Module-5

IOT CASE STUDIES AND

FUTURE TRENDS
 Syllabus

• Vehicular IoT – Introduction

• Healthcare IoT – Introduction, Case Studies
• IoT Analytics – Introduction

03-09-2023 M-5 2
 Vehicular IoT-Introduction
• The use of connected vehicles is increasing rapidly across the globe.
• Consequently, the number of on-road accidents and mismanagement of
traffic is also increasing. The increasing number of vehicles gives rise to
the problem of parking.
• The evolution of IoT helps to form a connected vehicular environment to
manage the transportation systems efficiently. Vehicular IoT systems have
penetrated different aspects of the transportation ecosystem, including
on-road to off-road traffic management, driver safety for heavy to small
vehicles, and security in public transportation.
• In a connected vehicular environment, vehicles are capable of
communicating and sharing their information. Moreover, IoT enables a
vehicle to sense its internal and external environments to make certain
autonomous decisions.
• With the help of modern-day IoT infrastructure, a vehicle owner residing
in Earth’s northern hemisphere can very easily track his vehicular asset
remotely, even if it is in the southern hemisphere.

03-09-2023 M-5 3
 A simple architecture of a vehicular IoT system

03-09-2023 M-5 4
• The architecture of the vehicular IoT is divided into three sublayers:
device, fog, and cloud.
• Device: The device layer is the bottom-most layer, which consists of the
basic infrastructure of the connected vehicle. This layer includes the
vehicles and road side units (RSU). These vehicles contain certain sensors
which gather the internal information of the vehicles. The RSU works as a
local centralized unit that manages the data from the vehicles.
• Fog: In vehicular IoT systems, fast decision making is pertinent to avoid
accidents and traffic mismanagement. In such situations, fog computing
plays a crucial role by providing decisions in real-time, much near to the
devices. Consequently, the fog layer helps to minimize data transmission
time in a vehicular IoT system.
• Cloud: Fog computing handles the data processing near the devices to
take decisions instantaneously. For the processing of huge data, fog
computing is not enough. Therefore, in such a situation, cloud computing
is used. In a vehicular IoT system, cloud computing helps to handle
processes that involve a huge amount of data. Further, for long-term
storage, cloud computing is used as a scalable resource in vehicular IoT
systems.
03-09-2023 M-5 5
 Components of vehicular IoT
• Modern cars come equipped with different types of sensors and electronic
components. These sensors sense the internal environment of the car and
transmit the sensed data to a processor.
• The on-road deployed sensors sense the external environment and transmit the
sensed data to the centralized processor.
• Thereafter, based on requirements, the processor delivers these sensed data to
fog or cloud to perform necessary functions. These processes seem to be simple,
but practically, several components, along with their challenges, are involved in a
vehicular IoT system. Figure depicts the components required for vehicular IoT
systems.

03-09-2023 M-5 6
Sensors: In vehicular IoT, sensors monitor different environmental conditions and help
to make the system more economical, efficient, and robust. Traditionally, two types of
sensors, internal and external, are used in vehicular IoT systems.
(i) Internal: These types of sensors are placed within the vehicle. The sensors are
typically used to sense parameters that are directly associated with the vehicle. Along
with the sensors, the vehicles are equipped with different electronic components
such as processing boards and actuators. The internal sensors in a vehicle are
connected with the processor board, to which they transmit the sensed data. Further,
the sensed data are processed by the board to take certain predefined actions.
A few examples of internal sensors are GPS, fuel gauge, ultrasonic sensors, proximity
sensors, accelerometer, pressure sensors, and temperature sensors.
(ii) External: External sensors quantify information of the environment outside the
vehicle. For example, there are sensors used in the smart traffic system that are
capable of sensing vacant parking lots in a designated parking area. The still images
and videos from cameras are important inputs to generate decisions in a vehicular IoT
system. Therefore, on-road cameras are widely used as external sensors to capture
still images and videos. The captured images and videos are processed further, either
in the fog or in the cloud layer, to take certain pre-programmed actions.
As an example, camera sensor can capture the image of the license plate of an over-
speeding vehicle at a traffic signal; the image can be processed to identify the owner
of the vehicle to charge a certain amount of fine. Similarly, temperature, rainfall, and
light sensors are also used in the vehicular IoT infrastructure.
03-09-2023 M-5 7
• Satellites: In vehicular IoT systems, automatic vehicle tracking and
crash detection are among the important available features.
Satellites help the system to track vehicles and detect on-road
crashes. The satellite image is also useful for detecting on-road
congestions and road blocks.
• Wireless connectivity: As vehicular IoT deals with connected
vehicles, communication is an important enabling component. For
taking any action or making decisions, the collective data from any
sensors need processing. For transmitting the sensed data from
multiple sensors to RSU and from RSUs to the cloud. Moreover, the
high mobility of the vehicles necessitates the connectivity type to
be wireless for practical and real-time data transmission. Different
communication technologies, such as Wi-Fi, Bluetooth, and GSM,
are common in the vehicular IoT systems.
• Road Side Unit (RSU): The RSU is a static entity that works
collaboratively with any sensors. Typically, the RSUs are equipped
with sensors, communication units, and fog devices. Vehicular IoT
systems deal with time-critical applications, which need to take
decisions in real time. In such a situation, the fog devices attached
to the RSUs process the sensed data and take necessary action. If a
vehicular system involves heavy computation, the RSU transmits
the sensed data to the cloud end. Sometimes, these RSUs also work
as an intermediate communication agent between two vehicles.

03-09-2023 M-5 8
• Cloud and fog computing: In vehicular IoT systems, fog computing
handles the light-weight processes geographically closer to the
vehicles than the cloud. Consequently fog computing is used, for
faster decision making. However, for a heavy-weight process, cloud
computing is more adept for vehicular IoT systems. Cloud computing
provides more scalability of resources as compared to fog
computing.
• For example, the location and extent of short on-road congestion
from a certain location can be determined by fog computing with the
help of sensed data. Further, the congestion information can be
shared by the RSU among other onroad vehicles, thereby suggesting
that they avoid the congested road. On the other hand, for
determining regular on-road congestion, predictions are typically
handled with the help of cloud computing.
• Analytics: Similar to different IoT application domains, in vehicular
IoT, analytics is a crucial component. Vehicular IoT systems can be
made to predict different dynamic and static conditions using
analytics. For example, strong data analytics is required to predict
on-road traffic conditions that may occur at a location after an hour.

03-09-2023 M-5 9
 Advantages of vehicular IoT

The advantages of IoT are inherently included in vehicular IoT

environments. A few selected advantages of vehicular IoT are depicted
in Figure.
(i) Easy tracking: The tracking of vehicles is an essential part of
vehicular IoT. Moreover, the system must know from which location
and which vehicle the system is receiving the information. The
tracking of vehicles is straightforward; the system can collect
information at a remote location.
(ii) Fast decision making: Most of the decisions in the connected
vehicle environment are time critical. Therefore, for such an
application, fast and active decision making are pertinent for avoiding
accidents. In the vehicular IoT environment, cloud and fog computing
help to make fast decisions with the data received from the sensor-
based devices.
03-09-2023 M-5 10
(iii) Connected vehicles: A vehicular IoT system provides an opportunity
to remain connected and share information among different vehicles.
(iv) Easy management: In vehicular IoT systems consist of different types
of sensors, a communication unit, processing devices, and GPS, the
management of the vehicle becomes easy. The connectivity among
different components in a vehicular IoT enables systems to track every
activity in and around the vehicle. The IoT infrastructure helps in
managing the huge number of users located at different geographical
coordinates.
(v) Safety: Safety is one of the most important advantages of a vehicular
IoT system. With easy management of the system, all sensors placed at
different locations play an important role in providing safety to the
vehicle, its occupants, as well as the people around it.
(vi) Record: Storing different data of the transportation system is an
essential component of a vehicular IoT. The record may be of any form,
such as video footage, still images, and documentation. By taking
advantage of cloud and fog computing architecture, the vehicular IoT
systems keep all the required records in its database.

03-09-2023 M-5 11
 Crime assistance in a smart IoT
transportation system
• A case study on smart safety in a vehicular IoT infrastructure. The system
highlights a fog-framework for intelligent public safety in vehicular
environments (fog-FISVER).
• The primary aim of this system is to ensure smart transportation safety
(STS) in public bus services. The system works through the following three
steps:
(i) The vehicle is equipped with a smart surveillance system, which is capable
of executing video processing and detecting criminal activity in real time.
(ii) A fog computing architecture works as the mediator between a vehicle
and a police vehicle.
(iii) A mobile application is used to report the crime to a nearby police agent.
• Architecture: The fog-FISVER consists of different IoT components.
Moreover, the developers utilized the advantages of the low-latency fog
computing architecture for designing their system. Fog-FISVER is based on
a three-tiered architecture, as shown in Figure.
03-09-2023 M-5 12
(i) Tier 1—In-vehicle FISVER STS Fog: A fog node is placed for detecting
criminal activities. This tier accumulates the real sensed data from
within the vehicle and processes it to detect possible criminal activities
inside the vehicle. Further, this tier is responsible for creating crime-
level metadata and transferring the required information to the next
tier. For performing all the activities, Tier 1 consists of two subsystems:
Image processor and event dispatcher.
03-09-2023 M-5 13
• Image Processor: The image processor inside Tier 1 is a potent component,
which has a capability similar to the human eye for detecting criminal
activities. Developers of the system used a deep-learning-based approach for
enabling image processing techniques in the processor.
• To implement the fog computing architecture in the vehicle, a Raspberry-Pi-3
processor board is used, which is equipped with a high-quality camera.
Further, this architecture uses template matching and correlation to detect
the presence of dangerous articles (such as a pistol or a knife) in the sub-
image of a video frame. Typically, the image processor stores a set of crime
object templates in the fog-FISVER STS fog infrastructure, which is present in
Tier 2 of the system. The image processor is divided into the following three
parts:
(a) Crime definition downloader: This component periodically checks for the
presence of new crime object template definitions in fog-FISVER STS fog
infrastructure. If a new crime object template is available, it is stored locally.
(b) Crime definition storage: In order to use template matching, the crime
object template definition is required to be stored in the system. The crime
definition storage is used to store all the possible crime object template
definitions.
(c) Algorithm launcher: This component initiates the instances of the registered
algorithm in order to match the template with the video captured by the
camera attached in the vehicles. If a crime object is matched with the video,
criminal activity is confirmed.

03-09-2023 M-5 14
• Event dispatcher: This is another key component of Tier 1. The event
dispatcher is responsible for accumulating the data sensed from vehicles
and the image processor. After the successful detection of criminal activity,
the information is sent to the fog-FISVER STS fog infrastructure. The
components of the event dispatcher are as follows:
(a) Event notifier: It transfers the data to the fog-FISVER STS fog
infrastructure, after receiving it from the attached sensor nodes in the
vehicle.
(b) Data gatherer: This is an intermediate component between the event
notifier and the physical sensor; it helps to gather sensed data.
(c) Virtual sensor interface: Multiple sensors that sense data from different
locations of the vehicle are present in the system. The virtual sensor interface
helps to maintain a particular procedure to gather data. This component also
cooperates to register the sensors in the system.
(ii) Tier 2—FISVER STS Fog Infrastructure: Tier 2 works on top of the fog
architecture. Primarily, this tier has three responsibilities—keep updating the
new object template definitions, classifying events, and finding the most
suitable police vehicle to notify the event. FISVER STS fog infrastructure is
divided into two sub-components:

03-09-2023 M-5 15
• Target Object Training: Practically, there are different types of
crime objects. The system needs to be up-to-dated regarding all
crime objects. This subcomponent of Tier 2 is responsible for
creating, updating, and storing the crime object definition. The
algorithm launcher uses these definitions in Tier 1 for the
template matching process. The template definition includes
different features of the crime object such as color gradient and
shape format. A new object definition is stored in the definition
database. The database requires to be updated based on the
availability of new template definitions.
• Notification Factory: This sub-component receives notification
about the events in a different vehicle with the installed system.
Further, this component receives and validates the events. In
order to handle multiple events, it maintains a queue.
(iii) Tier 3 consists of mobile applications that are executed on the
users’ devices. The application helps a user, who witnesses a crime,
to notify the police.

03-09-2023 M-5 16
 Healthcare IoT
• Internet of Things (IoT) had a huge impact on the medical field, especially
wearable healthcare.
• The salient features of IoT encourage researchers and industries to develop new
IoT-based technologies for healthcare. These technologies have given rise to
small, power-efficient, health monitoring and diagnostic systems.
• Consequently, the development of numerous healthcare technologies and
systems has rapidly increased over the last few years. Currently, various IoT-
enabled healthcare devices are used for diagnosing human diseases, monitoring
human health conditions, caring/monitoring for elders, children, and even
infants.
• Moreover, IoT-based healthcare systems and services help to increase the
quality of life for common human beings. IoT-based healthcare devices provide
access and knowledge about human physiological conditions through hand held
devices.
• With this development, users can be aware of the risks in acquiring various
diseases and take necessary precautions to avoid preventable diseases. The IoT-
based healthcare system is very similar to the conventional IoT architectures.
• However, for IoT-based healthcare services, the sensors are specifically designed
to measure and quantify different physiological conditions of its users/patients.

03-09-2023 M-5 17
 A typical architecture for healthcare IoT is shown in Figure

03-09-2023 M-5 18
(i) Layer 1: This layer contains different physiological sensors that are
placed on the human body. These sensors collect the values of various
physiological parameters. The physiological data are analyzed to
extract meaningful information.
(ii) Layer 2: Layer 1 delivers data to Layer 2 for short-term storage and
low-level processing. The devices that belong to Layer 2 are commonly
known as local processing units (LPU) or centralized hubs. These units
collect the sensed data from the physiological sensors attached to the
body and process it based on the architecture’s requirement.
(iii) Layer 3: This layer receives the data from Layer 2 and performs
application specific high-level analytics. Typically, this layer consists of
cloud architecture or high-end servers. The data from multiple
patients, which may be from the same or different locations, are
accumulated in this layer. Post analysis of data, some inferences or
results are provided to the application in Layer 4.
(iv) Layer 4: The end-users directly interact with Layer 4 through
receiver-side applications. The modes of accessibility of these services
by an end user are typically through cellphones, computers, and
tablets.

03-09-2023 M-5 19
 Components of healthcare IoT
A typical IoT healthcare architecture is composed of several
components that are essential to generate the whole
architecture. Figure depicts different components and their usage
in an IoT healthcare system. Each of these components plays a
distinct role in the smooth execution of the system as a whole.

03-09-2023 20
M-5
(i) Sensors: We have already explained that Layer 1 mainly consists of
physiological sensors that collect the physiological parameters of the patient.
(ii) Wireless Connectivity: The communication between the wearable sensors
and the LPU is through either wired or wireless connectivity. The wireless
communication between the physiological sensors and LPU occurs with the
help of Bluetooth and ZigBee. On the other hand, the communication between
the LPU and the cloud or server takes place with Internet connectivity such as
Wi- Fi and WLAN. These communication protocols vary depending on the type
of device in use. For example, when a service is received by a cellphone, it uses
GSM. On the other hand, if the same service is received on a desktop, it can be
through Ethernet or Wi-Fi.
(iii) Privacy and Security: The privacy and security of health data is a major
concern in healthcare IoT services. In a healthcare IoT architecture, several
devices connect with the external world. Moreover, between LPU and the
server/cloud, different networking devices work via network hops to transmit
the data. In order to increase the security of the healthcare data, different
healthcare service providers and organizations are implementing healthcare
data encryption and protection schemes.

03-09-2023 M-5 21
 Few commonly used physiological sensors and their uses

03-09-2023 M-5 22
(iv) Analytics: For converting the raw data into information, analytics plays an
important role in healthcare IoT. The doctors, nurses, and patients, access the
healthcare information in a different customized format. This customization
allows each person in the system to access only the information pertinent to
their job/role. In such a scenario, analytics plays a vital role in providing
different persons in the system access to meaningful information extracted
from the raw healthcare data. Analytics is also used for diagnosing a disease
from the raw physiological data available.
(v) Cloud and Fog Computing: In a healthcare IoT system, several physiological
sensors are attached to a patient’s body and it continuously produce a huge
amount of heterogeneous data. For storing these health data, efficient storage
space is essential. These data are used for checking the patient’s history,
current health status, and future for diagnosing different diseases and the
symptoms of the patient. Typically, the cloud storage space is scalable, where
payment is made as per the usage of space. Consequently, to store health
data, cloud storage space is used. Analytics on the stored data in cloud storage
space is used for drawing various inferences. The major challenges in storage
are security and delay in accessing the data.
(vi) Interface: The interface is the most important component for users in a
healthcare IoT system. The user interface must be designed in such a way that
it can depict all the required information clearly and, if necessary, reformat or
represent it such that it is easy to understand. Moreover, an interface must
also contain all the useful information related to the services.
03-09-2023 M-5 23
 Advantages and risk of healthcare IoT
• IoT has already started to penetrate the domain of medical science. In
healthcare, IoT has become significantly popular due to its various features.
The different advantages and risks of healthcare IoT as depicted in Figure.
The major advantages of healthcare IoT can be listed as follows:
• Low cost: Healthcare IoT systems facilitate users with different services at
low cost. For example, an authorized user can easily find the availability of
the beds in a hospital with simple Internet connectivity and a web-browser-
based portal. Moreover, multiple registered users can retrieve the same
information simultaneously.

03-09-2023 M-5 24
• Real-time: In healthcare sectors, different components, such as the
condition of the patients, availability of doctors and beds in a hospital,
medical facilities with their monetary charges, can vary dynamically with
time. In such a dynamic scenario, one of the important characteristics of an
IoT-based healthcare system is real-timeliness.
• A healthcare IoT system enables users, such as doctors, end users at the
patient-side, and staff in a healthcare unit, to receive real-time updates
about the healthcare IoT components. Moreover, a healthcare IoT system
can enable a doctor to observe a patient’s health condition in real-time
even from a remote location. On the other hand, users at the patient-end
can easily take different decisions, such as where to take a patient during
critical situations. Moreover, the staff in a healthcare unit are better aware
of the current situation of their unit, which includes the number of patients
admitted, availability of the doctors and bed, total revenue of the unit, and
other such information.
• Easy management: The management of numerous tangible and intangible
entities (such as users, medical devices, facilities, costs, and security) is a
challenging task. However, healthcare IoT facilitates easy and robust
management of all the entities.

03-09-2023 M-5 25
• Automatic processing: A healthcare unit consists of multiple subsystems,
for which manual interventions are required. For example, to register a
patient with a hospital, the user may be required to enter his/her details
manually. However, automatic processing features can remove such
manual intervention with a fingerprint sensor/device. Healthcare IoT
enables end-to-end automatic processing in different units and also
consolidates the information across the whole chain: from a patient’s
registration to discharge.
• Easy record-keeping: In a healthcare IoT system, it includes a huge number
of patients, doctors, and other staff. Different patients suffer from
different types of diseases. A particular disease requires particular
treatment, which requires knowledge of a patient’s health history, along
with other details about them. Therefore, the timely delivery of health
data of the patient to the doctor is important. In such a situation, the
permanent storage of the patients’ health data along with their respective
details is essential. A healthcare IoT enables the user to keep these
records in a safe environment and deliver them to the authorized user as
per requirement. Moreover, these recorded data are accessible from any
part of the globe.
• Easy diagnosis: We have already explained that a healthcare IoT system
stores the data of the patient in a secure manner. In a healthcare IoT
system, the diagnosis of the disease becomes easier with the help of
certain learning mechanisms along with the availability of prior datasets.

03-09-2023 M-5 26
• However, in a healthcare IoT system, there are multiple risks as well. Here, we
discuss the various risks associated with a healthcare IoT system.
• Loss of connectivity: A healthcare IoT system consists of different physiological
sensors that sense and transmit the sensed data to a centralized unit.
Moreover, continuous data transmission from the patient is expected in a good
healthcare system. Intermittent connectivity may result in data loss, which may
result in a life-threatening situations for the patient. Proper and continuous
connectivity is essential in a healthcare IoT system.
• Security: A healthcare IoT system contains the health data of different patients
associated with the system. The healthcare system must keep the data
confidential. This data should not be accessible to any unauthorized person. On
the other hand, different persons and devices are associated with a healthcare
IoT system. In such a system, the risk of data tampering and unauthorized
access is quite high.
• Error: Data analytics helps a healthcare IoT system to predict the patients’
condition and diagnosis of diseases. A huge amount of data needs to be fed
into the system in order to perform accurate analytics. Moreover, the
management of a huge amount of data is a crucial task in any IoT-based
system. Particularly, in the healthcare system, errors in data may lead to
misinterpretation of symptoms and lead to the wrong diagnosis of the patient.
It is a challenging task to construct an error-free healthcare IoT architecture.
03-09-2023 M-5 27
 Case Studies
AmbuSens system
• Patients need to be transferred from primary-care to tertiary-care hospitals for
proper diagnosis and treatment.
• During the transit, the hospitals at both ends—the referring one as well as the
referred one—do not have any information about the patient’s health condition
during transit. In such situations, the hospitals are unable to suggest any
precautionary measures in the event of some emergency during transit.
• Consequently, many patients die during the transit due to lack of proper
suggestive care by medical experts. To overcome these shortcomings, the Smart
Wireless Applications and Networking (SWAN) laboratory at the IIT, Kharagpur
developed a system: AmbuSens. The system was primarily funded by the MHRD
of the Government of India.
• The primary objectives of the AmbuSens system are summarized as follows:
• Digitization and standardization of the healthcare data, which can be easily accessed by
the registered hospital authorities.
• Real-time monitoring of the patients who are in transit from one hospital to another. At
both hospitals, doctors can access the patients’ health conditions.
• Accessibility by which multiple doctors can access the patient’s health data at the same
time.
• Provision of confidentiality to the health data of the patients in the cloud.
• In the AmbuSens system, wireless physiological sensor nodes are used. These sensor
nodes make the system flexible and easy to use.
03-09-2023 M-5 28
 Architecture
• The AmbuSens system is equipped with different physiological sensors along with a
local hub. These sensors sense the physiological parameters from the patient’s body
and transmit those to a local data processing unit (LDPU). The physiological sensors
and LDPU form a wireless body area network (WBAN).
• Further, this local hub forwards the physiological data to the cloud for storing and
analyzing the health parameters. Finally, the data are accessed by different users. The
detailed layered architecture of the AmbuSens system is depicted in Figure.

03-09-2023 M-5 29
(i) Layer 1: This layer consists of multiple WBANs attached to a patient’s body.
These WBANs acquire the physiological data from the patient and transmit them to
the upper layer. The physiological sensors are heterogeneous, that is, each of these
sensors senses different parameters of the body. Moreover, the physiological
sensors require calibration for acquiring the correct data from a patient’s body. The
formation of WBAN takes place by connecting multiple physiological sensor nodes
to the LDPU so that the sensors can transmit the data to the LDPU, simultaneously.
(ii) Layer 2: In the AmbuSens system, cloud computing has an important role. Layer
2 is responsible for handling the cloud-related functions. From Layer 1, WBANs
attached to the different patients deliver data to the cloud end. The cloud is used
for the long-term analysis and storage of data in the AmbuSens system. Moreover,
the previous health records of the patients are stored in the cloud in order to
perform patient-specific analysis. A huge volume of health data is produced by the
WBANs, which are handled by the cloud with the help of big data analytics for
providing real-time analysis.
(iii) Layer 3: In the AmbuSens system, the identity of the patients remains
anonymous. An algorithm is designed to generate a dynamic hash value for each
patient in order to keep the patient’s identity anonymous. Moreover, in the
AmbuSens system, at different time instants, a new hash value is generated for the
patients. The entire hashing mechanism of the AmbuSens is performed in this
layer.
(iv) Layer 4: The users simply register into the system and use it as per
requirement.

03-09-2023 M-5 30
 Hardware
• In the AmbuSens system, a variety of hardware components are used such
as sensors, communication units, and other computing devices.
• Sensors: The sensors used in the AmbuSens system are non-invasive. The
description of the sensors used for forming the WBAN in the AmbuSens
system are as follows:
(i) Optical Pulse Sensing Probe: It senses the photoplethysmogram (PPG) signal and
transmits it to a GSR expansion module. Typically, PPG signals are sensed from the ear
lobe, fingers, or other location of the human body. Further, the GSR expansion module
transfers the sensed data to a device in real-time.
(ii) Electrocardiogram (ECG) unit and sensor: The ECG module used in AmbuSens is in
the form of a kit, which contains ECG electrodes, biophysical 9” leads, biophysical 18”
leads, alcohol swabs, and wrist strap. Typically, the ECG sensor measures the pathway
of electrical impulses through the heart to sense the heart’s responses to physical
exertion and other factors affecting cardiac health.
(iii) Electromyogram (EMG) sensor: This sensor is used to analyze and measure the
biomechanics of the human body. Particularly, the EMG sensor is used to measure
different electrical activity related to muscle contractions; it also assesses nerve
conduction, and muscle response in injured tissue.
(iv) Temperature sensor: The body temperature of patients changes with the condition
of the body. Therefore, a temperature sensor is included in the AmbuSens system,
which can easily be placed on the body of the patient.
(v) Galvanic Skin Response (GSR) sensor: The GSR sensor is used for measuring the
change in electrical characteristics of the skin.

03-09-2023 M-5 31
• Local Data Processing Unit (LDPU): In AmbuSens, all the sensors
attached to the human body sense and transmit the sensed data to a
centralized device, which is called an LDPU. An LDPU is a small
processing board with limited computation capabilities. The
connectivity between the sensors and the LDPU follows a single-hop
star topology. The LDPU is programmed in such a way that it can receive
the physiological data from multiple sensor nodes, simultaneously.
Further, it transmits the data to the cloud for long-term storage and
heavy processing.
• Communication Module: Each sensor node consists of a Bluetooth (IEEE
802.15.1 standard) module. The communication between the sensor
nodes and the LDPU takes place with the help of Bluetooth, which
supports a maximum communication range of 10 meters in line-of-sight.
The LDPU delivers the data to the cloud with 3G/4G communication.

03-09-2023 M-5 32
• In the AmbuSens system, three actors—doctor, paramedic/nurse, and
patient—are able to participate and use the services.
• The web interface is designed as per the requirements of the actors of the
system. Each of the actors has an option to log in and access the system.
The confidentiality of a patient and their physiological data is important in
a healthcare system. Therefore, the system provides different scopes for
data accessibility based on the category of an actor.
• For example, the detailed health data of a patient is accessible only to the
assigned doctor. These data may not be required for the nurse; therefore, a
nurse is unable to access the same set of data a doctor can access. The
system provides the flexibility to a patient to log in to his/her account and
download the details of his/her previous medical/treatment details.
Therefore, in AmbuSens, the database is designed in an efficient way such
that it can deliver the customized data to the respective actor.
• Each of the users has to register with the system to avail of the service of
the AmbuSens. Therefore, in this system, the registration process is also
designed in a customized fashion, that is, the details of a user to be
entered into the registration form is different for different actors. For
example, a doctor must enter his/her registration number in the
registration form.
03-09-2023 M-5 33
 IoT Analytics
• The sensors collect data from the environment and serve
different IoT-based applications. The raw data from a sensor
require processing to draw inferences.
• However, an IoT based system generates data with complex
structures; therefore, conventional data processing on these
data is not sufficient. Sophisticated data analytics are necessary
to identify hidden patterns.
• These tools include k-means, decision tree (DT), random forest
(RF), k-nearest neighbor (KNN), and density-based spatial
clustering of applications with noise (DBSCAN) algorithms.

03-09-2023 M-5 34
 Machine learning
• The term “machine learning” was coined by Arthur Lee Samuel, in
1959. He defined machine learning as a “field of study that gives
computers the ability to learn without being explicitly
programmed”.
• ML is a powerful tool that allows a computer to learn from past
experiences and its mistakes and improve itself without user
intervention.
• Typically, researchers envision IoT-based systems to be
autonomous and self-adaptive, which enhances services and user
experience. To this end, different ML models play a crucial role in
designing intelligent systems in IoT by leveraging the massive
amount of generated data and increasing the accuracy in their
operations.
• The main components of ML are statistics, mathematics, and
computer science for drawing inferences, constructing ML models,
and implementation, respectively.
• ML is an important tool, which is used by different social
networking websites such as facebook and twitter.
• Autonomous vehicles use ML to determine their paths and speeds.

03-09-2023 M-5 35
 Advantages of ML
• Applications fueled by ML open a plethora of opportunities in IoT-based
systems, from triggering actuators to identifying chronic diseases from
images of an eye.
• ML also enables a system to identify changes and to take intelligent
actions that relatively imitates that of a human. As ML demonstrates a
myriad of advantages, its popularity in IoT applications is increasing
rapidly.

03-09-2023 M-5 36
(i) Self-learner: An ML-empowered system is capable of learning from
its prior and run-time experiences, which helps in improving its
performance continuously. For example, an ML-assisted weather
monitoring system predicts the weather report of the next seven days
with high accuracy from data collected in the last six months.
(ii) Time-efficient: ML tools are capable of producing faster results as
compared to human interpretation. For example, the weather
monitoring system generates a weather prediction report for the
upcoming seven days, using data that goes back to 6–9 months. A
manual analysis of such sizeable data for predicting the weather is
difficult and time-consuming. ML is beneficial in predicting the
weather with less delay and accuracy as compared to humans.
(iii) Self-guided: An ML tool uses a huge amount of data for producing
its results. These tools have the capability of analyzing the huge
amount of data for identifying trends autonomously. As an example,
when we search for a particular item on an online e-commerce
website, an ML tool analyzes our search trends. As a result, it shows a
range of products similar to the original item that we searched for
initially.

03-09-2023 M-5 37
(iv) Minimum Human Interaction Required: In an ML algorithm, the human
does not need to participate in every step of its execution. The ML algorithm
trains itself automatically, based on available data inputs. For instance, let us
consider a healthcare system that predicts diseases. In old systems, humans
need to determine the disease by analyzing different symptoms using standard
“if– else” observations. However, the ML algorithm determines the same
disease, based on the health data available in the system and matching the
same with the symptoms of the patient.
(v) Diverse Data Handling: Typically, IoT systems consist of different sensors
and produce diverse and multi-dimensional data, which are easily analyzed by
ML algorithms. For example, consider the profit of an industry in a financial
year. Profits in such industries depend on the attendance of laborers,
consumption of raw materials, and performance of heavy machineries. The
attendance of workers is associated with an RFID based system. On the other
hand, industrial sensors help in the detection of machine failures, and a
scanner helps in tracking the consumption of raw materials. ML algorithms use
these diverse and multi-dimensional data to determine the profit of the
industry in the financial year.
(vi) Diverse Applications: ML is flexible and can be applied to different
application domains such as healthcare, industry, smart traffic, smart home,
and many others. Two similar ML algorithms may serve two different
applications.

03-09-2023 M-5 38
 Challenges in ML
• An ML algorithm utilizes a model and its corresponding input data to
produce an output. A few major challenges in ML are listed as follows:
(i) Data Description: The data acquired from different sensors are required to
be informative and meaningful. Description of data is a challenging part of
ML.
(ii) Amount of Data: In order to provide an accurate output, a model must
have sufficient amount of data. The availability of a huge amount of data is a
challenge in ML.
(iii) Erroneous Data: A dataset may contain noisy or erroneous data. The
learning of a model is heavily dependent on the quality of data. Since
erroneous data misleads the ML model, its identification is crucial.
(iv) Selection of Model: Multiple models may be suitable for serving a
particular purpose. However, one model may perform better than others. In
such cases, the proper selection of the model is pertinent for ML.
(v) Quality of Model: After the selection of a model, it is difficult to
determine the quality of the selected model. However, the quality of the
model is essential in an ML-based system.
03-09-2023 M-5 39
 Types of ML
• Typically, ML algorithms consist of four categories: (i) Supervised (ii)
Unsupervised (iii) Semi-supervised (iv) Reinforcement Learning.
• We determine the meaning of labeled- and unlabeled-data. As the
name suggests, labeled data contain certain meaningful tags, known
as labels. Typically, the labels correspond to the characteristics or
properties of the objects. For example, in a dataset containing the
images of two birds, a particular sample is tagged as a crow or a
pigeon. On the other hand, the unlabeled dataset does not have any
tags associated with them. For example, a dataset containing the
images of a bird without mentioning its name.
(i) Supervised Learning: This type of learning supervises or directs a
machine to learn certain activities using labeled datasets. The labeled
data are used as a supervisor to make the machine understand the
relation of the labels with the properties of the corresponding input
data. Consider an example of a student who tries to learn to solve
equations using a set of labeled formulas. The labels indicate the
formulae necessary for solving an equation.
03-09-2023 M-5 40
03-09-2023 M-5 41
• The student learns to solve the equation using suitable formulae from the
set. In the case of a new equation, the student tries to identify the set of
formulae necessary for solving it.
• Similarly, ML algorithms train themselves for selecting efficient formulae
for solving equations. The selection of these formulae depends primarily
on the nature of the equations to be solved.
• Supervised ML algorithms are popular in solving classification and
regression problems. Typically, the classification deals with predictive
models that are capable of approximating a mapping function from input
data to categorical output. On the other hand, regression provides the
mapping function from input data to numerical output. There are different
classification algorithms in ML.
• We discuss 3 popular classification algorithms: (i) k-nearest neighbor
(KNN), (ii) decision tree (DT), and (iii) random forest (RF).
• Regression to estimate the relationship among a set of dependent
variables with independent variables, as shown in Figure. The dependent
variables are the primary factors that we want to predict. However, these
dependent variables are affected by the independent variables.
• Let x and y be the independent and dependent variables, respectively.
Mathematically, a simple regression model is represented as:

03-09-2023 M-5 42
where β represents the amount of impact of variable x on y and ε denotes
an error. In the given equation, x0 creates β0 impact on y, which indicates
that the value of y can never be 0. Similarly, for multiple variables, say n, the
regression model is represented as:

Regression Model

03-09-2023 M-5 43
(ii) Unsupervised Learning: Unsupervised learning algorithms
use unlabeled datasets to find scientific trends.
Let us consider an example of the student similar to that
described in the case of supervised learning. As already
mentioned, unsupervised learning does not use any labels in its
operations. Instead, the ML algorithms in this category try to
identify the nature and properties of the input equation and the
nature of the formulae responsible for solving it.
Unsupervised learning algorithms try to create different clusters
based on the features of the formulae and relate it with the
input equations.
Unsupervised learning is usually applied to solve two types of
problems: clustering and association. Clustering divides the data
into multiple groups. In contrast, association discovers the
relationship or association among the data in a dataset.

03-09-2023 M-5 44
(iii) Semi-Supervised Learning: Semi-supervised learning belongs to a
category between supervised and unsupervised learning. Algorithms under
this category use a combination of both labeled and unlabeled datasets for
training. Labeled data are typically expensive and are relatively difficult to
label correctly. Unlabeled data is less expensive than labeled data.
Therefore, semi-supervised learning includes both labeled and unlabeled
dataset to design the learning model. Traditionally, semi-supervised learning
uses mostly unlabeled data, which makes it efficient to use, and capable of
overcoming samples with missing labels.
(iv) Reinforcement Learning: Reinforcement learning establishes a pattern
with the help of its experiences by interacting with the environment.
Consequently, the agent performs a crucial role in reinforcement learning
models. It aims to achieve a particular goal in an uncertain environment.
Typically, the model starts with an initial state of a problem, for which
different solutions are available.
Based on the output, the model receives either a reward or a penalty from
the environment. The output and reward act as inputs for proceeding to the
next state. Thus, reinforcement learning models continue learning iteratively
from their experiences while inducing correctness to the output.

03-09-2023 M-5 45