i i

“CP05” — 2020/11/7 — 12:08 — page 97 — #1

i i

Chapter 5
IoT Sensing and Actuation

Learning Outcomes
After reading this chapter, the reader will be able to:
• List the salient features of transducers
• Differentiate between sensors and actuators
• Characterize sensors and distinguish between types of sensors
• List the multi-faceted considerations associated with sensing
• Characterize actuators and distinguish between types of actuators
• List the multi-faceted considerations associated with actuation

5.1 Introduction
A major chunk of IoT applications involves sensing in one form or the other. Almost
all the applications in IoT—be it a consumer IoT, an industrial IoT, or just plain
hobby-based deployments of IoT solutions—sensing forms the first step. Incidentally,
actuation forms the final step in the whole operation of IoT application deployment
in a majority of scenarios. The basic science of sensing and actuation is based on
the process of transduction. Transduction is the process of energy conversion from
one form to another. A transducer is a physical means of enabling transduction.
Transducers take energy in any form (for which it is designed)—electrical, mechanical,
chemical, light, sound, and others—and convert it into another, which may be
electrical, mechanical, chemical, light, sound, and others. Sensors and actuators
are deemed as transducers. For example, in a public announcement (PA) system,
a microphone (input device) converts sound waves into electrical signals, which is
amplified by an amplifier system (a process). Finally, a loudspeaker (output device)
outputs this into audible sounds by converting the amplified electrical signals back

i i

i i
 i i

“CP05” — 2020/11/7 — 12:08 — page 98 — #2

i i

98 Introduction to Internet of Things

into sound waves. Table 5.1 outlines the basic terminological differences between
transducers, sensors, and actuators.

Table 5.1 Basic outline of the differences between transducers, sensors, and actuators

Parameters Transducers Sensors Actuators

Definition Converts Converts various forms of Converts electrical
energy from energy into electrical signals. signals into
one form to various forms of
another. energy, typically
mechanical
energy.
Domain Can be used It is an input transducer. It is an output
to represent a transducer.
sensor as well
as an actuator.
Function Can work as Used for quantifying Used for
a sensor or an environmental stimuli into converting signals
actuator but not signals. into proportional
simultaneously. mechanical or
electrical outputs.
Examples Any sensor or Humidity sensors, Temperature Motors (convert
actuator sensors, Anemometers electrical energy
(measures flow velocity), to rotary motion),
Manometers (measures fluid Force heads
pressure), Accelerometers (which impose
(measures the acceleration of a a force), Pumps
body), Gas sensors (measures (which convert
concentration of specific gas or rotary motion of
gases), and others shafts into either a
pressure or a fluid
velocity).

5.2 Sensors
Sensors are devices that can measure, or quantify, or respond to the ambient changes
in their environment or within the intended zone of their deployment. They generate
responses to external stimuli or physical phenomenon through characterization of the
input functions (which are these external stimuli) and their conversion into typically
electrical signals. For example, heat is converted to electrical signals in a temperature
sensor, or atmospheric pressure is converted to electrical signals in a barometer. A

i i

i i
 i i

“CP05” — 2020/11/7 — 12:08 — page 99 — #3

i i

IoT Sensing and Actuation 99

sensor is only sensitive to the measured property (e.g., a temperature sensor only
senses the ambient temperature of a room). It is insensitive to any other property
besides what it is designed to detect (e.g., a temperature sensor does not bother about
light or pressure while sensing the temperature). Finally, a sensor does not influence
the measured property (e.g., measuring the temperature does not reduce or increase
the temperature). Figure 5.1 shows the simple outline of a sensing task. Here, a
temperature sensor keeps on checking an environment for changes. In the event of a
fire, the temperature of the environment goes up. The temperature sensor notices this
change in the temperature of the room and promptly communicates this information
to a remote monitor via the processor.

Environment Sensing Processing Monitoring

Enent: Fire Temperature Sensor node

sensor

Figure 5.1 The outline of a simple sensing operation

The various sensors can be classified based on: 1) power requirements, 2) sensor
output, and 3) property to be measured.
• Power Requirements: The way sensors operate decides the power requirements
that must be provided for an IoT implementation. Some sensors need to be
provided with separate power sources for them to function, whereas some
sensors do not require any power sources. Depending on the requirements of
power, sensors can be of two types.
(i) Active: Active sensors do not require an external circuitry or mechanism to
provide it with power. It directly responds to the external stimuli from its
ambient environment and converts it into an output signal. For example, a
photodiode converts light into electrical impulses.
(ii) Passive: Passive sensors require an external mechanism to power them
up. The sensed properties are modulated with the sensor’s inherent
characteristics to generate patterns in the output of the sensor. For example,
a thermistor’s resistance can be detected by applying voltage difference
across it or passing a current through it.
• Output: The output of a sensor helps in deciding the additional components
to be integrated with an IoT node or system. Typically, almost all modern-day
processors are digital; digital sensors can be directly integrated to the processors.

i i

i i
 i i

“CP05” — 2020/11/7 — 12:08 — page 100 — #4

i i

100 Introduction to Internet of Things

However, the integration of analog sensors to these digital processors or IoT

nodes requires additional interfacing mechanisms such as analog to digital
converters (ADC), voltage level converters, and others. Sensors are broadly
divided into two types, depending on the type of output generated from these
sensors, as follows.

(i) Analog: Analog sensors generate an output signal or voltage, which is

proportional (linearly or non-linearly) to the quantity being measured and is
continuous in time and amplitude. Physical quantities such as temperature,
speed, pressure, displacement, strain, and others are all continuous
and categorized as analog quantities. For example, a thermometer or a
thermocouple can be used for measuring the temperature of a liquid (e.g., in
household water heaters). These sensors continuously respond to changes
in the temperature of the liquid.
(ii) Digital: These sensors generate the output of discrete time digital
representation (time, or amplitude, or both) of a quantity being measured,
in the form of output signals or voltages. Typically, binary output signals in
the form of a logic 1 or a logic 0 for ON or OFF, respectively are associated
with digital sensors. The generated discrete (non-continuous) values may
be output as a single “bit” (serial transmission), eight of which combine to
produce a single “byte” output (parallel transmission) in digital sensors.

• Measured Property: The property of the environment being measured by

the sensors can be crucial in deciding the number of sensors in an IoT
implementation. Some properties to be measured do not show high spatial
variations and can be quantified only based on temporal variations in the
measured property, such as ambient temperature, atmospheric pressure, and
others. Whereas some properties to be measured show high spatial as well
as temporal variations such as sound, image, and others. Depending on the
properties to be measured, sensors can be of two types.

(i) Scalar: Scalar sensors produce an output proportional to the magnitude

of the quantity being measured. The output is in the form of a signal or
voltage. Scalar physical quantities are those where only the magnitude of
the signal is sufficient for describing or characterizing the phenomenon and
information generation. Examples of such measurable physical quantities
include color, pressure, temperature, strain, and others. A thermometer or
thermocouple is an example of a scalar sensor that has the ability to detect
changes in ambient or object temperatures (depending on the sensor’s
configuration). Factors such as changes in sensor orientation or direction
do not affect these sensors (typically).
(ii) Vector: Vector sensors are affected by the magnitude as well as the direction
and/or orientation of the property they are measuring. Physical quantities
such as velocity and images that require additional information besides

i i

i i
 i i

“CP05” — 2020/11/7 — 12:08 — page 101 — #5

i i

IoT Sensing and Actuation 101

their magnitude for completely categorizing a physical phenomenon are

categorized as vector quantities. Measuring such quantities are undertaken
using vector sensors. For example, an electronic gyroscope, which is
commonly found in all modern aircraft, is used for detecting the changes
in orientation of the gyroscope with respect to the Earth’s orientation along
all three axes.

Points to ponder

A sensor node is made up of a combination of sensor/sensors, a processor unit, a

radio unit, and a power unit. The nodes are capable of sensing the environment
they are set to measure and communicate the information to other sensor nodes
or a remote server. Typically, a sensor node should have low-power requirements
and be wireless. This enables them to be deployed in a vast range of scenarios
and environments without the constant need for changing their power sources
or managing wires. The wireless nature of sensor nodes would also allow them
to be freely relocatable and deployed in large numbers without bothering about
managing wires. The functional outline of a typical IoT sensor node is shown in
Figure 5.2.

DC AC

Light AC-DC Converter

WiFi
Power
Temp. Bluetooth
Radio

Force Zigbee
ADC

Position Processor LoRa

Sensor

Speed RFID/NFC

Pressure Z-Wave
Actuator
(Optional)
Chemical GSM/3G/5G
Interface

Pneumatic Hyderaulic

Electrical Mechanical

Figure 5.2 The functional blocks of a typical sensor node in IoT

i i

i i
 i i

“CP05” — 2020/11/7 — 12:08 — page 102 — #6

i i

102 Introduction to Internet of Things

Figure 5.3 shows some commercially available sensors used for sensing applications.

(a) Camera sensor (b) Color sensor (c) Compass and (d) Current sensor (e) Digital temperature
barometer and humidity sensor

(f) Flame sensor (g) Gas sensor (h) Infrared sensor (i) Rainfall sensor (j) Ultrasonic
distance
measurement
sensor

Figure 5.3 Some common commercially available sensors used for IoT-based sensing applications

5.3 Sensor Characteristics

All sensors can be defined by their ability to measure or capture a certain phenomenon
and report them as output signals to various other systems. However, even within the
same sensor type and class, sensors can be characterized by their ability to sense the
phenomenon based on the following three fundamental properties.
• Sensor Resolution: The smallest change in the measurable quantity that a sensor
can detect is referred to as the resolution of a sensor. For digital sensors, the
smallest change in the digital output that the sensor is capable of quantifying is
its sensor resolution. The more the resolution of a sensor, the more accurate is the
precision. A sensor’s accuracy does not depend upon its resolution. For example,
a temperature sensor A can detect up to 0.5◦ C changes in temperature; whereas
another sensor B can detect up to 0.25◦ C changes in temperature. Therefore, the
resolution of sensor B is higher than the resolution of sensor A.
• Sensor Accuracy: The accuracy of a sensor is the ability of that sensor to measure
the environment of a system as close to its true measure as possible. For example,
a weight sensor detects the weight of a 100 kg mass as 99.98 kg. We can say that
this sensor is 99.98% accurate, with an error rate of ±0.02%.
• Sensor Precision: The principle of repeatability governs the precision of a sensor.
Only if, upon multiple repetitions, the sensor is found to have the same error

i i

i i
 i i

“CP05” — 2020/11/7 — 12:08 — page 103 — #7

i i

IoT Sensing and Actuation 103

rate, can it be deemed as highly precise. For example, consider if the same
weight sensor described earlier reports measurements of 98.28 kg, 100.34 kg, and
101.11 kg upon three repeat measurements for a mass of actual weight of 100 kg.
Here, the sensor precision is not deemed high because of significant variations in
the temporal measurements for the same object under the same conditions.

Points to ponder

The more the resolution of a sensor, the more accurate is the precision. A sensor’s
accuracy does not depend upon its resolution.

5.4 Sensorial Deviations

In this section, we will discuss the various sensorial deviations that are considered as
errors in sensors. Most of the sensing in IoT is non-critical, where minor deviations
in sensorial outputs seldom change the nature of the undertaken tasks. However,
some critical applications of IoT, such as healthcare, industrial process monitoring,
and others, do require sensors with high-quality measurement capabilities. As the
quality of the measurement obtained from a sensor is dependent on a large number of
factors, there are a few primary considerations that must be incorporated during the
sensing of critical systems.
In the event of a sensor’s output signal going beyond its designed maximum and
minimum capacity for measurement, the sensor output is truncated to its maximum
or minimum value, which is also the sensor’s limits. The measurement range between
a sensor’s characterized minimum and maximum values is also referred to as the full-
scale range of that sensor. Under real conditions, the sensitivity of a sensor may differ
from the value specified for that sensor leading to sensitivity error. This deviation is
mostly attributed to sensor fabrication errors and its calibration.
If the output of a sensor differs from the actual value to be measured by a constant,
the sensor is said to have an offset error or bias. For example, while measuring an actual
temperature of 0◦ C, a temperature sensor outputs 1.1◦ C every time. In this case, the
sensor is said to have an offset error or bias of 1.1◦ C.
Similarly, some sensors have a non-linear behavior. If a sensor’s transfer function
(TF) deviates from a straight line transfer function, it is referred to as its non-linearity.
The amount a sensor’s actual output differs from the ideal TF behavior over the full
range of the sensor quantifies its behavior. It is denoted as the percentage of the
sensor’s full range. Most sensors have linear behavior. If the output signal of a sensor
changes slowly and independently of the measured property, this behavior of the
sensor’s output is termed as drift. Physical changes in the sensor or its material may
result in long-term drift, which can span over months or years. Noise is a temporally
varying random deviation of signals.

i i

i i
 i i

“CP05” — 2020/11/7 — 12:08 — page 104 — #8

i i

104 Introduction to Internet of Things

In contrast, if a sensor’s output varies/deviates due to deviations in the sensor’s

previous input values, it is referred to as hysteresis error. The present output of
the sensor depends on the past input values provided to the sensor. Typically, the
phenomenon of hysteresis can be observed in analog sensors, magnetic sensors, and
during heating of metal strips. One way to check for hysteresis error is to check how
the sensor’s output changes when we first increase, then decrease the input values
to the sensor over its full range. It is generally denoted as a positive and negative
percentage variation of the full-range of that sensor.
Focusing on digital sensors, if the digital output of a sensor is an approximation
of the measured property, it induces quantization error. This error can be defined as
the difference between the actual analog signal and its closest digital approximation
during the sampling stage of the analog to digital conversion. Similarly, dynamic
errors caused due to mishandling of sampling frequencies can give rise to aliasing
errors. Aliasing leads to different signals of varying frequencies to be represented as
a single signal in case the sampling frequency is not correctly chosen, resulting in the
input signal becoming a multiple of the sampling rate.
Finally, the environment itself plays a crucial role in inducing sensorial deviations.
Some sensors may be prone to external influences, which may not be directly linked
to the property being measured by the sensor. This sensitivity of the sensor may lead
to deviations in its output values. For example, as most sensors are semiconductor-
based, they are influenced by the temperature of their environment.

5.5 Sensing Types

Sensing can be broadly divided into four different categories based on the nature
of the environment being sensed and the physical sensors being used to do so
(Figure 5.4): 1) scalar sensing, 2) multimedia sensing, 3) hybrid sensing, and 4) virtual
sensing—[2].

5.5.1 Scalar sensing

Scalar sensing encompasses the sensing of features that can be quantified simply by
measuring changes in the amplitude of the measured values with respect to time [3].
Quantities such as ambient temperature, current, atmospheric pressure, rainfall, light,
humidity, flux, and others are considered as scalar values as they normally do not have
a directional or spatial property assigned with them. Simply measuring the changes
in their values with passing time provides enough information about these quantities.
The sensors used for measuring these scalar quantities are referred to as scalar sensors,
and the act is known as scalar sensing. Figures 5.3(b), 5.3(d), 5.3(e), 5.3(f), 5.3(g), 5.3(h),
5.3(i), and 5.3(j) show scalar sensors. A simple scalar temperature sensing of a fire
detection event is shown in Figure 5.4(a).

i i

i i


IoT Sensing and Actuation 105

(a) Scalar sensing (b) Multimedia sensing

(c) Hybrid sensing (d) Virtual sensing

Figure 5.4 The different sensing types commonly encountered in IoT

5.5.2 Multimedia sensing

Multimedia sensing encompasses the sensing of features that have a spatial variance
property associated with the property of temporal variance [4]. Unlike scalar sensors,
multimedia sensors are used for capturing the changes in amplitude of a quantifiable
property concerning space (spatial) as well as time (temporal). Quantities such
as images, direction, flow, speed, acceleration, sound, force, mass, energy, and
momentum have both directions as well as a magnitude. Additionally, these quantities
follow the vector law of addition and hence are designated as vector quantities. They
might have different values in different directions for the same working condition
at the same time. The sensors used for measuring these quantities are known as
vector sensors. Figures 5.3(a) and 5.3(c) are vector sensors. A simple camera-based
multimedia sensing using surveillance as an example is shown in Figure 5.4(b).

5.5.3 Hybrid sensing

The act of using scalar as well as multimedia sensing at the same time is referred to
as hybrid sensing. Many a time, there is a need to measure certain vector as well as
scalar properties of an environment at the same time. Under these conditions, a range
of various sensors are employed (from the collection of scalar as well as multimedia
sensors) to measure the various properties of that environment at any instant of


 i i

“CP05” — 2020/11/7 — 12:08 — page 106 — #10

i i

106 Introduction to Internet of Things

time, and temporally map the collected information to generate new information.
For example, in an agricultural field, it is required to measure the soil conditions at
regular intervals of time to determine plant health. Sensors such as soil moisture
and soil temperature are deployed underground to estimate the soil’s water retention
capacity and the moisture being held by the soil at any instant of time. However,
this setup only determines whether the plant is getting enough water or not. There
may be a host of other factors besides water availability, which may affect a plant’s
health. The additional inclusion of a camera sensor with the plant may be able
to determine the actual condition of a plant by additionally determining the color
of leaves. The aggregate information from soil moisture, soil temperature, and the
camera sensor will be able to collectively determine a plant’s health at any instant
of time. Other common examples of hybrid sensing include smart parking systems,
traffic management systems, and others. Figure 5.4(c) shows an example of hybrid
sensing, where a camera and a temperature sensor are collectively used to detect and
confirm forest fires during wildlife monitoring.

5.5.4 Virtual sensing

Many a time, there is a need for very dense and large-scale deployment of sensor
nodes spread over a large area for monitoring of parameters. One such domain is
agriculture [5]. Here, often, the parameters being measured, such as soil moisture,
soil temperature, and water level, do not show significant spatial variations. Hence, if
sensors are deployed in the fields of farmer A, it is highly likely that the measurements
from his sensors will be able to provide almost concise measurements of his neighbor
B’s fields; this is especially true of fields which are immediately surrounding A’s
fields. Exploiting this property, if the data from A’s field is digitized using an
IoT infrastructure and this system advises him regarding the appropriate watering,
fertilizer, and pesticide regimen for his crops, this advisory can also be used by B for
maintaining his crops. In short, A ’s sensors are being used for actual measurement
of parameters; whereas virtual data (which does not have actual physical sensors
but uses extrapolation-based measurements) is being used for advising B. This is the
virtual sensing paradigm. Figure 5.4(d) shows an example of virtual sensing. Two
temperature sensors S1 and S3 monitor three nearby events E1, E2, and E3 (fires). The
event E2 does not have a dedicated sensor for monitoring it; however, through the
superposition of readings from sensors S1 and S3, the presence of fire in E2 is inferred.

5.6 Sensing Considerations

The choice of sensors in an IoT sensor node is critical and can either make or break the
feasibility of an IoT deployment. The following major factors influence the choice of
sensors in IoT-based sensing solutions: 1) sensing range, 2) accuracy and precision, 3)
energy, and 4) device size. These factors are discussed as follows:

i i

i i
 i i

“CP05” — 2020/11/7 — 12:08 — page 107 — #11

i i

IoT Sensing and Actuation 107

(i) Sensing Range: The sensing range of a sensor node defines the detection fidelity
of that node. Typical approaches to optimize the sensing range in deployments
include fixed k-coverage and dynamic k-coverage. A lifelong fixed k-coverage
tends to usher in redundancy as it requires a large number of sensor nodes,
the sensing range of some of which may also overlap. In contrast, dynamic k-
coverage incorporates mobile sensor nodes post detection of an event, which,
however, is a costly solution and may not be deployable in all operational areas
and terrains [1].
Additionally, the sensing range of a sensor may also be used to signify the upper
and lower bounds of a sensor’s measurement range. For example, a proximity
sensor has a typical sensing range of a couple of meters. In contrast, a camera
has a sensing range varying between tens of meters to hundreds of meters. As
the complexity of the sensor and its sensing range goes up, its cost significantly
increases.
(ii) Accuracy and Precision: The accuracy and precision of measurements provided
by a sensor are critical in deciding the operations of specific functional processes.
Typically, off-the-shelf consumer sensors are low on requirements and often very
cheap. However, their performance is limited to regular application domains.
For example, a standard temperature sensor can be easily integrated with
conventional components for hobby projects and day-to-day applications, but it
is not suitable for industrial processes. Regular temperature sensors have a very
low-temperature sensing range, as well as relatively low accuracy and precision.
The use of these sensors in industrial applications, where a precision of up to
3–4 decimal places is required, cannot be facilitated by these sensors. Industrial
sensors are typically very sophisticated, and as a result, very costly. However,
these industrial sensors have very high accuracy and precision score, even under
harsh operating conditions.
(iii) Energy: The energy consumed by a sensing solution is crucial to determine the
lifetime of that solution and the estimated cost of its deployment. If the sensor
or the sensor node is so energy inefficient that it requires replenishment of its
energy sources quite frequently, the effort in maintaining the solution and its
cost goes up; whereas its deployment feasibility goes down. Consider a scenario
where sensor nodes are deployed on the top of glaciers. Once deployed, access
to these nodes is not possible. If the energy requirements of the sensor nodes are
too high, such a deployment will not last long, and the solution will be highly
infeasible as charging or changing of the energy sources of these sensor nodes is
not an option.
(iv) Device Size: Modern-day IoT applications have a wide penetration in all
domains of life. Most of the applications of IoT require sensing solutions which
are so small that they do not hinder any of the regular activities that were
possible before the sensor node deployment was carried out. Larger the size
of a sensor node, larger is the obstruction caused by it, higher is the cost and

i i

i i


108 Introduction to Internet of Things

energy requirements, and lesser is its demand for the bulk of the IoT applications.
Consider a simple human activity detector. If the detection unit is too large to
be carried or too bulky to cause hindrance to regular normal movements, the
demand for this solution would be low. It is because of this that the onset of
wearables took off so strongly. The wearable sensors are highly energy-efficient,
small in size, and almost part of the wearer’s regular wardrobe.

Check yourself

Principle of virtualization, MEMS

5.7 Actuators
An actuator can be considered as a machine or system’s component that can affect
the movement or control the said mechanism or the system. Control systems affect
changes to the environment or property they are controlling through actuators. The
system activates the actuator through a control signal, which may be digital or analog.
It elicits a response from the actuator, which is in the form of some form of mechanical
motion. The control system of an actuator can be a mechanical or electronic system,
a software-based system (e.g., an autonomous car control system), a human, or any
other input. Figure 5.5 shows the outline of a simple actuation system. A remote
user sends commands to a processor. The processor instructs a motor controlled
robotic arm to perform the commanded tasks accordingly. The processor is primarily
responsible for converting the human commands into sequential machine-language
command sequences, which enables the robot to move. The robotic arm finally moves
the designated boxes, which was its assigned task.

Monitoring Processing Actuation Environment

Event: Factory
Sensor node Motor-driven automation
mechanism

Figure 5.5 The outline of a simple actuation mechanism


 i i

“CP05” — 2020/11/7 — 12:08 — page 109 — #13

i i

IoT Sensing and Actuation 109

5.8 Actuator Types

Broadly, actuators can be divided into seven classes: 1) Hydraulic, 2) pneumatic, 3)
electrical, 4) thermal/magnetic, 5) mechanical, 6) soft, and 7) shape memory polymers.
Figure 5.6 shows some of the commonly used actuators in IoT applications.

5.8.1 Hydraulic actuators

A hydraulic actuator works on the principle of compression and decompression
of fluids. These actuators facilitate mechanical tasks such as lifting loads through
the use of hydraulic power derived from fluids in cylinders or fluid motors. The
mechanical motion applied to a hydraulic actuator is converted to either linear,
rotary, or oscillatory motion. The almost incompressible property of liquids is used
in hydraulic actuators for exerting significant force. These hydraulic actuators are also
considered as stiff systems. The actuator’s limited acceleration restricts its usage.

5.8.2 Pneumatic actuators

A pneumatic actuator works on the principle of compression and decompression
of gases. These actuators use a vacuum or compressed air at high pressure and
convert it into either linear or rotary motion. Pneumatic rack and pinion actuators are
commonly used for valve controls of water pipes. Pneumatic actuators are considered
as compliant systems. The actuators using pneumatic energy for their operation are
typically characterized by the quick response to starting and stopping signals. Small
pressure changes can be used for generating large forces through these actuators.
Pneumatic brakes are an example of this type of actuator which is so responsive that
they can convert small pressure changes applied by drives to generate the massive
force required to stop or slow down a moving vehicle. Pneumatic actuators are
responsible for converting pressure into force. The power source in the pneumatic
actuator does not need to be stored in reserve for its operation.

5.8.3 Electric actuators

Typically, electric motors are used to power an electric actuator by generating
mechanical torque. This generated torque is translated into the motion of a motor’s
shaft or for switching (as in relays). For example, actuating equipments such as
solenoid valves control the flow of water in pipes in response to electrical signals.
This class of actuators is considered one of the cheapest, cleanest and speedy actuator
types available. Figures 5.6(a), 5.6(b), 5.6(c), 5.6(d), 5.6(e), 5.6(f), 5.6(i), and 5.6(j) show
some of the commonly used electrical actuators.

i i

i i
 i i

“CP05” — 2020/11/7 — 12:08 — page 110 — #14

i i

110 Introduction to Internet of Things

(a) Brushless DC (b) Brushless DC (c) Stepper motor (d) Geared (e) DC motor
motor motor stepper motor

(f) Relay array (g) Hydroelectric (h) Hydroelectric (i) Solenoid-based (j) Solenoid-based
generator generator flow valve flow valve

(k) DPDT switch (l) Push button

switch

Figure 5.6 Some common commercially available actuators used for IoT-based control
applications

5.8.4 Thermal or magnetic actuators

The use of thermal or magnetic energy is used for powering this class of actuators.
These actuators have a very high power density and are typically compact,
lightweight, and economical. One classic example of thermal actuators is shape
memory materials (SMMs) such as shape memory alloys (SMAs). These actuators do
not require electricity for actuation. They are not affected by vibration and can work
with liquid or gases. Magnetic shape memory alloys (MSMAs) are a type of magnetic
actuators.

5.8.5 Mechanical actuators

In mechanical actuation, the rotary motion of the actuator is converted into linear
motion to execute some movement. The use of gears, rails, pulleys, chains, and
other devices are necessary for these actuators to operate. These actuators can be
easily used in conjunction with pneumatic, hydraulic, or electrical actuators. They
can also work in a standalone mode. The best example of a mechanical actuator is
a rack and pinion mechanism. Figures 5.6(g), 5.6(h), 5.6(k), and 5.6(l) show some of
the commonly available mechanical actuators. The hydroelectric generator shown in

i i

i i
 i i

“CP05” — 2020/11/7 — 12:08 — page 111 — #15

i i

IoT Sensing and Actuation 111

Figures 5.6(g) and 5.6(h) convert the water-flow induced rotary motion of a turbine
into electrical energy. Similarly, the mechanical switches shown in Figures 5.6 (k) and
5.6(l) uses the mechanical motion of the switch to switch on or off an electrical circuit.

5.8.6 Soft actuators

Soft actuators (e.g., polymer-based) consists of elastomeric polymers that are used as
embedded fixtures in flexible materials such as cloth, paper, fiber, particles, and others
[7]. The conversion of molecular level microscopic changes into tangible macroscopic
deformations is the primary working principle of this class of actuators. These
actuators have a high stake in modern-day robotics. They are designed to handle
fragile objects such as agricultural fruit harvesting, or performing precise operations
like manipulating the internal organs during robot-assisted surgeries.

5.8.7 Shape memory polymers

Shape memory polymers (SMP) are considered as smart materials that respond to
some external stimulus by changing their shape, and then revert to their original shape
once the affecting stimulus is removed [6]. Features such as high strain recovery,
biocompatibility, low density, and biodegradability characterize these materials.
SMP-based actuators function similar to our muscles. Modern-day SMPs have been
designed to respond to a wide range of stimuli such as pH changes, heat differentials,
light intensity, and frequency changes, magnetic changes, and others.
Photopolymer/light-activated polymers (LAP) are a particular type of SMP, which
require light as a stimulus to operate. LAP-based actuators are characterized by their
rapid response times. Using only the variation of light frequency or its intensity, LAPs
can be controlled remotely without any physical contact. The development of LAPs
whose shape can be changed by the application of a specific frequency of light have
been reported. The polymer retains its shape after removal of the activating light. In
order to change the polymer back to its original shape, a light stimulus of a different
frequency has to be applied to the polymer.

5.9 Actuator Characteristics

The choice or selection of actuators is crucial in an IoT deployment, where a control
mechanism is required after sensing and processing of the information obtained from
the sensed environment. Actuators perform the physically heavier tasks in an IoT
deployment; tasks which require moving or changing the orientation of physical
objects, changing the state of objects, and other such activities. The correct choice
of actuators is necessary for the long-term sustenance and continuity of operations,
as well as for increasing the lifetime of the actuators themselves. A set of four
characteristics can define all actuators:

i i

i i
 i i

“CP05” — 2020/11/7 — 12:08 — page 112 — #16

i i

112 Introduction to Internet of Things

• Weight: The physical weight of actuators limits its application scope. For
example, the use of heavier actuators is generally preferred for industrial
applications and applications requiring no mobility of the IoT deployment. In
contrast, lightweight actuators typically find common usage in portable systems
in vehicles, drones, and home IoT applications. It is to be noted that this is not
always true. Heavier actuators also have selective usage in mobile systems, for
example, landing gears and engine motors in aircraft.
• Power Rating: This helps in deciding the nature of the application with which
an actuator can be associated. The power rating defines the minimum and
maximum operating power an actuator can safely withstand without damage
to itself. Generally, it is indicated as the power-to-weight ratio for actuators.
For example, smaller servo motors used in hobby projects typically have a
maximum rating of 5 VDC, 500 mA, which is suitable for an operations-driven
battery-based power source. Exceeding this limit might be detrimental to the
performance of the actuator and may cause burnout of the motor. In contrast to
this, servo motors in larger applications have a rating of 460 VAC, 2.5 A, which
requires standalone power supply systems for operations. It is to be noted that
actuators with still higher ratings are available and vary according to application
requirements.
• Torque to Weight Ratio: The ratio of torque to the weight of the moving part of
an instrument/device is referred to as its torque/weight ratio. This indicates the
sensitivity of the actuator. Higher is the weight of the moving part; lower will be
its torque to weight ratio for a given power.
• Stiffness and Compliance: The resistance of a material against deformation
is known as its stiffness, whereas compliance of a material is the opposite of
stiffness. Stiffness can be directly related to the modulus of elasticity of that
material. Stiff systems are considered more accurate than compliant systems as
they have a faster response to the change in load applied to it. For example,
hydraulic systems are considered as stiff and non-compliant, whereas pneumatic
systems are considered as compliant.

Check yourself

Operation of PLC and SCADA, Working principle of electric motors, applications

of pneumatic and hydraulic actuators, Differences between pneumatic, hydraulic,
electrical, and mechanical actuators

Summary
This chapter covered the basics of sensing and actuation in order to help the readers
grasp the intricacies of designing an IoT solution keeping in mind the need to select

i i

i i
 i i

“CP05” — 2020/11/7 — 12:08 — page 113 — #17

i i

IoT Sensing and Actuation 113

the proper sensors and actuators. The first part of this chapter discusses sensors,
sensing characteristics, considerations of various sensorial deviations, and the sensing
types possible in a typical IoT-based implementation of a sensing solution. This part
concludes with a discussion on the various considerations to be thought of while
selecting sensors for architecting a viable IoT-based sensing solution. The second
part of this chapter focuses on actuators and the broad classes of actuators available.
This part concludes with a discussion on the various considerations to be thought
of while selecting actuators for architecting a viable IoT-based control solution using
actuators. After completing this chapter, the reader will be able to decide upon
the most appropriate sensing and actuation solutions to use with their IoT-based
applications.

Exercises
(i) Differentiate between sensors and actuators.
(ii) Differentiate between sensors and transducers.
(iii) How is sensor resolution different from its accuracy?
(iv) Differentiate between scalar and vector sensors.
(v) Differentiate between analog and digital sensors.
(vi) What is a an offset error?
(vii) What is a hysteresis error?
(viii) What is a quantization error?
(ix) What is aliasing error?
(x) Differentiate between hydraulic and pneumatic actuators with examples.
(xi) What are shape memory alloys (SMA)?
(xii) What are soft actuators?
(xiii) What are the main features of shape memory polymers?
(xiv) What are light activated polymers?

References
[1] Alam, Kh Mahmudul, Joarder Kamruzzaman, Gour Karmakar, and Manzur Murshed.
2014. “Dynamic Adjustment of Sensing Range for Event Coverage in Wireless Sensor
Networks.” Journal of Network and Computer Applications 46: 139–153. Elsevier.
[2] Popović, T., N. Latinović, A. Pes̆ić, Z. Zec̆ević, B. Krstajić, and S. Djukanović. 2017.
“Architecting an IoT-enabled Platform for Precision Agriculture and Ecological
Monitoring: A Case Study.” Computers and Electronics in Agriculture 140: 255–265.

i i

i i
 i i

“CP05” — 2020/11/7 — 12:08 — page 114 — #18

i i

114 Introduction to Internet of Things

[3] Kelly, S. D. T., N. K. Suryadevara, and S. C. Mukhopadhyay. 2013. “Towards the

Implementation of IoT for Environmental Condition Monitoring in Homes.” IEEE Sensors
Journal 13(10): 3846–3853.
[4] Rosário, D., Z. Zhao, A. Santos, T. Braun, and E. Cerqueira. 2014. “A Beaconless
Opportunistic Routing based on a Cross-layer Approach for Efficient Video Dissemination
in Mobile Multimedia IoT Applications.” Computer Communications 45: 21–31.
[5] Ojha, T., S. Misra, N. S. Raghuwanshi, and H. Poddar. 2019. “DVSP: Dynamic Virtual
Sensor Provisioning in Sensor-Cloud based Internet of Things.” IEEE Internet of Things
Journal 6 (3): 5265–5272.
[6] Lendlein, A., and O. E. Gould. 2019. “Reprogrammable Recovery and Actuation Behaviour
of Shape-memory Polymers.” Nature Reviews Materials 4(2): 116–133.
[7] Lessing, J. A., R. V. Martinez, A. S. Tayi, J. M. Ting, and G. M. Whitesides. 2019.
“Flexible and Stretchable Electronic Strain-limited Layer for Soft Actuators.” U. S. Patent
Application 15/972, 412.

i i

i i
 i i

“CP06” — 2020/11/7 — 12:08 — page 115 — #1

i i

Chapter 6
IoT Processing Topologies and
Types

Learning Outcomes
After reading this chapter, the reader will be able to:
• List common data types in IoT applications
• Understand the importance of processing
• Explain the various processing topologies in IoT
• Understand the importance of processing off-loading toward achieving scalability
and cost-effectiveness of IoT solutions
• Determine the importance of choosing the right processing topologies and
associated considerations while designing IoT applications
• Determine the requirements that are associated with IoT-based processing of
sensed and communicated data.

6.1 Data Format

The Internet is a vast space where huge quantities and varieties of data are generated
regularly and flow freely. As of January 2018, there are a reported 4.021 billion Internet
users worldwide. The massive volume of data generated by this huge number of users
is further enhanced by the multiple devices utilized by most users. In addition to these
data-generating sources, non-human data generation sources such as sensor nodes
and automated monitoring systems further add to the data load on the Internet. This
huge data volume is composed of a variety of data such as e-mails, text documents
(Word docs, PDFs, and others), social media posts, videos, audio files, and images,
as shown in Figure 6.1. However, these data can be broadly grouped into two types

i i

i i
 i i

“CP06” — 2020/11/7 — 12:08 — page 116 — #2

i i

116 Introduction to Internet of Things

based on how they can be accessed and stored: 1) Structured data and 2) unstructured
data.

Databases Cloud

Internet

Industry Transport Buildings Traffic Transaction Human

Figure 6.1 The various data generating and storage sources connected to the Internet and the
plethora of data types contained within it

6.1.1 Structured data

These are typically text data that have a pre-defined structure [1]. Structured data
are associated with relational database management systems (RDBMS). These are
primarily created by using length-limited data fields such as phone numbers, social
security numbers, and other such information. Even if the data is human or machine-
generated, these data are easily searchable by querying algorithms as well as human-
generated queries. Common usage of this type of data is associated with flight or
train reservation systems, banking systems, inventory controls, and other similar
systems. Established languages such as Structured Query Language (SQL) are used
for accessing these data in RDBMS. However, in the context of IoT, structured data
holds a minor share of the total generated data over the Internet.

6.1.2 Unstructured data

In simple words, all the data on the Internet, which is not structured, is categorized as
unstructured. These data types have no pre-defined structure and can vary according
to applications and data-generating sources. Some of the common examples of
human-generated unstructured data include text, e-mails, videos, images, phone

i i

i i
 i i

“CP06” — 2020/11/7 — 12:08 — page 117 — #3

i i

IoT Processing Topologies and Types 117

recordings, chats, and others [2]. Some common examples of machine-generated

unstructured data include sensor data from traffic, buildings, industries, satellite
imagery, surveillance videos, and others. As already evident from its examples, this
data type does not have fixed formats associated with it, which makes it very difficult
for querying algorithms to perform a look-up. Querying languages such as NoSQL
are generally used for this data type.

6.2 Importance of Processing in IoT

The vast amount and types of data flowing through the Internet necessitate the need
for intelligent and resourceful processing techniques. This necessity has become even
more crucial with the rapid advancements in IoT, which is laying enormous pressure
on the existing network infrastructure globally. Given these urgencies, it is important
to decide—when to process and what to process? Before deciding upon the processing to
pursue, we first divide the data to be processed into three types based on the urgency
of processing: 1) Very time critical, 2) time critical, and 3) normal. Data from sources
such as flight control systems [3], healthcare, and other such sources, which need
immediate decision support, are deemed as very critical. These data have a very low
threshold of processing latency, typically in the range of a few milliseconds.
Data from sources that can tolerate normal processing latency are deemed as time-
critical data. These data, generally associated with sources such as vehicles, traffic,
machine systems, smart home systems, surveillance systems, and others, which can
tolerate a latency of a few seconds fall in this category. Finally, the last category of
data, normal data,can tolerate a processing latency of a few minutes to a few hours
and are typically associated with less data-sensitive domains such as agriculture,
environmental monitoring, and others.
Considering the requirements of data processing, the processing requirements
of data from very time-critical sources are exceptionally high. Here, the need for
processing the data in place or almost nearer to the source is crucial in achieving
the deployment success of such domains. Similarly, considering the requirements of
processing from category 2 data sources (time-critical), the processing requirements
allow for the transmission of data to be processed to remote locations/processors such
as clouds or through collaborative processing. Finally, the last category of data sources
(normal) typically have no particular time requirements for processing urgently and
are pursued leisurely as such.

Check yourself

Difference between microprocessors and microcontrollers, network bandwidth,

network latency

i i

i i
 i i

“CP06” — 2020/11/7 — 12:08 — page 118 — #4

i i

118 Introduction to Internet of Things

6.3 Processing Topologies

The identification and intelligent selection of processing requirement of an IoT
application are one of the crucial steps in deciding the architecture of the deployment.
A properly designed IoT architecture would result in massive savings in network
bandwidth and conserve significant amounts of overall energy in the architecture
while providing the proper and allowable processing latencies for the solutions
associated with the architecture. Regarding the importance of processing in IoT as
outlined in Section 6.2, we can divide the various processing solutions into two large
topologies: 1) On-site and 2) Off-site. The off-site processing topology can be further
divided into the following: 1) Remote processing and 2) Collaborative processing.

6.3.1 On-site processing

As evident from the name, the on-site processing topology signifies that the data is
processed at the source itself. This is crucial in applications that have a very low
tolerance for latencies. These latencies may result from the processing hardware or
the network (during transmission of the data for processing away from the processor).
Applications such as those associated with healthcare and flight control systems (real-
time systems) have a breakneck data generation rate. These additionally show rapid
temporal changes that can be missed (leading to catastrophic damages) unless the
processing infrastructure is fast and robust enough to handle such data. Figure 6.2
shows the on-site processing topology, where an event (here, fire) is detected utilizing
a temperature sensor connected to a sensor node. The sensor node processes the
information from the sensed event and generates an alert. The node additionally has
the option of forwarding the data to a remote infrastructure for further analysis and
storage.

Environment Sensing Processing

On-site
processing

Event: Fire Temperature Sensor node

sensor

Figure 6.2 Event detection using an on-site processing topology

i i

i i
 i i

“CP06” — 2020/11/7 — 12:08 — page 119 — #5

i i

IoT Processing Topologies and Types 119

6.3.2 Off-site processing

The off-site processing paradigm, as opposed to the on-site processing paradigms,
allows for latencies (due to processing or network latencies); it is significantly cheaper
than on-site processing topologies. This difference in cost is mainly due to the low
demands and requirements of processing at the source itself. Often, the sensor
nodes are not required to process data on an urgent basis, so having a dedicated
and expensive on-site processing infrastructure is not sustainable for large-scale
deployments typical of IoT deployments. In the off-site processing topology, the
sensor node is responsible for the collection and framing of data that is eventually
to be transmitted to another location for processing. Unlike the on-site processing
topology, the off-site topology has a few dedicated high-processing enabled devices,
which can be borrowed by multiple simpler sensor nodes to accomplish their tasks. At
the same time, this arrangement keeps the costs of large-scale deployments extremely
manageable [5]. In the off-site topology, the data from these sensor nodes (data
generating sources) is transmitted either to a remote location (which can either be a
server or a cloud) or to multiple processing nodes. Multiple nodes can come together
to share their processing power in order to collaboratively process the data (which
is important in case a feasible communication pathway or connection to a remote
location cannot be established by a single node).

Remote processing
This is one of the most common processing topologies prevalent in present-day IoT
solutions. It encompasses sensing of data by various sensor nodes; the data is then
forwarded to a remote server or a cloud-based infrastructure for further processing
and analytics. The processing of data from hundreds and thousands of sensor nodes
can be simultaneously offloaded to a single, powerful computing platform; this results
in massive cost and energy savings by enabling the reuse and reallocation of the
same processing resource while also enabling the deployment of smaller and simpler
processing nodes at the site of deployment [4]. This setup also ensures massive
scalability of solutions, without significantly affecting the cost of the deployment.
Figure 6.3 shows the outline of one such paradigm, where the sensing of an event is
performed locally, and the decision making is outsourced to a remote processor (here,
cloud). However, this paradigm tends to use up a lot of network bandwidth and relies
heavily on the presence of network connectivity between the sensor nodes and the
remote processing infrastructure.

Collaborative processing
This processing topology typically finds use in scenarios with limited or no network
connectivity, especially systems lacking a backbone network. Additionally, this
topology can be quite economical for large-scale deployments spread over vast areas,
where providing networked access to a remote infrastructure is not viable. In such
scenarios, the simplest solution is to club together the processing power of nearby

i i

i i
 i i

“CP06” — 2020/11/7 — 12:08 — page 120 — #6

i i

120 Introduction to Internet of Things

Environment Sensing Network Remote

processing

Event: Fire Temperature Sensor node Internet

sensor

Figure 6.3 Event detection using an off-site remote processing topology

processing nodes and collaboratively process the data in the vicinity of the data
source itself. This approach also reduces latencies due to the transfer of data over
the network. Additionally, it conserves bandwidth of the network, especially ones
connecting to the Internet. Figure 6.4 shows the collaborative processing topology
for collaboratively processing data locally. This topology can be quite beneficial for
applications such as agriculture, where an intense and temporally high frequency
of data processing is not required as agricultural data is generally logged after
significantly long intervals (in the range of hours). One important point to mention
about this topology is the preference of mesh networks for easy implementation of
this topology.

Environment Sensing

Collaborative
network/mesh

Event: Fire Temperature Sensor

sensor node

Figure 6.4 Event detection using a collaborative processing topology

6.4 IoT Device Design and Selection Considerations

The main consideration of minutely defining an IoT solution is the selection of the
processor for developing the sensing solution (i.e., the sensor node). This selection is
governed by many parameters that affect the usability, design, and affordability of the

i i

i i
 i i

“CP06” — 2020/11/7 — 12:08 — page 121 — #7

i i

IoT Processing Topologies and Types 121

designed IoT sensing and processing solution. In this chapter, we mainly focus on the
deciding factors for selecting a processor for the design of a sensor node. The main
factor governing the IoT device design and selection for various applications is the
processor. However, the other important considerations are as follows.

• Size: This is one of the crucial factors for deciding the form factor and the
energy consumption of a sensor node. It has been observed that larger the form
factor, larger is the energy consumption of the hardware. Additionally, large form
factors are not suitable for a significant bulk of IoT applications, which rely on
minimal form factor solutions (e.g., wearables).
• Energy: The energy requirements of a processor is the most important
deciding factor in designing IoT-based sensing solutions. Higher the energy
requirements, higher is the energy source (battery) replacement frequency. This
principle automatically lowers the long-term sustainability of sensing hardware,
especially for IoT-based applications.
• Cost: The cost of a processor, besides the cost of sensors, is the driving force
in deciding the density of deployment of sensor nodes for IoT-based solutions.
Cheaper cost of the hardware enables a much higher density of hardware
deployment by users of an IoT solution. For example, cheaper gas and fire
detection solutions would enable users to include much more sensing hardware
for a lesser cost.
• Memory: The memory requirements (both volatile and non-volatile memory) of
IoT devices determines the capabilities the device can be armed with. Features
such as local data processing, data storage, data filtering, data formatting, and
a host of other features rely heavily on the memory capabilities of devices.
However, devices with higher memory tend to be costlier for obvious reasons.
• Processing power: As covered in earlier sections, processing power is vital
(comparable to memory) in deciding what type of sensors can be accommodated
with the IoT device/node, and what processing features can integrate on-site
with the IoT device. The processing power also decides the type of applications
the device can be associated with. Typically, applications that handle video and
image data require IoT devices with higher processing power as compared to
applications requiring simple sensing of the environment.
• I/O rating: The input–output (I/O) rating of IoT device, primarily the processor,
is the deciding factor in determining the circuit complexity, energy usage, and
requirements for support of various sensing solutions and sensor types. Newer
processors have a meager I/O voltage rating of 3.3 V, as compared to 5 V for the
somewhat older processors. This translates to requiring additional voltage and
logic conversion circuitry to interface legacy technologies and sensors with the
newer processors. Despite low power consumption due to reduced I/O voltage
levels, this additional voltage and circuitry not only affects the complexity of the
circuits but also affects the costs.

i i

i i
 i i

“CP06” — 2020/11/7 — 12:08 — page 122 — #8

i i

122 Introduction to Internet of Things

• Add-ons: The support of various add-ons a processor or for that matter, an IoT
device provides, such as analog to digital conversion (ADC) units, in-built clock
circuits, connections to USB and ethernet, inbuilt wireless access capabilities, and
others helps in defining the robustness and usability of a processor or IoT device
in various application scenarios. Additionally, the provision for these add-ons
also decides how fast a solution can be developed, especially the hardware part
of the whole IoT application. As interfacing and integration of systems at the
circuit level can be daunting to the uninitiated, the prior presence of these options
with the processor makes the processor or device highly lucrative to the users/
developers.

Check yourself

RISC versus CISC processors, volative versus non-volatile memory

6.5 Processing Offloading

The processing offloading paradigm is important for the development of densely
deployable, energy-conserving, miniaturized, and cheap IoT-based solutions for
sensing tasks. Building upon the basics of the off-site processing topology covered
in the previous sections in this chapter, we delve a bit further into the various nuances
of processing offloading in IoT.
Figure 6.5 shows the typical outline of an IoT deployment with the various layers of
processing that are encountered spanning vastly different application domains—from
as near as sensing the environment to as far as cloud-based infrastructure. Starting
from the primary layer of sensing, we can have multiple sensing types tasked with
detecting an environment (fire, surveillance, and others). The sensors enabling these
sensing types are integrated with a processor using wired or wireless connections
(mostly, wired). In the event that certain applications require immediate processing of
the sensed data, an on-site processing topology is followed, similar to the one in Figure
6.2. However, for the majority of IoT applications, the bulk of the processing is carried
out remotely in order to keep the on-site devices simple, small, and economical.
Typically, for off-site processing, data from the sensing layer can be forwarded to
the fog or cloud or can be contained within the edge layer [6]. The edge layer makes
use of devices within the local network to process data that which is similar to the
collaborative processing topology shown in Figure 6.4. The devices within the local
network, till the fog, generally communicate using short-range wireless connections.
In case the data needs to be sent further up the chain to the cloud, long-range wireless
connection enabling access to a backbone network is essential. Fog-based processing is
still considered local because the fog nodes are typically localized within a geographic
area and serve the IoT nodes within a much smaller coverage area as compared to the

i i

i i


IoT Processing Topologies and Types 123

cloud. Fog nodes, which are at the level of gateways, may or may not be accessed by
the IoT devices through the Internet.

On-site Local Local Global

Environment

Temperature
sensor Database
Server

Cloud processing
Local processing

Edge processing

Fog processing
Event: Fire
Local
Sensing

network
clusters
Camera
Environment sensor Sensor Internet Server
node

Event: Surveillance
Temperature Server
sensor Database

Wired/
Communication

wireless Short
range
wireless Long range wireless/backbone

Backbone

Long range wireless/backbone

Long range wireless/backbone

Figure 6.5 The various data generating and storage sources connected to the Internet and the
plethora of data types contained within it

Finally, the approach of forwarding data to a cloud or a remote server, as shown

in the topology in Figure 6.3, requires the devices to be connected to the Internet
through long-range wireless/wired networks, which eventually connect to a backbone
network. This approach is generally costly concerning network bandwidth, latency, as
well as the complexity of the devices and the network infrastructure involved.
This section on data offloading is divided into three parts: 1) offload location
(which outlines where all the processing can be offloaded in the IoT architecture), 2)
offload decision making (how to choose where to offload the processing to and by how
much), and finally 3) offloading considerations (deciding when to offload).

6.5.1 Offload location

The choice of offload location decides the applicability, cost, and sustainability of the
IoT application and deployment. We distinguish the offload location into four types:


 i i

“CP06” — 2020/11/7 — 12:08 — page 124 — #10

i i

124 Introduction to Internet of Things

• Edge: Offloading processing to the edge implies that the data processing is
facilitated to a location at or near the source of data generation itself. Offloading
to the edge is done to achieve aggregation, manipulation, bandwidth reduction,
and other data operations directly on an IoT device [7].
• Fog: Fog computing is a decentralized computing infrastructure that is utilized
to conserve network bandwidth, reduce latencies, restrict the amount of data
unnecessarily flowing through the Internet, and enable rapid mobility support
for IoT devices. The data, computing, storage and applications are shifted to a
place between the data source and the cloud resulting in significantly reduced
latencies and network bandwidth usage [8].
• Remote Server: A simple remote server with good processing power may
be used with IoT-based applications to offload the processing from resource-
constrained IoT devices. Rapid scalability may be an issue with remote servers,
and they may be costlier and hard to maintain in comparison to solutions such
as the cloud [4].
• Cloud: Cloud computing is a configurable computer system, which can get
access to configurable resources, platforms, and high-level services through a
shared pool hosted remotely. A cloud is provisioned for processing offloading
so that processing resources can be rapidly provisioned with minimal effort over
the Internet, which can be accessed globally. Cloud enables massive scalability of
solutions as they can enable resource enhancement allocated to a user or solution
in an on-demand manner, without the user having to go through the pains of
acquiring and configuring new and costly hardware [9].

6.5.2 Offload decision making

The choice of where to offload and how much to offload is one of the major deciding
factors in the deployment of an offsite-processing topology-based IoT deployment
architecture. The decision making is generally addressed considering data generation
rate, network bandwidth, the criticality of applications, processing resource available
at the offload site, and other factors. Some of these approaches are as follows.
• Naive Approach: This approach is typically a hard approach, without too much
decision making. It can be considered as a rule-based approach in which the data
from IoT devices are offloaded to the nearest location based on the achievement
of certain offload criteria. Although easy to implement, this approach is never
recommended, especially for dense deployments, or deployments where the
data generation rate is high or the data being offloaded in complex to handle
(multimedia or hybrid data types). Generally, statistical measures are consulted
for generating the rules for offload decision making.
• Bargaining based approach: This approach, although a bit processing-intensive
during the decision making stages, enables the alleviation of network traffic
congestion, enhances service QoS (quality of service) parameters such as

i i

i i
 i i

“CP06” — 2020/11/7 — 12:08 — page 125 — #11

i i

IoT Processing Topologies and Types 125

bandwidth, latencies, and others. At times, while trying to maximize multiple

parameters for the whole IoT implementation, in order to provide the most
optimal solution or QoS, not all parameters can be treated with equal importance.
Bargaining based solutions try to maximize the QoS by trying to reach a point
where the qualities of certain parameters are reduced, while the others are
enhanced. This measure is undertaken so that the achieved QoS is collaboratively
better for the full implementation rather than a select few devices enjoying very
high QoS. Game theory is a common example of the bargaining based approach.
This approach does not need to depend on historical data for decision making
purposes.
• Learning based approach: Unlike the bargaining based approaches, the learning
based approaches generally rely on past behavior and trends of data flow
through the IoT architecture. The optimization of QoS parameters is pursued by
learning from historical trends and trying to optimize previous solutions further
and enhance the collective behavior of the IoT implementation. The memory
requirements and processing requirements are high during the decision making
stages. The most common example of a learning based approach is machine
learning.

6.5.3 Offloading considerations

There are a few offloading parameters which need to be considered while deciding
upon the offloading type to choose. These considerations typically arise from the
nature of the IoT application and the hardware being used to interact with the
application. Some of these parameters are as follows.

• Bandwidth: The maximum amount of data that can be simultaneously

transmitted over the network between two points is the bandwidth of that
network. The bandwidth of a wired or wireless network is also considered to
be its data-carrying capacity and often used to describe the data rate of that
network.
• Latency: It is the time delay incurred between the start and completion of an
operation. In the present context, latency can be due to the network (network
latency) or the processor (processing latency). In either case, latency arises due
to the physical limitations of the infrastructure, which is associated with an
operation. The operation can be data transfer over a network or processing of
a data at a processor.
• Criticality: It defines the importance of a task being pursued by an IoT
application. The more critical a task is, the lesser latency is expected from the
IoT solution. For example, detection of fires using an IoT solution has higher
criticality than detection of agricultural field parameters. The former requires a
response time in the tune of milliseconds, whereas the latter can be addressed
within hours or even days.

i i

i i
 i i

“CP06” — 2020/11/7 — 12:08 — page 126 — #12

i i

126 Introduction to Internet of Things

• Resources: It signifies the actual capabilities of an offload location. These

capabilities may be the processing power, the suite of analytical algorithms, and
others. For example, it is futile and wasteful to allocate processing resources
reserved for real-time multimedia processing (which are highly energy-intensive
and can process and analyze huge volumes of data in a short duration) to scalar
data (which can be addressed using nominal resources without wasting much
energy).
• Data volume: The amount of data generated by a source or sources that can be
simultaneously handled by the offload location is referred to as its data volume
handling capacity. Typically, for large and dense IoT deployments, the offload
location should be robust enough to address the processing issues related to
massive data volumes.

Summary
This chapter started with an overview of the various data formats available on
the Internet and to which various IoT solutions are exposed. The complexities in
handling the numerous data formats available present a significant challenge to the
design of IoT-based solutions. In order to address these challenges, the importance
of processing in IoT is discussed. This discussion is followed by an introduction
to various processing topologies, which can be chosen to address the challenges of
IoT processing. These topologies are broadly made up of two categories: 1) On-site
processing and 2) Off-site processing. The off-site processing is typically composed
of approaches to offload data to locations which are not the same as the one from
which the data was generated. A discussion on processing offloading follows these
topologies. Various offload location types, means of deciding offload location and
quantity are explained. Finally, the various parameters to be considered for offloading
are discussed to enable the reader to grasp the nuances of processing in IoT.

Exercises
(i) What are the different data formats found in IoT network traffic streams?
(ii) Depending on the urgency of data processing, how are IoT data classified?
(iii) Highlight the pros and cons of on-site and off-site processing.
(iv) Differentiate between structured and unstructured data.
(v) How is collaborative processing different from remote processing?
(vi) What are the critical factors to be considered during the design of IoT devices?
(vii) What are the typical data offload locations available in the context of IoT?
(viii) What are the various decision making approaches chosen for offloading data in
IoT?

i i

i i
 i i

“CP06” — 2020/11/7 — 12:08 — page 127 — #13

i i

IoT Processing Topologies and Types 127

(ix) What factors are to be considered while deciding on the data offload location?

References
[1] Misra, S., A. Mukherjee, and A. Roy. 2018. “Knowledge Discovery for Enabling Smart
Internet of Things: A Survey.” Wiley Interdisciplinary Reviews: Data Mining and Knowledge
Discovery 8(6): 1276.
[2] Jiang, L., L. Da Xu, H. Cai, Z. Jiang, F. Bu, and B. Xu. 2014. “An IoT-oriented Data Storage
Framework in Cloud Computing Platform.” IEEE Transactions on Industrial Informatics
10(2): 1443–1451.
[3] Mukherjee, A., S. Misra, V. S. P. Chandra, and M. S. Obaidat. 2018. “Resource-Optimized
Multi-Armed Bandit Based Offload Path Selection in Edge UAV Swarms.” IEEE Internet
of Things Journal 6(3): 4889–4896.
[4] Mukherjee, A., S. Misra, N. S. Raghuwanshi, and S. Mitra. 2018. “Blind Entity Identification
for Agricultural IoT Deployments.” IEEE Internet of Things Journal 6(2): 3156–3163.
[5] Mukherjee, A., N. Pathak, S. Misra, and S. Mitra. 2018. “Predictive Intra-Edge Packet-
Source Mapping in Agricultural Internet of Things.” In 2018 IEEE Globecom Workshops
(GC Wkshps) (pp. 1–6). IEEE. doi: 10.1109/GLOCOMW.2018.8644296
[6] Cheng, N., F. Lyu, W. Quan, C. Zhou, H. He, W. Shi, and X. Shen. 2019. “Space/Aerial-
Assisted Computing Offloading for IoT Applications: A Learning-Based Approach.”
IEEE Journal on Selected Areas in Communications 37(5): 1117–1129.
[7] Huang, L., X. Feng, C. Zhang, L. Qian, and Y. Wu. 2019. “Deep Reinforcement Learning-
based Joint Task Offloading and Bandwidth Allocation for Multi-user Mobile Edge
Computing.” Digital Communications and Networks 5(1): 10–17.
[8] Adhikari, M., M. Mukherjee, and S. N. Srirama. 2019. “DPTO: A Deadline and Priority-
aware Task Offloading in Fog Computing Framework Leveraging Multi-level Feedback
Queueing.” IEEE Internet of Things Journal. doi: 10.1109/JIOT.2019.2946426.
[9] Mahmoodi, S. E., K. Subbalakshmi, and R. N. Uma. 2019. “Classification of Mobile Cloud
Offloading.” In Spectrum-Aware Mobile Computing (pp. 7–11). Springer, Cham.

i i

i i