful in an indus-
trial setting. By drawing information from different sensors in or on
machines, machine learning calculations can learn the usual running
conditions of the machines. So, when some irregularities occur, it
can identify the machine and raise an alarm. This would save cost
as well as avoid accidents. An organization called Augury does pre-
cisely this with vibration and ultrasonic sensors installed on their
equipments [33] and saves money by predicting any malfunction
before it happens.
In plain words we want an ‘Internet of Things’ where both the
‘Internet’ and the ‘Things’ have the power to think [32]. That incul-
cation of thought is where the ‘intelligence’ flavour of IoT lies.
This may seem too over-rated, but this is what research on current
artificial intelligence all about.
7 Challenges
After all the idea has been developed, the gap between an idea and
a working prototype is huge. Even if the working prototype is set,
one may need resources to develop the prototype. Even if one is
able to strive past all of this, the question that stands now is how
do we know if this hot new technology will succeed or fail. Most
of us, even experts in that particular domain, get it wrong most of
the time. Recent IoT trends indicate that data is coming at a faster
speed, from various sources in various forms. It obviously exceeds
the abilities of an information system to imbibe, store, analyze, and
process it. It is not hard to find databases with some petabytes of
data, but the main objective is to not let all of these data go to waste.
Therefore, efforts are being made to recognize and extract meaning-
ful information (patterns, structure, underlying relationships, etc.)
from them. This task is quite complicated and it needs advanced stor-
age and processing techniques due to the unimaginable volumes of
data. In this scenario, new algorithms are being devised and well-
known techniques are being revisited and tailored in the fields of
IET Research Journals, pp. 1–11
The Institution of Engineering and Technology 2015 9
ReView by River Valley Technologies CAAI Transactions on Intelligence Technology
2018/09/27 14:56:24 IET Review Copy Only 10
This article has been accepted for publication in a future issue of this journal, but has not been fully edited.
Content may change prior to final publication in an issue of the journal. To cite the paper please use the doi provided on the Digital Library page.
Pattern Recognition, Machine Learning, Data Mining, etc. to handle
the new challenges posed by these data. There are other concerns
too regarding such a union of technologies. The field of AI needs
more development with the advent of such IoT-CPS infrastructure.
As example cases, we can discuss some issues. Complex adaptive
AI systems may lead to self sustaining malicious evolution of sys-
tems that can mimic a cancerous growth in human body. We would
therefore need more research to combat such systems using superior
evolving AI systems. Cyber-security will also be a major concern as
in this age of technology, there will be cyber wars. Any autonomous
system will be used for malicious reasons if hacked. Such vulnerabil-
ities of AI systems should be checked so that they stay safeguarded
against such attacks. Mock attacking AI systems should be devel-
oped that would immunize the existing safeguarding AI system.
There should also be systems that would predict the new type of
attacks that can arise. Any organization must develop systems to
quickly recover from cyber events that disrupt usual business oper-
ations. Critical information that may be maliciously used against an
individual should be automatically identified and should be kept pri-
vate even if people share it publicly. AI systems can be used to ensure
such privacy as well.
7.1 Challenges of CPS
CPS can be thought of as a reality to virtual mapping where the phys-
ical world is connected to the virtual world of information processing
via some actuators and sensors. Thus, there is a need for Internet of
Things (IoT) infrastructure inside every CPS. Now, we also need the
IoT infrastructure to connect various CPSs with one another. The
sensors inside a CPS will continuously spew real-time data. Many
CPSs around the world would therefore be a massive generator of
big data and will demand real-time processing. In this scenario, a
number of specific challenges are raised due to the large-scale nature
of CPS.
The physical world is governed by the elementary laws of physics.
The changes made here are static. Therefore, the interaction between
objects in the real physical world are well known. Whereas, the cyber
world is governed mainly by basic laws of data. The data is either
volatile or continuously changing. Any physical system may be pre-
dicted partially by simulation, but for a cyber process, it is difficult
to predict the behavior. When the two worlds come together in a
CPS, they need to operate in synchronicity. It poses a big challenge
for global system control. Moreover, the interconnectedness and data
sharing features demand for well defined platforms and control inter-
faces. There eventually builds up a need of an even tighter cyber
security in such a large-scale platform.
There are even bigger challenges with the ‘big data’ that the CPS
creates. The data should be efficiently stored, cleaned, processed and
analysed in real time. After all these steps, it is essential to explore
the heterogenous nature of CPS models. This can be done by the
compositional verification and testing methods. There is an urgent
need of standardised architectures and abstractions that enables effi-
cient design and development of cyber physical systems. Even after
all these challenges, people are striving to fight all odds and make
this work for the promising future that it holds.
7.2 Challenges of IoT
IoT may seem fascinating to talk about. We argue that most of the
technical challenges have been addressed like having the ultra-low
power microcontrollers, advanced sensors, wireless technologies
etc., but still we are speculating about the IoT rocket to take off.
There are still some major issues which have held the world to be
encompassed by this IoT yet.
7.2.1 Connectivity: Though the world has advanced its scope
and use of the Internet, it is still not available in many small vil-
lages and remote locations around the world. Though companies like
Google has invested in ideas like ‘balloon poweredInternet’, it is still
a challenge to bring the whole earth within the coverage of the Inter-
net. The whole idea of IoT is built on the assumption that there will
be constantly reliable and fast network connectivity. This is one of
the biggest roadblocks IoT is facing presently.
7.2.2 Security and Trust: Trust and security in the world of
connected devices, are the two key problems to mass adoption of
IoT. Users are quite concerned about IoT being a safe option to share
their data. Once every information and device is interconnected and
is available on a network, it can be accessed by hackers and can
be used for various fraud. For example, an IoT connected home
might increase the security risk of a burglar intrusion. Or a com-
pany’s privacy may get breached when the competitors get access to
its production data. Even if all the security measures are taken, there
will still be trustability issue.
7.2.3 Interoperability: It is hard to make meaningful connec-
tions between many random devices. IoT requires standards to
enable platforms that are connected, communicable, operable from
distant locations, programmable across devices and should be inde-
pendent of the model, manufacturer or industry it is coming from.
In other words, the IoT should be platform independent and should
work even if the devices have different OSs (operating systems), dif-
ferent OEMs (original equipment manufacturers), different types of
connectors, different versions and different protocol standards.
7.2.4 Scale: IoT means millions of connected devices. This will
call for two major trends: data integration and ‘big little data’. The
former means that all types of data will be generated from this sys-
tem, and they need to be combined as and when analysis is needed.
The latter however indicates the tremendous multiplication of small
datasets across the web. The blending of small datasets with small
datasets, small datasets with big datasets and big datasets with big
datasets will require different approaches. Moreover, as new inno-
vations will be made, newer types of devices will emerge. IoT will
have to keep up the pace with these too. The scale of devices will also
grow and will impose a heavy burden on the connectivity aspect. The
initial system should be able to handle changes in the structure of
the system. It should have scope for change in future without much
modification in the original structure.
7.2.5 Energy and Environment: Most of the devices that we
currently use operate on battery power and have a very limited shelf
life. As IoT gains popularity, the number of devices and the size
of the network will grow quite fast. Based on the current energy
availability, it would be impossible to power these billions of tiny
devices along with the full fledged network across the globe. We
would eventually have to shift to more unconventional sources of
energy for prolonged use.
If the future devices also have a very short shelf life like the
devices now, that would generate a huge amount of e-waste. It would
then be impossible for the environment to keep balance in nature
and be a hazard. Thus, for future development of IoT, the individual
devices as well as the total system should be eco-friendly in nature.
Development in IoT will demand research in the field of unconven-
tional or renewable energy resources, research to make devices that
will have a long shelf life and in the field of reusability and recycling
of resources.
7.3 Challenges of Data Analysis
With the advent of IoT, practically everything will have a virtual
existence in the cyber world. All of these entities will generate
data like nobody has imagined before. Analyzing these massive and
fast flowing data requires advanced technologies (like virtualiza-
tion softwares), adaptable cloud computing, etc. It also needs very
powerful high-performance computing devices that can provide the
mechanism to discover the underlying insights in big data.
The IoT connected devices will produce data at such a staggering
rate that in future the data volume from this first era of big data will
seem to be a dwarf. Machine learning can likewise help machines,
humans, devices, etc. collectively called ‘things’, get together to
comprehend what important information individuals may need from
the data. Additionally machine learning assumes a basic part in IoT
aspect to handle the immense volume of data produced by those
IET Research Journals, pp. 1–11
10 c
The Institution of Engineering and Technology 2015
ReView by River Valley Technologies CAAI Transactions on Intelligence Technology
2018/09/27 14:56:24 IET Review Copy Only 11
This article has been accepted for publication in a future issue of this journal, but has not been fully edited.
Content may change prior to final publication in an issue of the journal. To cite the paper please use the doi provided on the Digital Library page.
‘things’. It gives IoT and those ‘things’ a cerebrum to think, which
is called “embedded intelligence” by a few researchers [34].
Indeed, something to remember is that if someone is determined
to search for a pattern in a given data, he/she will eventually discover
one. This inclination to discover patterns might be the reason for
human knowledge i.e., having the capacity to identify conveniently
meaningful patterns out of things we have watched. However, maybe
the key issue with this sort of pattern discovering, both with humans
and artificially intelligent computers, is judging whether the patterns
we unavoidably find are at all significant, genuine and valuable.
In conjunction with proper AI structures and upgraded machine
learning models, this may in the long run prompt an exact imperson-
ation of the human brain on a significantly greater scale. Under the
attire of IoT, we may begin making complex systems that are suffi-
ciently canny to begin understanding things as mystifying as human
irrationality, crimes and even human’s dependence on machines.
Once these characteristics get assimilated in them, that renders them
smart, there is no reason why people would not be threatened and
inevitably subjugated by such machines.
8 Conclusion
In future, people will be wearing intelligent gadgets, eating intel-
ligent capsules that judge the impact of the medicine on the body,
living inside intelligent homes, and so on. This sounds like a science
fiction, but this is what all the present research is about. Everything
will be smart and will be connected to the Internet. All branches of
science will collaborate to create something of a big value. We will
have a ‘smart cyber revolution’. However, there is still a debate on
whether we are heading towards a creative destruction or not.
For instance, machines are now able to take on less-routine tasks,
and this transition is occurring during an era in which many workers
are already struggling. Nonetheless, with the right policies we can
get the best of both worlds: automation without rampant unemploy-
ment. Eventually, human ingenuity changes the role of productive
work. Educational opportunities will be promoted and there will be
more skilled labor with re-skilling and up-skilling.
As we will continuously deploy AI models in the wild we will
be forced to re-examine the effects of such automation on the condi-
tions of human life. Although these systems bring myriad benefits,
they also contain inherent risks, such as privacy breach, codifying
and entrenching biases, reducing accountability and hindering due
process and increasing the information asymmetry between data pro-
ducers and data holders. The IoT-CPS is a diverse and complex
network. Keeping track of every unethical or security breach inci-
dent will be difficult. Any failure or bugs in the software or hardware
will have serious consequences. Even power failure can cause a lot
of inconvenience. So, we may need another AI system on top of
such AI enabled IoT to monitor its whereabouts each instant. Some-
day, we may need a democracy of such systems which will prevent
themselves from not doing irrational things. Our lives will go on
to be increasingly controlled by technology, and we will depend
on them for everything. Whatever be the case, humans should still
have supremacy over all the man-made smartness. Only then we can
control this revolution without getting enslaved by it.
9 References
1 R. S. Michalski, J. G. Carbonell, and T. M. Mitchell, Machine Learning: An
Artificial Intelligence Approach. Springer Science & Business Media, 2013.
2 I. H. Witten and E. Frank, Data Mining: Practical Machine Learning Tools and
Techniques. Morgan Kaufmann, 2016.
3 L. Monostori, B. Kádár, T. Bauernhansl, S. Kondoh, S. Kumara, G. Rein-
hart, O. Sauer, G. Schuh, W. Sihn, and K. Ueda, “Cyber-Physical Systems in
Manufacturing,” CIRP Annals, vol. 65, no. 2, pp. 621–641, 2016.
4 E. A. Lee and S. A. Seshia, Introduction to Embedded Systems: A Cyber-Physical
Systems Approach. MIT Press, 2016.
5 Q. F. Hassan, A. R. Khan, and S. A. Madani, Internet of Things: Challenges,
Advances, and Applications. Chapman & Hall/CRC Computer and Information
Science Series, CRC Press, 2017.
6 G. Fortino and P. Trunfio,Internet of Things based on Smart Objects: Technology,
Middleware and Applications. Springer, 2014.
7 L. T. Yang, B. Di Martino, and Q. Zhang, “Internet of Everything,” Mobile
Information Systems, vol. 2017, 2017.
8 R. Baheti and H. Gill, “Cyber-Physical Systems,” The Impact of Control Technol-
ogy, vol. 12, pp. 161–166, 2011.
9 M. M. Gorman, Database Management Systems: Understanding and Applying
Database Technology. Elsevier Science, 2014.
10 S. Theodoridis and K. Koutroumbas, Pattern Recognition. Elsevier Science, 2008.
11 N. Marz and J. Warren, Big Data: Principles and Best Practices of Scalable Real-
Time Data Systems. Manning, 2015.
12 J. Leskovec, A. Rajaraman, and J. D. Ullman, Mining of Massive Datasets.
Cambridge university press, 2014.
13 J. Fan, F. Han, and H. Liu, “Challenges of Big Data Analysis,” National Science
Review, vol. 1, no. 2, pp. 293–314, 2014.
14 P. Zikopoulos, C. Eaton, et al.,Understanding Big Data: Analytics for Enterprise
Class Hadoop and Streaming Data. McGraw-Hill Osborne Media, 2011.
15 A. Ghosh, N. S. Mishra, and S. Ghosh, “Fuzzy Clustering Algorithms for Unsu-
pervised Change Detection in Remote Sensing Images,” Information Sciences,
vol. 181, no. 4, pp. 699–715, 2011.
16 A. Halder, S. Ghosh, and A. Ghosh, “AggregationPheromone Metaphor for Semi-
Supervised Classification,” Pattern Recognition, vol. 46, no. 8, pp. 2239–2248,
17 D. Cohn, “ActiveLearning,” Encyclopedia of Machine Learning and Data Mining,
pp. 9–14, 2017.
18 S. Jha and S. A. Seshia, “A Theory of Formal Synthesis via Inductive Learning,
Acta Informatica, vol. 54, no. 7, pp. 693–726, 2017.
19 S. J. Pan and Q. Yang, “A Survey on Transfer Learning,” IEEE Transactions on
Knowledge and Data Engineering, vol. 22, no. 10, pp. 1345–1359, 2010.
20 A. Ghosh and L. C. Jain, Evolutionary Computation in Data Mining. Studies in
Fuzziness and Soft Computing, Springer Berlin Heidelberg, 2006.
21 T. Hastie, R. Tibshirani, and J. Friedman, The Elements of Statistical Learning:
Data Mining, Inference, and Prediction. Springer Series in Statistics, Springer
New York, 2013.
22 J. Holler, V. Tsiatsis, C. Mulligan, S. Avesand, S. Karnouskos, and D. Boyle,
From Machine-to-Machine to the Internet of Things: Introduction to a New Age
of Intelligence. Academic Press, 2014.
23 J. Kaplan, Artificial Intelligence: What Everyone Needs to Know. What Everyone
Needs To Know, Oxford University Press, 2016.
24 D. Câmara and N. Nikaein, Wireless Public Safety Networks 2: A Systematic
Approach. Elsevier Science, 2016.
25 A. S. Appadurai and D. Kumar, “Performance Analysis of ZigBee and OWC in
Wireless Body Area Network,” Small, vol. 5, no. 3, 2016.
26 C. Gomez and J. Paradells, “Wireless Home Automation Networks: A Survey of
Architectures and Technologies,” IEEE Communications Magazine, vol.48, no. 6,
pp. 92–101, 2010.
27 Z. Shelby and C. Bormann, 6LoWPAN: The Wireless Embedded Internet, vol. 43.
John Wiley & Sons, 2011.
28 V. Coskun, B. Ozdenizci, and K. Ok, “A Survey on Near Field Communication
(NFC) Technology,Wireless Personal Communications, vol. 71, no. 3, pp. 2259–
2294, 2013.
29 C. Edwards, “Over the Hills & Far Away [Sensors and IoT],” Engineering &
Technology,vol. 11, no. 6, pp. 60–3, 2016.
30 M. Bor, J. E. Vidler, and U. Roedig, “LoRa for the Internet of Things,” in Pro-
ceedings of the 2016 International Conference on Embedded WirelessSystems and
Networks, EWSN ’16, pp. 361–366, Junction Publishing, 2016.
31 S. Jeong, The Internet of Garbage. Forbes Media, 2015.
32 N. Gershenfeld, When Things Start to Think: Integrating Digital Technology into
the Fabric of our lives. Henry Holt and Company, 2014.
33 S. Yoskovitzand S. Gal, “Automatic Mechanical System Diagnosis,” Oct. 22 2012.
US Patent App. 13/657,037.
34 B. Guo, D. Zhang, Z. Yu, Y. Liang, Z. Wang, and X. Zhou, “From the Internet of
Things to Embedded Intelligence,” World Wide Web, vol. 16, no. 4, pp. 399–420,
IET Research Journals, pp. 1–11
The Institution of Engineering and Technology 2015 11
ReView by River Valley Technologies CAAI Transactions on Intelligence Technology
2018/09/27 14:56:24 IET Review Copy Only 12
This article has been accepted for publication in a future issue of this journal, but has not been fully edited.
Content may change prior to final publication in an issue of the journal. To cite the paper please use the doi provided on the Digital Library page.
... According to Cisco (Cisco), it is expected that by 2025, the data generated by IoT devices worldwide at the edge of the network will reach 79.4 ZB [1]. Faced with the data explosion brought about by the rapid development, the cloud computing model converges IoT and AI by using the data generated by IoT devices to train machine learning models and then using the models to analyze the data (e.g., image recognition, intelligent prediction, etc.), which in turn drives the implementation of AI and IoT in various application scenarios [2]. However, this model requires all data to be aggregated to a cloud data center and then processed centrally, which can lead to huge transmission delays and privacy leaks. ...
... parameterizes the hypothetical class , i.e., the overall parameters of the classifier. In addition, the connection weights from to the output layer are denoted as = [ (1) , (2) , ⋯ , ( ) ] and ∈ . In each training iteration, we apply the backpropagation method to calculate the gradient of the loss function ( ) under . ...
Full-text available
Federated Semi-supervised Learning (FSSL) combines techniques from both fields of federated and semi-supervised learning to improve the accuracy and performance of models in a distributed environment by using a small fraction of labeled data and a large amount of unlabeled data. Without the need to centralize all data in one place for training, it collect updates of model training after devices train models at local, and thus can protect the privacy of user data. However, during the federal training process, some of the devices fail to collect enough data for local training, while new devices will be included to the group training. This leads to an unbalanced global data distribution and thus affect the performance of the global model training. Most of the current research is focusing on class imbalance with a fixed number of classes, while little attention is paid to data imbalance with a variable number of classes. Therefore, in this paper, we propose Federated Semi-supervised Learning for Class Variable Imbalance (FCVI) to solve class variable imbalance. The class-variable learning algorithm is used to mitigate the data imbalance due to changes of the number of classes. Our scheme is proved to be significantly better than baseline methods, while maintaining client privacy.
... The operation of the Internet is gradually shifting from the "Internet of Computers" (IoC) to the "Internet of things" (IoT). Massively linked systems, often called cyber-physical systems (CPSs), are also emerging due to the incorporation of many aspects such as interest, embedded devices, smart objects, humans, and physical surroundings [55]. Cloud computing is linked to the Internet of Things (IoT) and smart homes. ...
Full-text available
The normal development of “smart buildings,” which calls for integrating sensors, rich data, and artificial intelligence (AI) simulation models, promises to usher in a new era of architectural concepts. AI simulation models can improve home functions and users’ comfort and significantly cut energy consumption through better control, increased reliability, and automation. This article highlights the potential of using artificial intelligence (AI) models to improve the design and functionality of smart houses, especially in implementing living spaces. This case study provides examples of how artificial intelligence can be embedded in smart homes to improve user experience and optimize energy efficiency. Next, the article will explore and thoroughly analyze the thorough analysis of current research on the use of artificial intelligence (AI) technology in smart homes using a variety of innovative ideas, including smart interior design and a Smart Building System Framework based on digital twins (DT). Finally, the article explores the advantages of using AI models in smart homes, emphasizing living spaces. Through the case study, the theme seeks to provide ideas on how AI can be effectively embedded in smart homes to improve functionality, convenience, and energy efficiency. The overarching goal is to harness the potential of artificial intelligence by transforming how we live in our homes and improving our quality of life. The article concludes by discussing the unresolved issues and potential future research areas on the usage of AI in smart houses. Incorporating AI technology into smart homes benefits homeowners, providing excellent safety and convenience and increased energy efficiency.
... In recent years, there is an increasing demand for the convergence of AI and IoT to tackle high-performance data processing issues in IoT-driven engineering applications. The collective integration of AI and the IoT has greatly promoted the rapid development of AI-of-Things (AIoT) systems that evolve the existing IoT standards to form autonomous future communication architectures to support the intelligent exchange of data between millions of devices [1]- [2]. ...
Full-text available
As a modern communication paradigm, Artificial intelligence based Internet of Things (AIoT) can provide an interactive platform across the globe to enrich the quality of networking services. With the AIoT paradigm, coded distributed computing (CDC) has recently emerged to be a promising solution to address the straggling effects in conventional distributed computing systems. In this article, we propose a novel CDC control scheme in the AIoT platform. Based on the cooperative game theory, the main challenges of our scheme are i) the k value decision for the CDC process, and ii) edge node resource allocation for offloading tasks. Using the ideas of coalition game and weighted Nash social welfare solution (WNSWS) , our proposed scheme is developed as a two-phase game model to achieve a mutually desirable solution. At the first phase, a dynamic coalition formation is proceeded to select the most adaptable edge nodes for the offloading subtasks. At the second phase, the WNSWS is adopted to effectively share each edge node’s computing resource. Based on the jointly design of these two cooperative games, we explore the synergy effect to optimize the CDC process. In the edge assisted distributed computing infrastructure, our reciprocal combinative approach can provide a fair-efficient solution through the sequential interactions of edge nodes and AIoT devices. In the performance evaluation, we provide extensive simulation analyses to show our scheme’s superiority by comparing with the existing baseline protocols.
... related to the design of t he software, which, in addition, may imply the inability to use it for its intended purpose. These are the so-called 'ineffective products' [31]. In addition to a defect that causes damage according to the product liability rules, there is a non-conformity that allows the injured party to make a claim against the seller for such damage (Articles 6-8 of Directive 2019/771). ...
Full-text available
Introduction: when studying legal issues related to safety and adequacy in the application of artificial intelligence systems (AIS), it is impossible not to raise the subject of liability accompanying the use of AIS. In this paper we focus on the study of the civil law aspects of liability for harm caused by artificial intelligence and robotic systems. Technological progress necessitates revision of many legislative mechanisms in such a way as to maintain and encourage further development of innovative industries while ensuring safety in the application of artificial intelligence. It is essential not only to respond to the challenges of the moment but also to look forward and develop new rules based on short-term forecasts. There is no longer any reason to claim categorically that the rules governing the institute of legal responsibility will definitely not require fundamental changes, contrary to earlier belief. This is due to the growing autonomy of AIS and the expansion of the range of their possible applications. Artificial intelligence is routinely employed in creative industries, decision-making in different fields of human activity, unmanned transportation, etc. However, there remain unresolved major issues concerning the parties liable in the case of infliction of harm by AIS, the viability of applying no-fault liability mechanisms, the appropriate levels of regulation of such relations; and discussions over these issues are far from being over. Purpose: basing on an analysis of theoretical concepts and legislation in both Russia and other countries, to develop a vision of civil law regulation and tort liability in cases when artificial intelligence is used. Methods: empirical methods of comparison, description, interpretation; theoretical methods of formal and dialectical logic; specialscientific methods: legal-dogmatic and the method of interpretation of legal norms. Results: there is considerable debate over the responsibilities of AIS owners and users. In many countries, codes of ethics for artificial intelligence are accepted. However, what is required is legal regulation, for instance, considering an AIS as a source of increased danger; in the absence of relevant legal standards, it is reasonable to use a tort liability mechanism based on analogy of the law. Standardization in this area (standardization of databases, software, infrastructure, etc.) is also important – for identifying the AIS developers and operators to be held accountable; violation of standardization requirements may also be a ground for holding them liable under civil law. There appear new dimensions added to the classic legal notions such as the subject of harm, object of harm, and the party that has inflicted the harm, used with regard to both contractual and non-contractual liability. Conclusions: the research has shown that legislation of different countries currently provides soft regulation with regard to liability for harm caused by AIS. However, it is time to gradually move from the development of strategies to practical steps toward the creation of effective mechanisms aimed at minimizing the risks of harm without any persons held liable. Since the process of developing AIS involves many participants with an independent legal status (data supplier, developer, manufacturer, programmer, designer, user), it is rather difficult to establish the liable party in case something goes wrong, and many factors must be taken into account. Regarding infliction of harm to third parties, it seems logical and reasonable to treat an AIS as a source of increased danger; and in the absence of relevant legal regulations, it would be reasonable to use a tort liability mechanism by analogy of the law. The model of contractual liability requires the development of common approaches to defining the product and the consequences of violation of the terms of the contract.
... Artificial intelligence (AI) has advanced thanks to developments in Machine Learning (ML) algorithms that can mimic human thinking [10]. Deep learning (DL), a subset of machine learning (ML), is a widely used technique for extracting useful insight from enormous amounts of data by finding patterns. ...
Full-text available
The dramatic increase in the number of the Internet of Things (IoT) devices resulted in massive data being generated. This complexity mainly increases the need to offload the IoT tasks to minimize the higher latency, computation, and storage complexities of resourceful architectures such as cloud and edge computing. Even though edge computing minimizes latency-related issues, the model deployment adds new challenges when different offloading schemes or service architectures are utilized. The main aim of this paper is to minimize the latency of high-priority healthcare applications that needs immediate service using different steps. The improved Variational mode decomposition (VMD)-Random Forest (RF) architecture is used to classify the edge device application tasks into computationally intensive, time-sensitive, and priority-sensitive workloads. The tasks are mainly classified by taking different parameters as input such as the task length, network demand, delay sensitivity, and Virtual Machine (VM) utilization parameters. This step reduces the processing time of edge-based applications. For task offloading, a novel Dynamic arithmetic optimized double deep Q-network (DAO-DDQN) architecture is developed, which determines task offloading decisions based on the classification results from the VMD-optimized RF design. A Computational Access Point (CAP) has been formed using interconnected wireless access points and the CAP is used for executing the application requests sent from mobile edge devices. To improve the task processing and computational capabilities of edge devices, the Dynamic arithmetic optimization algorithm (DAOA) is employed to choose the optimal CAP for task offloading. These steps help to minimize the edge latency by simultaneously improving the edge network performance. The results show that the proposed methodology is efficient in improving the service parameters when terms of different parameters such as average delivery time, schedulability, computing delay, bandwidth consumption, communication delay, and latency. The proposed model offers a 32% improvement in scheduling rate, an 18% improvement in bandwidth consumption, and a 25% improvement in the average delivery time when compared to the existing techniques as per the simulation outcomes.
... With the continuous development of information technology, the emergence of computers with stronger computing power and more accurate sensors has also promoted the development of human beings. Cloud computing, big data, Internet of Things, and artificial intelligence have quickly rushed into various industries and disciplines to bring changes (Ghosh et al. 2018). At the same time, a more natural and harmonious human-computer interaction mode was also desired by the world (Shu et al. 2020). ...
Full-text available
The outward-bound training has been a popular manner to exercise in daily life. However, there lacks an intelligent assistant system to monitor the real-time status of users to avoid accidents during training. In order to fill this gap, this paper established an intelligent system to monitor fatigue status during outward-bound training by using surface electromyography (sEMG) signals. The system consists of three parts: a wearable device, edge node, and cloud server. First, the wearable device collects sEMG signals. Second, the edge node processes the collected sEMG signals and sends the sEMG signal features to the cloud server. Finally, the cloud server returns the results to edge node according to a stored classification model that learnt from massive historical sEMG signals. The experimental results show the effectiveness of the proposed system.
... While technological innovation is far-reaching in general, a substantial body of literature demonstrates that AI in particular can significantly impact organisations (e.g., Nortje & Grobbelaar, 2020). AI is a broad category of intelligent technologies and tools involving machine learning (Glikson & Woolley, 2020), deep learning models (Samek et al., 2018), genetic algorithms, the Internet of Things (Ghosh et al., 2018), smart robots, and virtual and augmented reality applications (Abou-Zahra et al., 2018). According to the PricewaterhouseCoopers survey, an increasing number of global businesses are beginning to see the value of AI in assisting workforce management (PwC, 2018). ...
Full-text available
Artificial Intelligence (AI) in the industry 4.0 based technologies is becoming omnipresent and changing the workforce environment from traditional to digital and virtual. While technology continues to evolve, new jobs will be created and the job landscape requires advanced competency elements for new employment forms and processes. The demand for some occupational skills will decline, while others grow, and the transition is likely disruptive, with skill requirements shifting significantly. In the Southeast Asia (SEA) region specifically, there has been a rapid increase in the adoption of AI but empirical approach to competencies is still underdeveloped. Hence, this study sought to fill this gap by conducting a systematic literature review on the competencies required during the intervention of AI in the SEA region. This study conducted a systematic review following the PRISMA publication standard. The articles were selected from two main databases of Scopus and Google Scholar. Based on the thematic analysis, the results revealed four main themes, namely i) technological competency; ii) cognitive competency; iii) social and emotional competency; and iv) change management competency. The four main themes have further produced 15 sub-themes highlighting the uniqueness of human soft skills that are still required although AI is adopted. Based on the pattern of past research, the review presented several recommendations for further consideration by scholars, organisations and communities.
The development of IoT systems based WSN denotes a significant issue on providing intelligent capabilities to verify nodes behaviors and battery constraints. Existing AI-based works have been recently emerged for the analysis of dynamic WSN systems. Unfortunately, they failed to capture the design of dynamic intelligent WSN requirements at a high abstraction level. They provide AI solutions which are related to the target system and focus on specific problems without supporting reusability and interoperability. The Model Driven Engineering (MDE) and in particular the UML/MARTE profile become promising solutions for high-level abstraction to ease the design of WSN. We propose an AI-based model driven approach for the analysis and the prediction of WSN nodes behaviors and its interaction. It starts with a high-level specification based on the UML/MARTE profile, which describes the adaptation of WSN nodes and their interaction. Then, Model-to-Text (M2T) transformations are used to generate simulation scripts for analysis of WSN on a target AI-based platform. This later focuses on the prediction of WSN nodes behaviors, network clusters interaction and analysis of battery constraints. The prediction is based on training dataset which are collected from the German Weather Service (DWD) and measured within Measurement and Sensor Technology (MST) professorship, in the Technology University of Chemnitz.
The optimal operation of photovoltaic solar panels requires efficient energy monitoring, in order to ensure perfect monitoring of energy production and its affecting factors we develop a real-time data acquisition, transmission, and logging system. In this work, a database has been generated to record the collected data related to a photovoltaic solar panel. This system is based on a powerful ESP32 microcontroller which ensures the transmission of all the collected parameters to the No SQL database managed by MongoDB software. The hardware prototype of this system requires a set of sensors to measure data related to temperature, humidity, irradiation, voltage, and current of the adopted load, and related power, the ESP32 module is in charge of reading the sensors, adapting, and sending the data to the database. Therefore, this generated database allows energy managers and data scientists to make reliable studies on the effect of recorded parameters on energy production by applying artificial intelligence methods.
Artificial intelligence is facilitating the future of smart systems and their associated services. The world is witnessing the rapid growth in smart systems with artificial intelligence (AI) algorithms. Services must include a set of features to handle future trends, intelligent decisions, security, adaptivity, scalability, user-friendliness, time-saving ability, and adaptability to new architectural models. To satisfy these objectives, smart systems and services need advancements in working models of IoTs and applications of cloud computing. The AI and IoT create a great combination in various fields; health systems, agriculture, weather forecasting, manufacturing units, prediction, etc. Some of its applications include smart homes, smart cities, and smart businesses. This chapter also discuss about the role of IoT and cloud computing in smart systems, followed by smart city development, education system, environment protection, healthcare, water reservation, security, etc.
Full-text available
Internet of Things: Challenges, Advances, and Applications provides a comprehensive introduction to IoT, related technologies, and common issues in the adoption of IoT on a large scale. It surveys recent technological advances and novel solutions for challenges in the IoT environment. Moreover, it provides detailed discussion of the utilization of IoT and its underlying technologies in critical application areas, such as smart grids, healthcare, insurance, and the automotive industry. The chapters of this book are authored by several international researchers and industry experts. This book is composed of 18 self-contained chapters that can be read, based on interest. Features: Introduces IoT, including its history, common definitions, underlying technologies, and challenges Discusses technological advances in IoT and implementation considerations Proposes novel solutions for common implementation issues Explores critical application domains, including large-scale electric power distribution networks, smart water and gas grids, healthcare and e-Health applications, and the insurance and automotive industries The book is an excellent reference for researchers and post-graduate students working in the area of IoT, or related areas. It also targets IT professionals interested in gaining deeper knowledge of IoT, its challenges, and application areas.
Over the coming decades, Artificial Intelligence will profoundly impact the way we live, work, wage war, play, seek a mate, educate our young, and care for our elderly. It is likely to greatly increase our aggregate wealth, but it will also upend our labor markets, reshuffle our social order, and strain our private and public institutions. Eventually it may alter how we see our place in the universe, as machines pursue goals independent of their creators and outperform us in domains previously believed to be the sole dominion of humans. Whether we regard them as conscious or unwitting, revere them as a new form of life or dismiss them as mere clever appliances, is beside the point. They are likely to play an increasingly critical and intimate role in many aspects of our lives. The emergence of systems capable of independent reasoning and action raises serious questions about just whose interests they are permitted to serve, and what limits our society should place on their creation and use. Deep ethical questions that have bedeviled philosophers for ages will suddenly arrive on the steps of our courthouses. Can a machine be held accountable for its actions? Should intelligent systems enjoy independent rights and responsibilities, or are they simple property? Who should be held responsible when a self-driving car kills a pedestrian? Can your personal robot hold your place in line, or be compelled to testify against you? If it turns out to be possible to upload your mind into a machine, is that still you? The answers may surprise you.
This carefully edited book reflects and advances the state of the art in the area of Data Mining and Knowledge Discovery with Evolutionary Algorithms. It emphasizes the utility of different evolutionary computing tools to various facets of knowledge discovery from databases, ranging from theoretical analysis to real-life applications. "Evolutionary Computation in Data Mining" provides a balanced mixture of theory, algorithms and applications in a cohesive manner, and demonstrates how the different tools of evolutionary computation can be used for solving real-life problems in data mining and bioinformatics.
Wireless Public Safety Networks, Volume Two: A Systematic Approach presents the latest advances in the wireless Public Safety Networks (PSNs) field, the networks established by authorities to either prepare the population for an eminent catastrophe, or those used for support during crisis and normalization phases. Maintaining communication capabilities in a disaster scenario is crucial for avoiding loss of lives and damages to property. This book examines past communication failures that have directly contributed to the loss of lives, giving readers in-depth discussions of the public networks that impact emergency management, covering social media, crowdsourcing techniques, wearable wireless sensors, moving-cells scenarios, mobility management protocols, 5G networks, broadband networks, data dissemination, and the resources of the frequency spectrum. Provides a focus on specific enabling technologies which can help the most on the deployment and usage of PSNs in real world scenarios Proposes a general framework that has the capability to fulfill the public safety requirements and dynamically adapt to different public safety situations Investigates the problem of data dissemination over PSNs, presenting a review of the state-of-the-art of different information and communication technologies.
The Internet of Things (IoT) usually refers to a world-wide network of interconnected heterogeneous objects (sensors, actuators, smart devices, smart objects, RFID, embedded computers, etc) uniquely addressable, based on standard communication protocols. Beyond such a definition, it is emerging a new definition of IoT seen as a loosely coupled, decentralized system of cooperating smart objects (SOs). A SO is an autonomous, physical digital object augmented with sensing/actuating, processing, storing, and networking capabilities. SOs are able to sense/actuate, store, and interpret information created within themselves and around the neighbouring external world where they are situated, act on their own, cooperate with each other, and exchange information with other kinds of electronic devices and human users. However, such SO-oriented IoT raises many in-the-small and in-the-large issues involving SO programming, IoT system architecture/middleware and methods/methodologies for the development of SO-based applications. This Book will specifically focus on exploring recent advances in architectures, algorithms, and applications for an Internet of Things based on Smart Objects. Topics appropriate for this Book include, but are not necessarily limited to: - Methods for SO development - IoT Networking - Middleware for SOs - Data Management for SOs - Service-oriented SOs - Agent-oriented SOs - Applications of SOs in Smart Environments: Smart Cities, Smart Health, Smart Buildings, etc. Advanced IoT Projects. © Springer International Publishing Switzerland 2014. All rights reserved.
Data Mining: Practical Machine Learning Tools and Techniques, Fourth Edition, offers a thorough grounding in machine learning concepts, along with practical advice on applying these tools and techniques in real-world data mining situations. This highly anticipated fourth edition of the most acclaimed work on data mining and machine learning teaches readers everything they need to know to get going, from preparing inputs, interpreting outputs, evaluating results, to the algorithmic methods at the heart of successful data mining approaches. Extensive updates reflect the technical changes and modernizations that have taken place in the field since the last edition, including substantial new chapters on probabilistic methods and on deep learning. Accompanying the book is a new version of the popular WEKA machine learning software from the University of Waikato. Authors Witten, Frank, Hall, and Pal include today's techniques coupled with the methods at the leading edge of contemporary research. Please visit the book companion website at It contains Powerpoint slides for Chapters 1-12. This is a very comprehensive teaching resource, with many PPT slides covering each chapter of the book Online Appendix on the Weka workbench; again a very comprehensive learning aid for the open source software that goes with the book Table of contents, highlighting the many new sections in the 4th edition, along with reviews of the 1st edition, errata, etc. Provides a thorough grounding in machine learning concepts, as well as practical advice on applying the tools and techniques to data mining projects Presents concrete tips and techniques for performance improvement that work by transforming the input or output in machine learning methods Includes a downloadable WEKA software toolkit, a comprehensive collection of machine learning algorithms for data mining tasks-in an easy-to-use interactive interface Includes open-access online courses that introduce practical applications of the material in the book.
For all the talk of the Internet of Things (IoT), most things will never really connect to the worldwide network. They will sit on an intranet and talk to the cloud through a gateway that sits mere metres away. But there will still be plenty that need their own direct connection because they will be too far away to make installing a gateway viable. Because of this demand, battle lines are now being drawn between the standards that want to bring sensors out of the field onto the IoT.