Unit – 2 NETWORK AND COMMUNICATION ASPECTS

Wireless Medium Access Issues

When it comes to communication using a wireless medium there is always a

concern about the interference due to other present wireless communication
technologies. Wireless means communication and message transfer
without the use of physical medium i.e., wires.

Let us understand how communication is done between them. Different Mobile

stations (MS) are attached to a transmitter/receiver which communicates via a
shared channel by other nodes. In this type of communication, it makes it
difficult for the MAC design rather than the wireline networks.

The very important issues which are observed are:

• Half Duplex operation

• Time-varying channel
• Burst channel errors.

These are explained as following below.

a) Half Duplex operation:

Half-duplex transmission means when the sender and receiver both are
capable of sharing data but one at a time. In wireless transmission, it is
difficult to receive data when the transmitter is sending the data because
during transmission a large amount or a large fraction of signal energy is
leaked while broadcasting. The magnitude of the transferred signal and
received signal differs a lot. Due to which collision detection is even not
possible by the sender as the intensity of the transferred signal is large than
the received one. Hence this causes the problem of collision and the prime
focus should be to minimize the collision.

b) Time-varying channel :
Time-varying channels include the three mechanisms for radio signal
propagations they are Reflection, Diffraction, and Scattering.

Page 1 of 12
• Reflection –
This occurs when propagating wave carrying information intrudes on an object
that has very large dimensions than the wavelength of the wave.

• Diffraction –
This occurs when the radio path between the transmitter and the receiver is
collided by the surface with sharp edges. This is a phenomenon which causes
the diffraction of the wave from the targeted position.

• Scattering –
This occurs when the medium through from the wave is traveling consists of
some objects which have dimensions smaller than the wavelength of the wave.

While transmitting the signal by the node these are time shifted and this is
called multipath propagation. While when this node signals intensity is dropped
below a threshold value, then this is termed as fade. As a result Handshaking
strategy is widely used so as a healthy communication can be set up.

c) Burst channel errors:

Burst channel errors are called as a contiguous sequence of symbols, which
are received in a communication channel, in which the first and last symbols
has an error and there is no evidence of contiguous sub-sequence of corrected
received symbols. When time-varying channels are used then signals
strengths are introduced due to which errors are observed in transmission. For
these channels in wireline networks, the Bit rate is high as 10 -3.

MAC protocol survey:

MAC protocol is the first protocol layer above the Physical Layer in ad hoc.
The primary task of any MAC protocol is to control the access of the nodes to
shared medium.

Classification of MAC protocols:

These are as following below.

1. Contention-based protocols without reservation/scheduling –

• Bandwidths are not reserved.
• No guarantees.
2. Contention-based protocols with reservation mechanisms –
• Bandwidth is reserved for transmission.
• Guarantees can be given.

Page 2 of 12
3. Contention-based protocols with scheduling mechanisms –
• Distributed scheduling is done between nodes.
• Guarantees can be given.
4. Other protocols –
• Combine multiple features of other protocols.
 It can also use a completely new approach.

Figure 2.1: MAC Protocol survey

Survey Routing protocols

A routing protocol specifies how routers communicate with each other to

distribute information that enables them to select routes between nodes on a
computer network. There exist lots of routing protocols each having a unique
operating standard with significant performance for Wireless Sensor Networks
that can be deployed for IoT with few modifications for bandwidth and power
consumption. We discuss few of the broad categories of routing protocols in
this section.

Page 3 of 12
• Naive Routing
The idea deployed in naïve routing is flooding. Each node can overhear its
neighbors within its range. The source node floods the network with route
request packets called as beacons. Destination nodes respond with a route reply
message to the beacon and communication link is established between these
nodes. Beaconing is typically utilized for location tracking, discovering routes
to destinations and tracking neighbors through keep-alive requests. One of the
most important factors that affect performance is the beacon interval in the route
discovery process. If the beacon interval is too small, the number of beacons
generated becomes huge. On the other hand, a higher beacon interval incurs a
lesser number of generated beacons. Popular routing protocols such as DSR,
DSDV and AODV fall under this category.

• Hierarchical Routing
Nodes form clusters based on polling. The cluster head is responsible for all
communications on behalf of the members of the cluster. Group mobility can be
achieved by the cluster head following some metric to devise the mobility
pattern of the nodes in the cluster. LEACH is a common example, where the
cluster head is rotated among the members to facilitate load balancing that can
be deployed for IoT environments.

• Query based Routing

The underlying principle of query based routing is data dissemination within the
network. A querying node can retrieve data from any node in the network.
Common examples are SPIN and Direct Diffusion.

• Multipath routing
Protocols employing multipath routing seek to and use alternate paths towards
every destination. This distributes the cost of forwarding packets among more
nodes, saving the energy of individual, highly-frequented nodes.

• Probabilistic routing
Routing decision is based on the calculated probabilistic value. A primitive
method to compute these values is by gossiping. Data packets are flooded into
the network like a rumor with a probability p. Unlike other flooding
mechanisms, these packets are forwarded only once and thereby the traffic
overhead is reduced. A highly structured approach is to refer the prior history of
packet delivery and mobility pattern, based on this we can decide which nodes
can form a route to the destination.

Page 4 of 12
 Ad-hoc On-Demand Distance Vector (AODV)
AODV computes a loop free single path on demand. A mobile node discovers
and maintains a route to another node only when it needs to communicate. One
observation of AODV is that, though the source actually discovers multiple
paths during the route discovery process, it chooses only the best route and
discards the rest. Also, frequent route breaks cause the intermediate nodes to
drop packets because no alternate path to the destination is available. This
reduces the overall throughput and the packet delivery ratio. Moreover, in high
mobility scenarios, the average end-to-end delay can be significantly high due
to frequent route discoveries. When route failures occur, the process of route
discovery has to begin from scratch consuming more network resources and
overhead.

Sensor Deployment & Node Discovery

Five static sensor deployment strategies are introduced. The deployment

strategies are one uniform random deployment and four regular deployments:
square grid, triangle grid, hexagon grid, and THT. These strategies are shown in
Figure 2.2. A square field is considered in this work.

Figure 2.2 Sensor Deployment Strategies

Page 5 of 12
• Uniform Random Deployment

• Regular Deployment
1. Square Grid
2. Triangle Grid
3. Hexagon Grid
4. Tri-Hexagon Tiling (THT)

a) Uniform Random Deployment:

In the random deployment scheme, sensors are deployed in a random way.
Therefore, the exact locations of nodes are not known. Random deployment is
usually suitable in the following cases:

• When the environment is harsh and therefore, deployment of nodes is

exceedingly difficult.
• When wireless sensor network is large scale.
• When the cost of sensors is not an issue

In this method, sensors are dropped form a helicopter, an airplane or an

unmanned vehicle. Uniform random deployment is random deployment where
nodes are placed in a sensing area randomly with uniform distribution. Figure
2.2 (a) illustrates the uniform random deployment method.

Regular Deployment:
Regular deployment refers to placing sensors in a regular form in a sensing area.
This type of deployment can usually be applied in non-harsh environment and
non-large-scale regions. In this paper, four regular deployment strategies are
introduced.

1) Square Grid: In the square grid deployment, a square sensing area is

divided into small squares and the nodes are placed at the points of grid
intersection as shown in Figure 2.2(b).
2) Triangle Grid: A square sensing area in this model is divided into small
triangles and the sensors are located in the points of triangle heads as
shown in Figure 2.2(c).
3) Hexagon Grid: In hexagon grid deployment scheme, a sensing area is
divided into hexagons, as illustrated in Figure 2.2(d) and the sensors are
placed at the vertices of hexagons.
4) Tri-Hexagon Tiling (THT): In this model, is a pattern of hexagonal stars
that consist of triangles and hexagons. Tiling refers to covering the area
without leaving any gaps and without any overlapping. Figure 2.2(e)
displays the THT deployment strategy.

Page 6 of 12
IoT Sensor Deployment Challenges:

The business benefits of IoT coupled with market trends are driving rapid IoT
adoption in every industry vertical like smart cities, building automation,
industrial, healthcare, etc. This growing demand for IoT connectivity is paving
the way to a plethora of sensor types for various use cases such as traffic
sensors, parking meters, pressure sensors, electricity sensors, and so on.
Efficient sensor deployment is one of the key success factors in every IoT
investment and that’s where most enterprises struggle a lot today.

Challenge #1: Variety of sensors and chipsets

There is an increasing number of commercial launches of cellular technologies
like NB-IoT, Cat-M1/M2, LTEM, LoRa, etc. Each of these technologies has
specific electronics for sensing endpoints. Although the cost of mobile chipsets
has been declining over time, currently there’s no cost-effective solution that
can work with the widespread in electronics of the cellular-connected IoT
sensors to measure connectivity parameters.

Challenge #2:Identify an optimal location to deploy sensors

Whether it is a factory floor or a smart building, it is never easy to identify the
perfect spot to deploy the IoT sensors. To successfully capture and transmit the
ambient inputs over-the-air, the sensor must be located near the input-source
and also where the network signal strength is reliable. To determine signal
quality spread, today operators mostly rely on statistical modeling using terrain
and clutter models. This results in statistical variabilities. Currently, there is no
way to capture empirical network data, the lack of which often leads to
installing sensors in sub optimal locations where the signal quality is poor and
unreliable. Unreliable connectivity results in poor sensor performance which in
turn affects the performance of the overall IoT solution and impacts the
customer experience.

Challenge # 3: Not easy to remediate sensor performance issues

IoT sensors are typically installed in hard-to-access locations. When a sensor
exhibits sub-optimal performance due to network connectivity, it is hard to root-
cause the problem. Currently, there is no way to obtain real-time visibility into
connectivity data to assess network health. The technicians may need to try a
different location hoping for better wireless connectivity or replace the sensor
itself. In such trial-and-error methodology, multiple truck rolls could be needed
before the problem is identified and rectified. This impacts both OpEx and TCO
and also skews up inventory management.

Page 7 of 12
Challenge # 4 Network validations for connectivity SLAs
It is difficult to guarantee a satisfactory level of service if the IoT devices fail to
deliver due to poor connectivity. Currently, RF and RAN design are done based
on statistical models. Once the IoT network is deployed, due to the lack of
network health data, it is not possible to validate your network design
assumptions and performance in the context of the initial SLAs.
In a highly competitive digital marketplace, businesses can’t live with these
challenges for too long.

NEIGHBOUR DISCOVERY FOR loT SCENARIOS:

1. Time Synchronised Protocols

2. Deterministic Approach
3. Colocation based Approaches
4. A Fully Distributed Opportunistic Approach
5. Learning Based Approach

Time Synchronised Protocols:

These protocols make use of synchronization methods to schedule
communication with neighboring nodes. With these methods it is possible for
the nodes to arrive at a common time reference.

Deterministic Approach:
Deterministic approaches make use of mathematical methods to guarantee
overlap. Though they require higher discovery latency, they do not need any
synchronization. The searchlight protocol by Bakht et al can ensure discovery
within a definite latency by sequential search. This protocol has two slots -
Active slots and probe slots. The nodes decide to remain awake during active
slots, and sleep during the remaining slots. During an active slot, the node may
send/receive or do both. Successful discovery takes place between two
neighboring nodes whenever their active slots overlap. When a discovery
scheme uses few active slots to discover neighbors within a reasonable time
limit, then it is said to be efficient. In searchlight, each node has two active slots
in t. The first active slot, called the anchor slot, is the first slot in the period. In
the symmetric case, since the position of this anchor slot is fixed in t but the
start times for the t vary for different nodes, the anchor slots for two nodes
overlap only if the difference between the start times of the two periods is less
than a timeslot. The two anchor slots would never meet since the offset remains
constant. The relative position of the anchor slot of one node remains the same
with respect to that of the other node always and is in the range [1, t - 1].
Searchlight introduces a second active slot in t called the probe slot, which

Page 8 of 12
searches for the anchor slot of the other node. Probe-anchor, probe-probe and
anchor-anchor overlaps all result in discovery.
One disadvantage of this scheme is that it is a mobile agnostic method and is
not well suited for mobile scenarios.

Colocation based Approaches:

These approaches make use of knowledge coming from neighbouring nodes.
This knowledge is used to coordinate schedule between the nodes.

A Fully Distributed Opportunistic Approach:

In order to optimize the discovery of constrained nodes in loT network, Djamaa
et al have proposed EADP (Efficient Application-layer Discovery Protocol).
The protocol targets low power applications. EADP architecture has the
following components - User agent, Service agent, Reply Agent. The user agent
delivers a limited-flooding algorithm for discovering nodes in a pull-mode. The
service agent is responsible for maintenance of remote service information. The
service agent also has a state-maintenance mechanism that allows the protocol
to react to topology changes. The reply agent is responsible for control and
delivery of unicast replies. In addition, the Service Agent uses Trickle-inspired
Algorithm which controls the frequency and size of messages that are used for
advertising. Thus, the algorithm allows minimizing the number of advertising
messages and also ensures a bound on the number of descriptions of the service.

Learning Based Approach:

Context Aware Resource Discovery framework leverages Q-Learning
techniques to do node discovery. The system is able to reduce energy wastage
when contacts are not expected while reducing discovery latency by employing
an approach in which both mobile and static devices can learn their contact
opportunities. Thus the time of contact useful for service is increased
provisioning data collection.

Q-Learning is becoming widely used in communication scenarios for solving

problems that require learning of knowledge to be exploited at a later time. This
algorithm belongs to the Temporal Difference (TD) methods of Reinforcement
Learning (RL) methods. The other two methods of RL are Dynamic
Programming and Monte Carlo methods. TD methods combine the pros of the
other two types of RL methods; that is, they don't require an accurate model of
the environment.

Page 9 of 12
 Figure 2.3: Learning based approach

After the completion of the learning phase, the framework enters the
experimental phase. In this phase, the framework is tested with the real world
data and optimal decisions are taken.

Data Aggregation:
Data Aggregation is the process of one or more sensor nodes and detects the
information result from the other sensor nodes. The aim of the data aggregation
is removes data redundancy and improves the energy lifetime in wireless sensor
network. Therefore reducing the number of data packets transmitted over the
network because aggregation need less power as compare to multiple packets
sending having same data.

Data Aggregation Approaches:

Data aggregation is further divided into four basic approaches.

a) Cluster based approach
b) Tree based approach
c) Multipath approach
d) Hybrid approach.

Cluster Based Approach: Cluster Based Approach is defined as the

hierarchical approach in which whole network is separated into various cluster.
Each cluster has cluster heads and which is cluster choose from members. The

Page 10 of 12
main role of cluster head aggregate data received from cluster members locally
and then transmits the result to base station. The cluster head can share
information with the sink directly via long range transmissions or multi
hopping using other cluster heads.

Tree Based Approach: The Tree Based Approach is actually defining

aggregation concept which is used to make aggregation tree. This tree define as
minimum spanning tree in which sink node act as root and source node act
as leaves. Data start flowing from leave nodes up to root nodes. The main
disadvantage of this approach is data packet loss at any level of tree which may
cause failure whole network.

Multipath Approach: This approach is used to overcome the drawbacks of tree

based approach. Accordingly to this approach each and every node could send
data packets over multiple paths using multiple neighbors in aggregation tree.
So a data packet sends from source to destination using multiple paths with the
help of intermediate nodes. The example of this approach like ring topology.
Overhead is the disadvantage of this approach.

Hybrid Approach: The Hybrid Approach is the mixture of cluster based

approach, multipath approach and tree based approach. This approach is mainly
used for adaptively for optimal performance of their data aggregation.

Data Dissemination
Data Dissemination is the process in which sensor nodes is collecting the data
and communicate to the base station or any other interested node. The source
node is generating the data and the information to be reported is known as
event. Those nodes which are interested in event and seek information are
known as sink. So in this whole process data are routed in sensor network. It is
two steps process; in first step interested nodes are broadcast to their neighbor
nodes in the network and in second step nodes after receiving the request nodes
sends requesting data.

Data Dissemination Approach:

There are many data dissemination methods or approaches which are following
as:

Flooding: If the destination node is not receive the data packet or specified
number of hops is not reached. Then each node broadcast the gathered data until
the packet is reached to their destination node. The main advantage of flooding
is not requires costly topology maintain or route discovery but it face several
problems like implosion, overlap and resource blindness.

Page 11 of 12
Gossiping: The gossiping is the version of flooding approach .In this approach
the packet is sent to a single neighbor chosen from neighbour table randomly
instead of broadcasting each packet to the entire neighbor. This process can take
long time from completion. Gossiping avoids the problem faced in flooding
approach like implosion.

SPIN: (Sensor Protocol for Information via Negotiation) this is the

enhancement of flooding protocols based on data centric routing. Flooding has
mainly three problems like: implosion, overlap and resource blindness. To
overcome these problems the spin family protocols used three ways: ADV,
REQ, DATA are used. The nodes which are interested in the event to transmit
REQ message for DATA. After receiving REQ message source node sends
DATA message to interested node. In this way data can reach to all interested
node in entire network. This technique prevents the problems implosion,
overlap and resource blindness which is faced by flooding. Wireless Sensor
Network is the important in the field networking

TEXT / REFERENCE BOOKS

1. Boswarthick,Omar Elloumi., The Internet of Things: Applications and
Protocols, Wiley publications., 2012
2. Dieter Uckelmann, Mark Harrison, Florian Michahelles., Architecting the
Internet of Things, Springer publications.2011
3. Marco Schwatrz Internet of Things with Arduino Cookbook, Packt
Publications.2016 .
4. Jan Holler, Vlasios Tsiatsis, Catherine Mulligan, Stefan Avesand, Stamatis
Karnouskos, David Boyle, “From Machine-to-Machine to the Internet of
Things: Introduction to a New Age of Intelligence”, 1st Edition, Academic
Press, 2014.
5. Vijay Madisetti, Arshdeep Bahga, “Internet of Things: A Hands-On
Approach” published by Vijay Madisetti 2014
.

Page 12 of 12