Seminar Report

Download as pdf or txt
Download as pdf or txt
You are on page 1of 47

VISVESVARAYA TECHNOLOGICAL UNIVERSITY

BELAGAVI - 590018

A Technical Seminar Report on


“Congestion Free Routing Mechanism for IoT-enabled Wireless
Sensor Networks for Smart Healthcare Applications”

Submitted in partial fulfillment of the requirements as per VTU curriculum of


BACHELOR OF ENGINEERING
IN
COMPUTER SCIENCE & ENGINEERING
By

ANILKUMAR B N 4AL17CS006

Under the Guidance of


Mrs. Vidya M
Assistant Professor

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING


ALVA’S INSTITUTE OF ENGINEERING AND TECHNOLOGY
MOODBIDRI-574225, KARNATAKA
2020– 2021
ALVA’S INSTITUTE OF ENGINEERING AND TECHNOLOGY
MIJAR, MOODBIDRI D.K. -574225
KARNATAKA

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING

CERTIFICATE

This is to certify that ANILKUMAR B N (4AL17CS006) has submitted Technical Seminar


Report on “Congestion Free Routing Mechanism for IoT-enabled Wireless Sensor
Networks for Smart Healthcare Applications” for the VII semester B.E. in Computer
Science & Engineering during the year 2020-21

Mr. Vidya M Mr. Harish Kunder Dr. Manjunath Kotari


Seminar Guide Seminar Coordinator Professor and Head
ACKNOWLEDGEMENT
The satisfaction and euphoria that accompany a successful completion of any task would be
incomplete without the mention of people who made it possible, success is the epitome of
hard work and perseverance, but steadfast of all is encouraging guidance.
So, with gratitude I acknowledge all those whose guidance and encouragement served
as beacon of light and crowned the effort with success.

I thank my Seminar Guide Mr. Vidya M, Assistant Professor in Department of


Computer Science & Engineering, who has been my source of inspiration. He has been
especially enthusiastic in giving his valuable guidance and critical reviews.

The selection of Technical Seminar Topic as well as the timely completion is mainly
due to the interest and persuasion of my Seminar Coordinator Mr. Harish Kunder,
Associate Professor, Department of Computer Science & Engineering. I will remember his
contribution for ever.

I sincerely thank, Dr. Manjunath Kotari, Professor and Head, Department of


Computer Science & Engineering who has been the constant driving force behind the
completion of the seminar.

I thank Principal Dr. Peter Fernandes, for his constant help and support throughout.

I am also indebted to Management of Alva’s Institute of Engineering and


Technology, Mijar, Moodbidri for providing an environment which helped me in
completing the seminar.

Also, I thank all the teaching and non-teaching staff of Department of Computer
Science & Engineering for the help rendered.

Finally I would like to thank my parents and friends whose encouragement and
support was invaluable
ANILKUMAR B N
4AL17CS006

i
ABSTRACT
Recently, the Internet of Things (IoT) topology has used to collect physical, physiological,
vital signs of patients in consumer-centric e-health or consumer’ wellness care services. In
such healthcare systems, varieties of medical sensors are attached to the patients to collect
vital signs from those who are under observation. The data gathering process in IoT-enabled
Wireless Sensor Network (WSN) suffers from the congestion problem. The effect of this
translates on missing packets, a decrease of reliability and throughput degradation in IoT-
enabled WSN. This paper proposes a distributed congestion control algorithm for IoT-
enabled WSNs to effectively resolve the congestion for healthcare applications. The
proposed scheme alleviates congestion by a priority-based data routing strategy.
Furthermore, this paper presents a priority queue based scheduling scheme for better
reliability. We analyze the properties of the proposed congestion control mechanism
mathematically and validate its performance through extensive simulation and real-life
experiments. The application of this work can be used to an early warning system in
detecting abnormal heart rate, blood pressure, ECG, EMG in the hospital/home care
environment to the state-of-art diagnosis.

ii
TABLE OF CONTENTS

DESCRIPTIONS PAGE NO.


CHAPTER NO.

ACKNOWLEDGEMENTS……………………………………………............ i

ABSTRACT……………………………………………………………………. ii

LIST OF FIGURES…………………………………………………………...... iii

LIST OF TABLES……………………………………………………………… iv

1. INTRODUCTION 1

1.1 STATEMENT OF THE PROBLEM 2

1.2 OBJECTIVES 2

2. RELATED WORK 4

3. DESIGN SCHEME 7

4. SYSTEM DESIGN 10
4.1 LOWER COMPUTER DESIGN 10
4.2 UPPER COMPUTER DESIGN 15

4.3 DATABASE DESIGN 15

4.4 DATA TRANSMISSION 18

5. FUNCTION TEST 20

6. APPLICATIONS 23

7. SYSTEM IMPROVEMENT PLAN 24

8. CONCLUSION 26
REFERENCES………………………………………………………………… 27

BASE PAPER …………………………………………………………………. 28


LIST OF FIGURES
Fig number Description Page number

2.1 Network Model 4

3.1 Percentage of successfully received packets 7

3.2 Source Data Rate 8

3.3 Distribution of average hop-by-hop delay according to 9


the sources data rate

4.1 Setup Phase 10


Comparison of Lifetime over traffic load 12
4.2
14
Traffic Load
4.3

5.1 Comparison of real testbed results 21

5.2 System Implementation and Testing 22

6.1 Comparison of real testbed lifetime over traffic load 25

iii
LIST OF TABLES
Table Description Page number
number

4.1 Energy Consumption of Nodes table 16

4.2 Parameters values used in the indoor tested table 17

4.3 Parameter setting of teacher information table 17

iv
CHAPTER 1

INTRODUCTION
RECENT advances in microelectronics fabrication and Very Large-Scale
Integrated (VLSI) circuit technologies allowed the development of smart electronics
devices with the capability of sensing, processing, and transmission [1]-[5]. These
electronic devices are combined to form a network that achieves some shared objectives.
They interact with the physical world pervasively with the aid of enhanced communication
protocols and distributed intelligence, which constitute a novel paradigm called the Internet
of Things (IoT). The IoT is possible through the integration of several heterogeneous
network infrastructures such as Radio Frequency Identification (RFID) systems, Machine-
to-Machine (M2M) communication systems, Wireless Sensor Networks (WSNs) and much
more. WSNs are used for tracking the status of the real-world object. All these different
network infrastructures together compose the backbone network infrastructure for IoT.
WSN has a lot of unique challenges and constraints such as self-management, energy
limitation, congested packet transmissions, ad hoc deployment, unattended operation, etc.
This unique infrastructure nature of WSN demands a protocol designed that should fit for
its specialties, especially for the consumer-centric e-health or consumer’ wellness care
services.

In the field of consumer electronics, several features in the IoT-based Healthcare


aware Wireless Sensor Networks (HWSNs) implementation are considered, such as (1)
Quality of Service (QoS) (2) low cost, (3) low power consumption, (4) easy integration,
and (5) low complexity [6], [7]. The primary challenge of HWSNs is congestion control
during physiological data routing that helps to provide a high QoS. Congestion control in
HWSNs is much more complicated compared to the congestion control in other traditional
networks. In HWSNs, different types of congestion may occur at different points and
locations of the networks. In a sensor network, deployed sensor nodes have limited
resources for computing, bandwidth, memory storage, energy supply, and communication
[8-10]. The event-driven character of the network channels to an unpredictable
communication load within the network. In healthcare applications, the network stocks low
traffic load when there are no special events but the happening of medical emergencies may
generate burst traffic which leads to congestion in the network [11], [12]. Therefore,
overloaded nodes exhaust more energy than other sensor nodes. Furthermore, they are
subject to premature death because of heavy traffic. Once the nodes with heavy load die,
some other sensor nodes should relay data to the gateway node originally forwarded by
Congestion Free Routing Mechanism for IoT-enabled Wireless Sensor Networks for
Smart Healthcare Applications
these dead nodes. It leads to dynamical changes in the routing paths, as well as creates

-enabled Wireless
congestion within the other sensor nodes. As a result, doctors and nurses cannot receive
Congestion Free Routing Mechanism for IoT
update information from patients under observation [13], [14]. Due to dynamical changes
Sensor Networks for Smart Healthcare
of routing paths, huge data loads can quickly generate congestion within the network.
Applications
Congestion creates considerable delays within the network. Therefore, data packets are
likely to be lost due to congestion and a high amount of energy is consumed in forwarding
packet toward the gateway node. A lot of research has been conducted to enhance the
performance of the HWSNs. A congestion avoidance scheme known as Relaxation
Theorywith Max-Min Fairness (RT-MMF) is presented to avoid congestion in the wireless
body sensor network [15]. A congestion avoidance scheme called Adaptive Duty-cycle
Based Congestion Control (ADCC) has proposed to handle congestion in home automation
networks which use both the resource control and traffic control mechanize to mitigate
packet loss [16]. However, existing congestion control mechanize does not perform well in
the health care application because they are unable to reduced transmission delay during
vital signs transmission from the monitoring patients to medical staff. Also, existing
approaches used huge message overhead to control burst traffic within the network that
significantly reduces network lifetime. Moreover, many approaches are already available
in the literature on congestion control routing for WSNs. However, the consumer-centric
IoT environment includes heterogeneous sensor devices. Hence, the traditional congestion
control approaches for WSNs must be enhanced to suit the WSNs in the consumer-centric
IoT environment. Thus, congestion control is one of the most critical problems in the
designing and implementation of WSNs, especially for the consumers-centric e-health or
consumer’ wellness care services where a vital event may occur any time and on arbitrary
nodes within the network.

To address these issues, this paper proposes a distributed traffic-aware congestion


control scheme for IoT-enabled WSNs. In the proposed scheme, deployed sensor nodes are
divided into different levels and transmit their sensing information through the congestion-
free path. The proposed algorithm minimizes the total energy consumption of the network
and improves the Quality of Services (QoS). For efficient and reliable communication in
IoT networks, we aim to frame a cross-layer design that spans different layers, i.e., stable
congestion-free routing in the network layer, effective access control and transmission
power control in MAC sub-layer, reliable and light-weight transmission control at the
transport layer, and a middle-ware technique to integrate the wireless personal area network
application subnet with the Internet backbone. In this paper, we propose a congestion-free

Dept. of CSE, AIET, Mijar Page 2


Congestion Free Routing Mechanism for IoT-enabled Wireless Sensor Networks for
Smart Healthcare Applications
data forwarding method at the network layer.

Congestion
The restFree Routing
of the paperMechanism -enabled Wireless
for IoT
is organized as follows. Section II introduces related works.
Sensor
Section Networks
III presents the network model for Smart
and problem Healthcare
formulation. Section IV gives an
analytical description of the proposed congestion control mechanism. Theoretical analysis
Applications
has presented in Section V. Section VI presents experimental results and analysis and
Section VII finally concludes this paper highlighting the main conclusions of the study.

1.1 PROBLEM STATEMENT


Congestion control is a primary objective in several multi-hop WSNs especially in
healthcare applications where deployed sensor nodes continually sense data from the
monitoring patients and forward this information to the Base Station (BS) to satisfy certain
requirements of the medical staff. This section provides an overview of the most recent
existing research works related to congestion control in WSNs..

1.2 OBJECTIVES
where a rate-based fairness aware congestion control was reported. This approach
divides all intermediate sensor nodes into near-source and near sink nodes. It detects
congestion according to the packet loss rate at the sink node. Kang et al. [8] proposed a
Topology Aware Resource Adaptation (TARA) scheme to alleviate congestion in WSNs.
This approach focuses on the adaptation of the extra resources to control congestion within
the network. This approach activates a particular sensor node whose mode is sleeping to
form a new topology

1.3 SCOPE FOR THE STUDY

• A new topology of the network can control the increased traffic of the network.
• It puts a large overhead for the large scale WSNs on the other hand.
• It builds a source-based tree to find all possible routes to the sink. HTAP selects the least
buffer node to forward the excess packets.
It is a data-centric and on-demand routing protocol where the routing phase selects based
on local communication.

Dept. of CSE, AIET, Mijar Page 3


Congestion Free Routing Mechanism for IoT-enabled Wireless Sensor Networks for
Smart Healthcare Applications

1.4 Congestion
RELATED WORK
Free Routing Mechanism for IoT-enabled Wireless
Sensor Networks for Smart Healthcare
Sharma et al. [18] presented a congestion control protocol for the WSNs called
Bidirectional Reliable and Congestion Control Transport Protocol (BRCCTP). In this
Applications
approach, a rate adjustment based scheme is used to manage congestion within the network.
It provides bidirectional reliability during data transmission. The proportion of average
packet service time over average packet inter-arrival time is used to detect the congestion
level in the nodes. In this approach, all data sources are assigned by an equal priority index
that helps to mitigate congestion. In this approach, if congestion is detected within the
network then it significantly increases data transmission latency.

Zhuang et al. [19] proposed a congestion control mechanize for data gathering in
sensor network called Congestion-Adaptive Data Collection scheme (CADC). In this
paper, an adaptive lossy compression based mechanism is developed to mitigate
congestion. Furthermore, a weighted CADC scheme is applied to manage cyber-physical
applications congestion. In CADC, data items are defined by the different priorities.
However, CADC does not focus on the energy consumption of the deployed sensor nodes.
Therefore, this approach suffers from a poor network lifetime.

It has been seen from the literature that the existing congestion control method leads
to a large number of message exchanges over the network for data and status exchange. It
increases data transmission delay within the network and energy consumption of the
deployed sensor nodes. Also, the quality of services (QoS) of the existing approaches is
very poor. Due to poor performance and high transmission delay of the existing set of
approaches, it is very challenging to implement in IoT-based healthcare applications. In
this paper, a distributed level based congestion control mechanism.

proposed for effectively classifying the data packets which help in congestion
avoidance. Furthermore, a level-based data routing scheme is used for making a more
suitable decision to implement an optimal value in the route identification process. The
existing congestion control routing protocols which were proposed for routing in WSNs
are not able to provide optimal energy consumption solution. Therefore, the proposed
model considered the sensors present in the IoT scenario. Hence, the proposed congestion
control data routing scheme provides optimal results by reducing the energy consumption
and delay. The proposed scheme is increasing the packet delivery ratio by avoiding
congestion in the network. This is achieved by the effective monitoring of different queues.

Dept. of CSE, AIET, Mijar Page 4


Congestion Free Routing Mechanism for IoT-enabled Wireless Sensor Networks for
Smart Healthcare Applications
It helps to avoid packet drops occurring due to congestion. Finally, it is noted that the

-enabled Wireless
proposed scheme gives optimal results in terms of QoS in IoT-based healthcare
Congestion Free Routing Mechanism for IoT
applications.
Sensor Networks for Smart Healthcare
We consider an IoT-based healthcare network consisting of N numbers of static
Applications
sensor nodes arbitrarily located in the hospital environment. In the initial network topology,
each node has transmission range Rmax and deployed network topology is represented with
an undirected weighted graph G= (V, E), where is the set of nodes and is the set of edges,
where shows the distance between nodes vi and vj. The data transmission links between
any two nodes do not have the same capacity and edge is expected to have a stipulated
capacity to transmit packets from node vi to node vj. Each deployed node periodically sends
the sensed data to the BS through multi-hop communication. The first order radio model is
used for energy consumption [20- 22]. So the energy consumed while transmitting a -bit
message to a receiver node located d distance apart can be formulated as the equation.

The authors say in this work proposed a matching game approach to deal with
accomplishing low energy utilization in IOT and fog condition. The authors made
utilization of cache technology and fog computing to enhance energy efficiency.

Several works have introduced some principles to implement the Fog, but few took
in consideration its energy consumption. Deng and al. [9] present a vision of the Fog-Cloud
interplay, while taking into account the trade-off between power consumption and latency.
The latency sensitive requests are processed locally on the Fog devices, whereas the others
are dispatched in the Cloud using wide area network (WAN).

The layer of architecture consists of edge, fog and cloud layers. At different layers,
processing and forwarding data by devices power are consumed differently. The problem
arises here how to identify which machine learning steps to be performed at which layer to
minimize the energy consumption.

The main objective of the proposed scheme is to control congestion in IoT-based


healthcare aware WSNs which can potentially increase the network’s lifetime. The
proposed scheme also considers the fairer delivery of sensing information to the gateway.
To achieve these objectives, the proposed scheme considers two main parameters energy
and delay. Furthermore, it considers two types of traffic: sensitive and non-sensitive.
Sensitive traffic is designed to transfer high priority data and non-sensitive traffic is
designed to transfer normal data.

Dept. of CSE, AIET, Mijar Page 5


Congestion Free Routing Mechanism for IoT-enabled Wireless Sensor Networks for
Smart Healthcare Applications
The proposed congestion control scheme consists of three phases: 1) setup phase,

-enabled Wireless
2) request distribution phase and 3) data routing and event occurrence report phase. In
Congestion Free Routing Mechanism for IoT
healthcare applications, due to different stages of vital signs and different priorities of
Sensor Networks for Smart Healthcare
several IoT based medical devices, the proposed scheme forwards sensed information to
Applications
the gateway node with different priorities. Before packets transmission, the source node
sets a priority according to the importance of the data. Intermediate nodes route these
packets according to the priority of the data packets.

The setup phase runs only once, during the initialization of the network. In this
phase, after deployment each node detects its single-hop neighbor nodes and deployed
nodes are also divided into different levels. For level detection, initially gateway sets its
level value once i.e. L(Gateway)=1 and sends a request message LEVEL to the sensor nodes
in the range Rmax.

sensor nodes also identify their single-hop child nodes and their current energy
condition. After level detection, all sensor nodes transmit their update level value, ID,
location information and current energy condition to gateway node through their
intermediate parent nodes. Due to mobility, sensor nodes may change their current
locations. If the sensor node moves from one level to another level, the mobile node updates
its current level value with the help of neighbor nodes.

Dept. of CSE, AIET, Mijar Page 6


CHAPTER 2

METHODOLOGY
2.1 NETWORK MODEL

2.1.1 Network Model system


We consider an IoT-based healthcare network consisting of N numbers of static
sensor nodes arbitrarily located in the hospital environment. In the initial network topology,
each node has transmission range Rmax and deployed network topology is represented with
an undirected weighted graph G= (V, E), where is the set of nodes and is the set of edges,
where shows the distance between nodes vi and vj. The data transmission links between
any two nodes do not have the same capacity and edge is expected to have a stipulated
capacity to transmit packets from node vi to node vj. Each deployed node periodically sends
the sensed data to the BS through multi-hop communication. The first order radio model is
used for energy consumption [20- 22]. So the energy consumed while transmitting a -bit
message to a receiver node located d distance apart can be formulated as the equation, 1 2
where Eelec denotes transmitting circuit loss and d0 is the threshold distance. and denote
the energy for power amplification in the models, respectively. The energy consumption
for receiving an –bit message depends on the dissipation in the circuit.

For idle listening, the energy consumption rate of the sensor nodes is denoted by . idle E
The main objective of a Fog architecture is to process users requests, named hereafter jobs.
Conversely to other Cloudoriented work presented, we consider here a fully-distributed
Fog environment, which means that jobs are not submitted to a central scheduler. Each new
job is sent from an IoT or enduser device to the closest Fog node. Jobs are continuously
submitted to the platform and treated as soon as they are received. We do not consider a
distributed file system such as a network file system. We assume that jobs do not require
to share data among them, as it is typically the case for videostream decoding applications
for instance [26].

2.1.2 Problem Formulation

Let xi is the packet sending rate of vi and if xi is detected as zero then vi is out of
work. Therefore, Gx=(V, E, X) graph simulate a topology of sensor-based IoT network, in
which have a static power consumption based on their idle consumption (Pidle), Ethernet
Optimizing Green Energy Consumption of Fog Computing Architecture

X=(x1, x2,…., xn)T. X presents the sending rate of all deployed nodes, which varies with
time. In the IoT network, when traffic demand exceeds the network’s bandwidth, data
packets will accumulate in the cache of nodes, which will cause congestion. This means
that when node vi sends data to vj, if , then the queue length of cache in the node vj will
gradually increase, the network congestion occurs. QH, QL, and QC are the specific queues
used for storing different types of medical data in relay nodes.

The congestion index helps in identifying the level of congestion in the IoT network and is
given by the following equation: P = P idle + P cpu (u) + P eth,idle + P wlan,idle

A virtual machine (VM) is created to host each job and this VM is deleted after the
end of the job. The only parameter for the creation of a VM is the number of required cores.
Cores are not shared between VMs (i.e. we do not consider over-commitment here as it
would modify the performance and energy consumption model of jobs).

The power model used for the network consumption comes from Guegan and al.
[13]. The energy consumption of a link corresponds to the consumption of the its two
Ethernet devices’ ports connected by the link.

It is the sum of the idle power consumption of both ports (P net idle), and its maximum
dynamic power consumption is given by the following equation depending on the
maximum packet size (Maximum Transmission Unit), the transmitted data volume (D), the
link bandwidth (BW), and an energy cost per processed byte (Byte) and per processed
packet (Epkt):

Ci=I,jen,Vi,JeV(Xj-Xi)+Qi

2.1.3 PROPOSED CONGESTION

The main objective of the proposed scheme is to control congestion in IoT-based


healthcare aware WSNs which can potentially increase the network’s lifetime. The
proposed scheme also considers the fairer delivery of sensing information to the gateway.
To achieve these objectives, the proposed scheme considers two main parameters energy
and delay. Furthermore, it considers two types of traffic: sensitive and non-sensitive.
Sensitive traffic is designed to transfer high priority data and non-sensitive traffic is
designed to transfer grid. Similarly, to simply examine the integration of renewable energy

Dept. of CSE, AIET, Mijar Page 8


Optimizing Green Energy Consumption of Fog Computing Architecture

The proposed congestion control scheme consists of three phases: 1) setup phase, 2) request
distribution phase and 3) data routing and event occurrence report phase. In healthcare
applications, due to different stages of vital signs and different priorities of several IoT
based medical devices, the proposed scheme forwards sensed information to the gateway
node with different priorities. Before packets transmission, the source node sets a priority
according to the importance of the data. Intermediate nodes route these packets according
to the priority of the data packets.

2.2 SETUP PHASE

The message contains its ID, level (L) and location information. When a sensor
node vi receives the LEVEL message then vi sets its level value to one higher than the
gateway, i.e. L(vi)= L(Gateway)+1 and also sets the gateway as its parent node (PN) i.e.
PN(vi)=Gateway. Similarly, all sensor nodes within the range 2Rmax to the gateway node
increase their level value to one higher than the gateway node and set the gateway as their
parent node. Recursively, node vi broadcasts a MOD_LEVEL message to the deployed
sensor nodes in the range 2Rmax. The message contains its ID, L(v1), current energy
condition (Ecurrent), and location information. If a sensor node vj receives the
MOD_LEVEL message and its level is less than the level value of the node vi, it sets as one
of the child nodes i.e. CN(vj)=vi. Otherwise, it updates its level to one more than the level
value of the node vi and sets it as one of the parent nodes, i.e., PN(vj)= vi. In the setup
phase, all deployed nodes detect their single hop parent node-set and their current energy
condition.

sensor nodes also identify their single-hop child nodes and their current energy
condition. After level detection, all sensor nodes transmit their update level value, ID,
location information and current energy condition to gateway node through their
intermediate parent nodes. Due to mobility, sensor nodes may change their current
locations. If the sensor node moves from one level to another level, the mobile node updates
its current level value with the help of neighbor nodes. The detailed description of the
process is summarized in algorithm 1.

2.2.1 Query Distribution Phase

In this phase, when the gateway node receives a query from the medical staff
(doctor/nurse), the gateway node distributes received requirements among deployed nodes
to fulfill the need of the medical staff. In healthcare applications, the type of data is very

Dept. of CSE, AIET, Mijar Page 9


Optimizing Green Energy Consumption of Fog Computing Architecture

important during the data gathering process. In some situations, health parameters may
include highly sensitive information. During the query distribution phase, gateway
identifies level and location information of the sensor node from the earlier record and
calculates the lifetime of a query packet within the network. The lifetime of the query
packet can potentially increase the reliability of data transmission and reduces message
overhead within the network. The lifetime of the query packet is calculated as follows

the previous packet’s acknowledgment or after a time-out from the former


submission. The link capacity, , is then approximated by the total number of acknowledged
packets divided by the period T. When a node vi receives RREQ message, it checks
destination node ID. If request node ID is the same with vi, vi sends its update information
to the gateway through the highest capacity link. Otherwise, it checks the level value of the
destination nodes and lifetime of the request packet. If the level of the node vi is equal or
less than the level of the destination node and lifetime of the request packet is expired, it
simply discards the message. Otherwise, it forwards the RREQ message within the child
nodes. In healthcare applications, the gateway may receive multiple requests at the same
time for the same or multiple IoT based medical devices. p i vvC , host. Termination
conditions are:

Therefore, the proposed scheme identifies the priority of the query according to the priority
of the patient. The proposed scheme also considers the request time to specify the execution
of the query for a part node. For example, in healthcare applications, vital signs related to
sensitive information (such as the heart rate, breathing condition, and blood sugar) have
high priority; therefore this type of request is assigned a high priority. Also proposed
scheme can consider other types of traffics for data related to non-sensitive (such as legs
sensors) which has a lower priority. In the proposed scheme, when the gateway node
receives a query from the medical staff, gateway sets query dissemination time concerning
the current time. On the other hand, each sensor node attached to the patient’s body should
report to the gateway when it detects vital signs out of the normal range. In that situation,
the source node sets high priority during the vital sign transmission. The detail description
of the query distribution process is summarized in the algorithm 2.

Dept. of CSE, AIET, Mijar Page 10


Optimizing Green Energy Consumption of Fog Computing Architecture

2.2.2 Data Routing and Event Report Phase

After the request dissemination phase, the destination node sends its update information to
the BS/gateway node through the energy-efficient highest capacity link. Similarly, if a node
senses a medical emergency based on its duty, it will report to the gateway according to the
specifications. The reports must have the required parameter values so that the gateway can
show the proper reaction and sends a medical emergency report to the doctor/nurse.

In this phase, the information related to the occurring medical emergency is sent to
the gateway. For this purpose, the node creates a packet containing the information related
to the sensed event and sends it to the nearest parent node through the highest capacity link.
According to the setup phase, a sensor node may have multiple parent nodes and multiple
paths to the gateway. Let {v1, v2, v3,…..vp} are the set of nodes belonging to set PN(vk).
To measure the impact of consolidation, we use the number of nodes that are placed on a
booting host, i.e., a host which is awaken by the placement of the job on it, or a host already
in the booting process. The unavailability of nodes can reduce the responsiveness of the
architecture.

Some jobs might have to wait for a machine to boot up until they can run on it.
However, once a job is allocated, the boot duration is not an issue anymore: it does not
impact the latency perceived by the user anymore. In this sense, we introduce the notions
of final latency: the time from initial node to the host after the allocation and eventual
booting duration, only taking into account the communication time, i.e., latency due to the
distance between users.

where p is the number of nodes in the PN(vk) and Ecurrent is current energy
condition. Let S={s1, s2, …, sp} be the parent nodes that residual energy is greater than or
equal to . Then vk sends all data packets to the parent node concerning the buffer status of
the parent sensor node (which is specified by congestion index Ci) and distance of the
parent sensor node. If the buffer status of the nearest parent node is overflow, vk selects the
next nearest parent from S. The pseudo-code of the routing algorithm is described as
follows.

Dept. of CSE, AIET, Mijar Page 11


Optimizing Green Energy Consumption of Fog Computing Architecture

------------------------------------------------------------------------------------------------------------

Algorithm 1: Level Detection Algorithm

Input: Deployed sensor nodes

Output: A level value has allotted to all sensor nodes

1. Initialize: L(Gateway)=1;// L(Gateway) initialized to one

2. Gateway broadcasts a LEVEL msg. in the range Rmax.

3. for each sensor node vi

4. if (D(Gateway,vi) ≤ 2Rmax) then

5. L (s)= L(Gateway)+1;

6. PN(vi)=Gateway;

7. end if

8. end for

9. Similarly, node vi broadcasts a MOD_LEVEL msg. in the range 2Rmax

10. if (node vj receives the msg. and L(vj)>L(vi)) then

11. L(vj)=L(vi)+1;

12. PN(vj)=vi

13. CN(vj) =vi.

14.else

15. Discord the msg.

16.end if

17. for each sensor node vi

18. vi sends its current level value, energy condition, ID, Location

information;

19. end for

Dept. of CSE, AIET, Mijar Page 12


Optimizing Green Energy Consumption of Fog Computing Architecture

Algorithm 2: Request Distribution Algorithm

Input: Query from the doctors and nurses

Output: Respond of the query

1. Gateway receives a query from the doctors and nurses

2. Start answering process according to request priority

3. if required data is not available then

4. Gateway identifies level and node ID of the vj node

5. Calculated LLife_Packet using eqn. 1

6. Set life time of the query packet

7. Broadcast RREQ msg. within the single-hop one node

8. for each node vi

9. if node ID(vi) = ID(vj) then

10. Send requested data to the gateway

11. end if

12. if L(vi) ≠ L(vj) then

13. Node vi forwards RREQ msg. within the child nodes

14. end if

15. if LLife_Packrt=0 and L(vi) < L(vj) then

16. Discord the message

17. end if

18. end for

Dept. of CSE, AIET, Mijar Page 13


Optimizing Green Energy Consumption of Fog Computing Architecture

Algorithm 3: Event Distribution Algorithm


Input: Event identifies by the sensor node
Output: Event report receive at BS/Gateway
1. for each sensor node vk
2. for each sensor vi belongs to PN(vk)
3. sum=sum + Ecurrent (vi)
4. end for
5. ( ) / ( ) k k v sum PN v
6. for each sensor node vi belongs to PN(vk)
7. if (Ecurrent (vi) ≥ ( ) k v
8. S=union (S,v1)
9. end if
10. end for
11. for each sensor node vl belongs to S
12. if (MAX> Ecurrent (vk) && min D(vl,vk) then
13. MAX=Ecurrent(vk)
14. Node vl selects data routing.
15. end if
16. end for

The main objective of this work is to provide a routing and congestion management
mechanism for IoT-based healthcare network. In this work, deployed nodes select their data
routing path in such a way that the congestion occurrence probability will be reduced. The
congestion control phase is executed with the data routing algorithm.
 The proposed congestion control scheme uses a classifier to classify different types
of data and route them into different queues. All data packets are classified into
three categories: a) high priority data packets, b) low priority data packets, and c)
control data packets. Each data packet contains its type in the packet header.
Classifier classifies received data packets and sends them into different queues
concerning their class. The proposed scheme uses a priority queue based scheduler
to transmit different data. If class 1 data is found within the queue, scheduler stop
class 2, class 3 data transmission and start class 1 data transmission. After
transmission of all class 1 data packets, the scheduler starts class 2 and class 3 data
packets transmission.

Dept. of CSE, AIET, Mijar Page 14


Optimizing Green Energy Consumption of Fog Computing Architecture

 data from its child node and the network suffers from the packet loss problem. To
solve this problem, the proposed scheme selects an alternative path when the buffer
level reaches the threshold value (which is identified by Ci). When a node receives
a data packet from the child node then it sends buffer status with the ACK message.
If a node v1 detects that class 1 queue has reached the threshold value, it sends
H_AlterPath message to all child nodes so that the high priority data is sent to
another immediate energy-efficient shortest path.
 When a node vk receives an alternative path selection message from the parent node
then vk selects the next nearest energy level node from the parent node set PN(vk)
and sends all high priority data from this parent node. Similarly, when node vk
detects it is class 2 queue and class 3 queue have overflows then it sends
L_AlterPath and C_AlterPath message to all child nodes. Child nodes also select
an alternative path for class 2 and class 3 data packet transmission.
 The threshold value for the alternative path selection process depends on the
number of child nodes hold by a parent node and receives the data rate of the node.
In this work, when a parent node detects that its queue is 95% full by received data
packets, then it sends an alternative path selection request to the child nodes.
 After some time when node vk again senses that its buffer is less than the threshold
value then also it sends REQ 1, REQ 2, and REQ 3 message to all child nodes for
resending class 1, class 2, and class 3 data packets.

The idea behind this node availability ratio consists of reducing the initial
latency: available nodes are kept powered on even if they are not used for speeding
up the deployment of new incoming jobs.

2.3 THEORETICAL ANALYSIS OF PROPOSED CONGESTION

CONTROL MECHANISM

2.3.1 EXPERIMENTAL SETUP


This section examines the complexity of the proposed congestion control algorithm
to provide a theoretical analysis of whether the proposed solutions are applicable in real-
world healthcare practice.
Theorem 1: The message complexity of the proposed congestion control algorithm in O
(N) where N is the number deployed IoT-enabled sensor nodes.

Dept. of CSE, AIET, Mijar Page 15


Optimizing Green Energy Consumption of Fog Computing Architecture

Proof: The total number of messages exchanged by the deployed IoT based sensor nodes
over the network for executing the proposed congestion control scheme is known as
message complexity. In the setup phase, the proposed congestion control scheme needs one
message exchange over the network that means each node vi broadcast a single message to
their neighbors. Based on the received data each sensor node vi calculates its level value
and parent node. Therefore, the proposed congestion control algorithm needs maximum O
(N) message exchange over the network.
Theorem 1: Assume that vi is in the small region of Ad with the width of . Denote d as the
distance between Ad and the base station. If each node receives one query packet per round,
the average data sent by vi in a round at S0.

IoT-based healthcare aware network is evaluated and compared with the existing well-
known TARA [8], REEP [11], HTAP [17], BRCCTP [18], and CADC [19] schemes in the
literature in terms of percentage of the successfully received packet, average throughput,
average hop-by-hop delay. Furthermore, the energy-saving performance of the proposed
algorithm and the network lifetime was evaluated where the initial energy of each sensor
node is E0=0.5 [Joules]. All the simulations are based on collision-free MAC protocol.

Simulation table

Parameter Value
Deployment area 100×100 m2
Number of nodes 20-100
Data packet size 500 bits
Initial energy of each 0.5 Joules
node (E0)
50nJ/bit
0.0013 pJ/bit/m4
10pJ/bit/m2
Control message size 100 bits
Duty cycle 10%
Duration of a data 10s
period
Energy consumption 0.88 mJ/s
rate for ideal
listening
Data transmission 512 Kb/s
rate
Sensing radius 20m
Antenna type Omni Antenna
Max packet in queue 50

Dept. of CSE, AIET, Mijar Page 16


Optimizing Green Energy Consumption of Fog Computing Architecture

PARAMETERS VALUES USED IN THE INDOOR TESTED EXPERIMENT

Parameter Value
Number of sensor 20
nodes
Deployment area 15 15
Base 1
Station/Gateway
Sending rate 1 packet/s
Packet size 500 bits
Initial voltage of 1.6.0V
battery

System Implementation and Testing

A testbed of the proposed congestion control scheme has been created for
demonstration where twenty medical sensor nodes are used for data collection. In this
experimental setup, we used a blood pressure sensor, pulse oximetry sensor, ECG sensor,
EMG sensor for data collection. The testbed consists of a gateway for query processing and
data collection from the deployed medical sensor nodes. Twenty medical sensors are
deployed into two different laboratory rooms where each person is monitored by four
medical sensors. The detailed parameters and corresponding values in real experiments are
reviewed in Table 2.

shows that the result of the practical test is almost similar to that of the simulation
case in Fig. 1. As predicted by the theoretical case in Fig.1, it can be seen that the percentage
of successfully received packets of the proposed scheme was decreased when sources data
rate increased. From these results, it can be observed that our proposed scheme shows
improved performance compared to the existing congestion control approaches in the real
IoT-based healthcare experiments.

illustrates that our proposed scheme has an increased average throughput compared to
BRCCTP, CADC, TARA, HTAP, REEP schemes, verifying the theoretical case in Fig. 2.
From the figures, it is clear that the proposed scheme demonstrates desirable performance
for the congestion control mechanism in an IoT-based healthcare system.

Values used for the power model are summarized in Table II. Note that P eth,idle is
negative to account for the periodic CPU activity even when Ethernet network is not used;
this activity stops when the network is active.

Dept. of CSE, AIET, Mijar Page 17


Optimizing Green Energy Consumption of Fog Computing Architecture

This power values come from Kaup and al. [17]. The job traces used in our simulation do not refer
to data exchange between nodes and we suppose that data exchange time is far lower than data
processing time for the considered Fog applications. It corresponds to highly distributed
applications where data is produced and consumed locally. During shutting down or booting, a node
consumes as much as if all its cores were fully loaded. Each node is considered to take 150 seconds
to boot and 10 seconds to shut down [5].

Fig. 1a and 1b present the percentage of successfully received packets of the proposed
scheme. This is measured with the number of data packets received at the gateway. Fig. 1a
shows that the percentage of successfully received packets of the proposed scheme is more
by 37% as compared with BRCCTP, 39.4% as compared with CADC, 41.3% as compared
with the REEP scheme, 42.5% as compared with the HTAP scheme, and 57% as compared
with the TARA algorithm. Also, Fig. 1b indicates that the proposed scheme increases the
percentage of successfully received packets up to 38% comparing with BRCCTP, up to
42% comparing with CADC, up to 44 % comparing with REEP, up to 45.7% comparing
with HTAP and up to 59% comparing with TARA. This causes due to the level-wise
elimination of data congestion during the data routing process and priority-based
congestion control mechanism.

Dept. of CSE, AIET, Mijar Page 18


Optimizing Green Energy Consumption of Fog Computing Architecture

Fig. 2a and 2b present the throughput of the proposed scheme using different percentages
of sensor nodes in the network. This is measured with the rate of data that reaches the
gateway node. Fig. 2a indicates that the proposed scheme improves throughput up to 29%
comparing with BRCCTP, up to 31% comparing with CADC, up to 33% comparing with
REEP, up to 34.33% comparing with TARA and up to 36% comparing with HTAP. Fig.
2b shows that the proposed scheme improves throughput up to up to 26.5% comparing with
BRCCTP, up to 29% comparing with CADC, up to 31.21% comparing with REEP, up to
32% comparing with TARA and up to 33.45 % comparing with HTAP. It is due to the
level-based congestion control mechanism where data transmission congestion is
controlled through the effective alternative data routing path selection process.

IoT-based healthcare aware network is evaluated and compared with the existing well-
known TARA [8], REEP [11], HTAP [17], BRCCTP [18], and CADC [19] schemes in the
literature in terms of percentage of the successfully received packet, average throughput,
average hop-by-hop delay. Furthermore, the energy-saving performance of the proposed
algorithm and the network lifetime was evaluated where the initial energy of each sensor
node is E0=0.5 [Joules]. All the simulations are based on collision-free MAC protocol.

We performed extensive experiments on the proposed scheme using NS-2 [23]. The
medical sensor nodes are randomly distributed in 2D space. An IoT-enabled gateway/BS
was used for data gathering and queries processing.

Dept. of CSE, AIET, Mijar Page 19


Optimizing Green Energy Consumption of Fog Computing Architecture

Fig. 3a and 3b present average hop-by-hop delay results using the tested algorithms in
scenario 1 (20 nodes) and scenario 2 (100 nodes). This performance metric is important for
how the proposed scheme controls inter-path interferences, link-layer retransmissions, and
overhead introduced by control packet exchanges. In Fig. 3a, it is seen that the proposed
scheme has a less average hop-by-hop delay of 31% as compared to BRCCTP, 33.6% as
compared to CADC, 35.4% as compared to REEP, 38.5 % as compared with HTAP, and
41% as compared with TARA algorithm. Fig. 3b indicates that the proposed scheme
reduces average hop-by-hop delay up to 28% comparing with BRCCTP, up to 31.3%
comparing with CADC, up to 32% comparing with REEP, up to 35% comparing with
HTAP, and up to 37% comparing with TARA algorithm. It is since the proposed scheme
efficiently selects the data routing path with the minimum number of control packet
exchanges. Therefore, the average data transmission delay is less in the proposed scheme .

Fig. 5b illustrates that our proposed scheme has an increased average throughput compared
to BRCCTP, CADC, TARA, HTAP, REEP schemes, verifying the theoretical case in Fig.
2. From the figures, it is clear that the proposed scheme demonstrates desirable performance
for the congestion control mechanism in an IoT-based healthcare system.

Dept. of CSE, AIET, Mijar Page 20


Optimizing Green Energy Consumption of Fog Computing Architecture

Fig. 5a shows that the result of the practical test is almost similar to that of the simulation
case in Fig. 1. As predicted by the theoretical case in Fig.1, it can be seen that the percentage
of successfully received packets of the proposed scheme was decreased when sources data
rate increased. From these results, it can be observed that our proposed scheme shows
improved performance compared to the existing congestion control approaches in the real
IoT-based healthcare experiments.

the throughput of the proposed scheme using different percentages of sensor nodes
in the network. This is measured with the rate of data that reaches the gateway node. Fig.
2a indicates that the proposed scheme improves throughput up to 29% comparing with
BRCCTP, up to 31% comparing with CADC, up to 33% comparing with REEP, up to
34.33% comparing with TARA and up to 36% comparing with HTAP. Fig. 2b shows that
the proposed scheme improves throughput up to up to 26.5% comparing with BRCCTP, up
to 29% comparing with CADC, up to 31.21% comparing with REEP, up to 32% comparing
with TARA and up to 33.45 % comparing with HTAP. It is due to the level-based
congestion control mechanism where data transmission congestion is controlled through
the effective alternative data routing path selection process.

Furthermore, the energy-saving performance of the proposed algorithm and the network
lifetime was evaluated where the initial energy of each sensor node is E0=0.5 [Joules]. All
the simulations are based on collision-free MAC protocol. Simulation parameters are listed.

Dept. of CSE, AIET, Mijar Page 21


Optimizing Green Energy Consumption of Fog Computing Architecture

Fig. 6a shows the average hop-by-hop delay in the practical test. It can be seen that the
average hop-by-hop delay has been reduced compared to BRCCTP, CADC, TARA, HTAP,
and REEP schemes. As described in the theoretical case, our proposed scheme shows an
improved result because of the class-based congestion management mechanism.

Fig. 6b shows the lifetime over traffic load. As predicted by the simulation tests, it can be
seen that the lifetime of the proposed scheme was increased when the traffic load was
increased within the network. From this result, it can be observed that our proposed scheme
shows improved performance in an IoT-based healthcare system compared to the state-of-
the-art congestion control approaches such as BRCCTP, CADC, TARA, HTAP, and REEP
in the real hardware experiments.

2.3.2 System Implementation and Testing

A testbed of the proposed congestion control scheme has been created for demonstration
where twenty medical sensor nodes are used for data collection. In this experimental setup,
we used a blood pressure sensor, pulse oximetry sensor, ECG sensor, EMG sensor for data
collection. The testbed consists of a gateway for query processing and data collection from
the deployed medical sensor nodes. Twenty medical sensors are deployed into two different
laboratory rooms where each person is monitored by four medical sensors.

Each configuration is simulated 10 times, the results provided hereafter exhibit the
average and standard deviation values of these 10 runs for each configuration.

Dept. of CSE, AIET, Mijar Page 22


Optimizing Green Energy Consumption of Fog Computing Architecture

Variables Values Expired

Allocation Algorithm Lowest Latency, First Green, Most Green

Consolidation Policy None, Consolidate All, Consolidate Brown,


Consolidate Ratio

Number of green producers per district 1, 3, 5

Latency Range (ms) 0, 5, 10, 20

Node availability ratio 0.1, 0.2, 0.3, 0.4

Per-node availability threshold 0.25, 0.5, 0.75

Table 2.4 : PARAMETERS EXPLORED IN THE SIMULATIONS

Energy Consumption of Nodes: The energy consumption of nodes for the allocation
algorithm First Green is shown in Fig 2.1. It presents the energy consumption of the
infrastructure for the nodes, depending on the consolidation policy.

The policies are those detailed in Section IV: NC: No Consolidation, CA:
Consolidate All, CB: Consolidate Brown, CR: Consolidation Ratio. The behavior of the
consolidation policy Consolidation Ratio depends on two parameters, as explained in
Section IV: node availability ratio and per-node availability threshold.

The other allocation algorithms follow the same trend and are omitted for saving
space. For each allocation algorithm, the energy consumed by the nodes is similar. Without
consolidation the infrastructure consumes significantly more energy. Among the
consolidation policies, Consolidate Ratio is the one consuming the most energy since it
consolidates less nodes by keeping some available nodes.

The overall green energy consumption is small compared to the brown energy
consumption, but it is coherent considering that only one user node per district produces
green energy and they can produce only during daytime.

Dept. of CSE, AIET, Mijar Page 23


Optimizing Green Energy Consumption of Fog Computing Architecture

The allocation algorithms First Green and Most Green have an additional parameter
compared to Lowest Latency: the latency range. Concerning more particularly the
allocation algorithm First Green, the results of the simulations exploring the impact of the
latency range and the number of green producers per district is shown in green.For these
experiments, we do not apply any consolidation policy in order to isolate the latency effect.

To solve this problem, the proposed scheme selects an alternative path when the
buffer level reaches the threshold value (which is identified by Ci). When a node receives
a data packet from the child node then it sends buffer status with the ACK message.

If a node v1 detects that class 1 queue has reached the threshold value, it sends
H_AlterPath message to all child nodes so that the high priority data is sent to another
immediate energy-efficient shortest path. When a node vk receives an alternative path
selection message from the parent node then vk selects the next nearest energy level node
from the parent node set PN(vk) and sends all high priority data from this parent node.
Similarly, when node vk detects it is class 2 queue and class 3 queue have overflows then
it sends L_AlterPath and C_AlterPath message to all child nodes. Child nodes also select
an alternative path for class 2 and class 3 data packet transmission.

The threshold value for the alternative path selection process depends on the
number of child nodes hold by a parent node and receives the data rate of the node. In this
work, when a parent node detects that its queue is 95% full by received data packets, then
it sends an alternative path selection request to the child nodes. After some time when node
vk again senses that its buffer is less than the threshold value then also it sends REQ 1, REQ
2, and REQ 3 message to all child nodes for resending class 1, class 2, and class 3 data
packets.

The main objective of this work is to provide a routing and congestion management
mechanism for IoT-based healthcare network. In this work, deployed nodes select their data
routing path in such a way that the congestion occurrence probability will be reduced. The
congestion control phase is executed with the data routing algorithm.

The proposed congestion control scheme uses a classifier to classify different types
of data and route them into different queues. All data packets are classified into three
categories: a) high priority data packets, b) low priority data packets, and c) control data
packets. Each data packet contains its type in the packet header. Classifier classifies
received data packets and sends them into different queues concerning their class. The

Dept. of CSE, AIET, Mijar Page 24


Optimizing Green Energy Consumption of Fog Computing Architecture

proposed scheme uses a priority queue based scheduler to transmit different data. If class 1
data is found within the queue, scheduler stop class 2, class 3 data transmission and start
class 1 data transmission. After transmission of all class 1 data packets, the scheduler starts
class 2 and class 3 data packets transmission.

WSNs called Bidirectional Reliable and Congestion Control Transport Protocol


(BRCCTP). In this approach, a rate adjustment based scheme is used to manage congestion
within the network. It provides bidirectional reliability during data transmission. The
proportion of average packet service time over average packet inter-arrival time is used to
detect the congestion level in the nodes. In this approach, all data sources are assigned by
an equal priority index that helps to mitigate congestion. In this approach, if congestion is
detected within the network then it significantly increases data transmission latency.

congestion control mechanize for data gathering in sensor network called


Congestion-Adaptive Data Collection scheme (CADC). In this paper, an adaptive lossy
compression based mechanism is developed to mitigate congestion. Furthermore, a
weighted CADC scheme is applied to manage cyber-physical applications congestion. In
CADC, data items are defined by the different priorities. However, CADC does not focus
on the energy consumption of the deployed sensor nodes. Therefore, this approach suffers
from a poor network lifetime.

Dept. of CSE, AIET, Mijar Page 25


CHAPTER 4

CONCLUSION AND FUTURE ENHANCEMENT

This paper proposed a congestion control mechanism for wireless sensor networks
applied in IoT-based healthcare scenarios. The proposed scheme used a priority-based data
routing strategy to control congestion within the network. It classifies data packets into
three different categories according to the priority of the data packets. Furthermore, the
proposed scheme used a priority queue based scheduler to transmit

shows that the result of the practical test is almost similar to that of the simulation
case in Fig. 1. As predicted by the theoretical case in Fig.1, it can be seen that the percentage
of successfully received packets of the proposed scheme was decreased when sources data
rate increased. From these results, it can be observed that our proposed scheme shows
improved performance compared to the existing congestion control approaches in the real
IoT-based healthcare experiments.

The lifetime over traffic load. As predicted by the simulation tests, it can be seen
that the lifetime of the proposed scheme was increased when the traffic load was increased
within the network. From this result, it can be observed that our proposed scheme shows
improved performance in an IoT-based healthcare system compared to the state-of-the-art
congestion control approaches such as BRCCTP, CADC, TARA, HTAP, and REEP in the
real hardware experiments.

A testbed of the proposed congestion control scheme has been created for
demonstration where twenty medical sensor nodes are used for data collection. In this
experimental setup, we used a blood pressure sensor, pulse oximetry sensor, ECG sensor,
EMG sensor for data collection. The testbed consists of a gateway for query processing and
data collection from the deployed medical sensor nodes. Twenty medical sensors are
deployed into two different laboratory rooms where each person is monitored by four
medical sensors.
REFERENCES

[1] Arif Ahmed et al. Fog computing applications: Taxonomy and requirements. CoRR,
abs/1907.11621, 2019.

[2] Sherif Akoush, Ripduman Sohan, Andrew Rice, and Andy Hopper. Evaluating the viability of
remote renewable energy in datacentre computing. Technical report, University of Cambridge,
2016.

[3] Anton Beloglazov, Jemal Abawajy, and Rajkumar Buyya. Energyaware Resource Allocation
Heuristics for Efficient Management of Data Centers for Cloud Computing. Future Gener. Comput.
Syst., 28(5):755– 768, May 2012.

[4] Benjamin Camus, Anne Blavette, Fanny Dufosse, and Anne-C ´ ecile Org- ´ erie. Self-
Consumption Optimization of Renewable Energy Production in Distributed Clouds. In IEEE Int.
Conf. on Cluster Computing, pages 1–11, 2018.

[5] Benjamin Camus, Fanny Dufosse, Anne Blavette, Martin Quinson, and ´ Anne-Cecile Orgerie.
Network-aware energy-efficient virtual machine ´ management in distributed Cloud infrastructures
with on-site photovoltaic production. In Int. Symp. on Computer Architecture and High-
Performance Computing (SBAC-PAD), pages 1–8, 2018.

[6] Henri Casanova, Arnaud Giersch, Arnaud Legrand, Martin Quinson, and Fred´ eric Suter.
Versatile, scalable, and accurate simulation of distributed ´ applications and platforms. Journal of
Parallel and Distributed Computing, 74(10):2899–2917, June 2014.

[7] L. Chiaraviglio, M. Mellia, and F. Neri. Energy-aware backbone networks: A case study. In
International Conference on Communications Workshops, pages 1–5, 2009.

[8] Ismael Cuadrado-Cordero, Anne-Cecile Orgerie, and Christine Morin. ´ GRaNADA: A


Network-Aware and Energy-Efficient PaaS Cloud Architecture. In IEEE International Conference
on Green Computing and Communications (GreenCom), 2015.

[9] R. Deng, R. Lu, C. Lai, T. H. Luan, and H. Liang. Optimal Workload Allocation in Fog-Cloud
Computing Toward Balanced Delay and Power Consumption. IEEE Internet of Things Journal,
3(6):1171–1181, 2016.

[10] U. Drolia, K. Guo, J. Tan, R. Gandhi, and P. Narasimhan. Cachier: EdgeCaching for
Recognition Applications. In International Conference on Distributed Computing Systems
(ICDCS), pages 276–286, 2017.

[11] Hughes Ferreboeuf and al. Lean ICT, pour une sobriet´ e num ´ erique. ´ Rapport intermediaire
du groupe de travail – The Shift Project, 2018.

[12] ´Inigo Goiri, Kien Le, Md. E. Haque, Ryan Beauchea, Thu D. Nguyen, ˜ Jordi Guitart, Jordi
Torres, and Ricardo Bianchini. GreenSlot: Scheduling Energy Consumption in Green Datacenters.
In Int. Conf. for High Performance Computing, Networking, Storage and Analysis, 2011.

[13] Loic Guegan, Betsegaw Lemma Amersho, Anne-Cecile Orgerie, and ´ Martin Quinson. A
Large-Scale Wired Network Energy Model for FlowLevel Simulations. In Int. Conf. on Advanced
Information Networking and Applications (AINA), volume 926, pages 1047–1058, March 2019.

[14] F. C. Heinrich, T. Cornebize, A. Degomme, A. Legrand, A. CarpenAmarie, S. Hunold, A.-C.


Orgerie, and M. Quinson. Predicting the Energy-Consumption of MPI Applications at Scale Using
Only a Single Node. In IEEE Int. Conf. on Cluster Computing, pages 92–102, 2017.

[15] T. Hirofuchi, A. Lebre, and L. Pouilloux. Simgrid vm: Virtual machine support for a simulation
framework of distributed systems. IEEE Transactions on Cloud Computing, 6(1):221–234, 2018.

[16] F. Jalali, K. Hinton, R. Ayre, T. Alpcan, and R. S. Tucker. Fog Computing May Help to Save
Energy in Cloud Computing. IEEE Journal on Selected Areas in Communications, 34(5):1728–
1739, 2016.

[17] F. Kaup, S. Hacker, E. Mentzendorff, C. Meurisch, and D. Hausheer. Energy models for NFV
and service provisioning on fog nodes. In IEEE/IFIP Network Operations and Management Symp.
(NOMS), 2018.

[18] Yunbo Li, Anne-Cecile Orgerie, and Jean-Marc Menaud. Balancing the ´ use of batteries and
opportunistic scheduling policies for maximizing renewable energy consumption in a Cloud data
center. In Euromicro Int. Conf. on Paral., Distr., and Network-Based Processing (PDP), 2017.

[19] Openfog consortium. out of the fog: Use case scenarios (visual security surveillance).
https://www.openfogconsortium.org/wpcontent/uploads/Video-Surveillance-Use-Case.pdf, 2018.

[20] Openfog consortium. process manufacturing in beverage industry.


https://www.openfogconsortium.org/wp-content/uploads/BeverageIndustry-Short.pdf, 2018.

[21] Anne-Cecile Orgerie, Marcos Dias de Assuncao, and Laurent Lefevre. A Survey on
Techniques for Improving the Energy Efficiency of Largescale Distributed Systems. ACM Comp.
Surveys, 46(4):47:1–47:31, 2014.

[22] Photovolta project. http://photovolta2.univ-nantes.fr/, accessed April 2020.

[23] Xueyan Tang, Changbing Chen, and Bingsheng He. Green-aware Workload Scheduling in
Geographically Distributed Data Centers. In IEEE International Conference on Cloud Computing
Technology and Science (CloudCom), pages 82–89, 2012.

[24] B. Thomas, B. Close, J. Donoghue, J. Squires, P. De Bondi, M. Morris, and W. Piekarski.


ARQuake: an outdoor/indoor augmented reality first person application. In International
Symposium on Wearable Computers, pages 139–146, 2000.

[25] Pedro Velho, Lucas Mello Schnorr, Henri Casanova, and Arnaud Legrand. On the validity of
flow-level tcp network models for grid and cloud simulations. ACM Trans. Model. Comput. Simul.,
23(4):1– 26, 2013.

[26] Jon Whiteaker, Fabian Schneider, Renata Teixeira, Christophe Diot, Augustin Soule, Fabio
Picconi, and Martin May. Expanding Home Services with Advanced Gateways. SIGCOMM
Comput. Commun. Rev., 42(5):37–43, 2012.

[27] R. Wolski and J. Brevik. Using parametric models to represent private cloud workloads. IEEE
Transactions on Services Computing, 7(4):714– 725, 2014.

[28] Y. Xiao and M. Krunz. Distributed Optimization for Energy-Efficient Fog Computing in the
Tactile Internet. IEEE Journal on Selected Areas in Communications, 36(11):2390–2400, 2018.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCE.2020.2987433, IEEE
Transactions on Consumer Electronics
1

 Congestion Free Routing Mechanism for IoT-


enabled Wireless Sensor Networks for Smart
Healthcare Applications
Prasenjit Chanak, Member, IEEE, Indrajit Banerjee, Member, IEEE

Abstract—Recently, the Internet of Things (IoT) topology has network infrastructures together compose the backbone
used to collect physical, physiological, vital signs of patients in network infrastructure for IoT. WSN has a lot of unique
consumer-centric e-health or consumer’ wellness care services. In challenges and constraints such as self-management, energy
such healthcare systems, varieties of medical sensors are attached
limitation, congested packet transmissions, ad hoc
to the patients to collect vital signs from those who are under
observation. The data gathering process in IoT-enabled Wireless deployment, unattended operation, etc. This unique
Sensor Network (WSN) suffers from the congestion problem. The infrastructure nature of WSN demands a protocol designed
effect of this translates on missing packets, a decrease of that should fit for its specialties, especially for the consumer-
reliability and throughput degradation in IoT-enabled WSN. centric e-health or consumer’ wellness care services.
This paper proposes a distributed congestion control algorithm In the field of consumer electronics, several features in the
for IoT-enabled WSNs to effectively resolve the congestion for IoT-based Healthcare aware Wireless Sensor Networks
healthcare applications. The proposed scheme alleviates
(HWSNs) implementation are considered, such as (1) Quality
congestion by a priority-based data routing strategy.
Furthermore, this paper presents a priority queue based of Service (QoS) (2) low cost, (3) low power consumption, (4)
scheduling scheme for better reliability. We analyze the easy integration, and (5) low complexity [6], [7]. The primary
properties of the proposed congestion control mechanism challenge of HWSNs is congestion control during
mathematically and validate its performance through extensive physiological data routing that helps to provide a high QoS.
simulation and real-life experiments. The application of this work Congestion control in HWSNs is much more complicated
can be used to an early warning system in detecting abnormal compared to the congestion control in other traditional
heart rate, blood pressure, ECG, EMG in the hospital/home care networks. In HWSNs, different types of congestion may occur
environment to the state-of-art diagnosis. at different points and locations of the networks. In a sensor
network, deployed sensor nodes have limited resources for
Index Terms— Healthcare system, Internet of Things (IoT),
computing, bandwidth, memory storage, energy supply, and
Wireless Sensor Network (WSN), Congestion Control.
communication [8-10]. The event-driven character of the
network channels to an unpredictable communication load
I. INTRODUCTION
within the network. In healthcare applications, the network

R advances in microelectronics fabrication and Very


ECENT
Large-Scale Integrated (VLSI) circuit technologies
allowed the development of smart electronics devices with the
stocks low traffic load when there are no special events but the
happening of medical emergencies may generate burst traffic
which leads to congestion in the network [11], [12]. Therefore,
capability of sensing, processing, and transmission [1]-[5]. overloaded nodes exhaust more energy than other sensor
These electronic devices are combined to form a network that nodes. Furthermore, they are subject to premature death
achieves some shared objectives. They interact with the because of heavy traffic. Once the nodes with heavy load die,
physical world pervasively with the aid of enhanced some other sensor nodes should relay data to the gateway node
communication protocols and distributed intelligence, which originally forwarded by these dead nodes. It leads to
constitute a novel paradigm called the Internet of Things dynamical changes in the routing paths, as well as creates
(IoT). The IoT is possible through the integration of several congestion within the other sensor nodes. As a result, doctors
heterogeneous network infrastructures such as Radio and nurses cannot receive update information from patients
Frequency Identification (RFID) systems, Machine-to- under observation [13], [14]. On the other hand, in healthcare
Machine (M2M) communication systems, Wireless Sensor applications, sensors attached to the patient’s body sense and
Networks (WSNs) and much more. WSNs are used for transmit vital signs (such as the heart rate and breathing
condition) very frequently and simultaneously to the gateway
tracking the status of the real-world object. All these different
node. Due to dynamical changes of routing paths, huge data
loads can quickly generate congestion within the network.
This work was supported in part by the Development of Science and Congestion creates considerable delays within the network.
Technology (DST), Govt. of India, under grant number SP/YO/036/2017. Therefore, data packets are likely to be lost due to congestion
P. Chanak is with the Department of Information and Communication and a high amount of energy is consumed in forwarding
Technology, ABV-Indian Institute of Information Technology and
Management, Gwalior-474015, M.P., India (e-mail: pchanak@iiitm.ac.in). packet toward the gateway node. A lot of research has been
I. Banerjee is with the Department of Information Technology, Indian conducted to enhance the performance of the HWSNs. A
Institute of Engineering Science and Technology, Shibpur, Howrah-03, India congestion avoidance scheme known as Relaxation Theory
(e-mail: ibanerjee@it.iiests.ac.in).

0098-3063 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Otago. Downloaded on May 31,2020 at 10:01:34 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCE.2020.2987433, IEEE
Transactions on Consumer Electronics
2

with Max-Min Fairness (RT-MMF) is presented to avoid A Fairness-Aware Congestion control scheme (FACC) [7]
congestion in the wireless body sensor network [15]. A was proposed by Huang et al. where a rate-based fairness
congestion avoidance scheme called Adaptive Duty-cycle aware congestion control was reported. This approach divides
Based Congestion Control (ADCC) has proposed to handle all intermediate sensor nodes into near-source and near sink
congestion in home automation networks which use both the nodes. It detects congestion according to the packet loss rate at
resource control and traffic control mechanize to mitigate the sink node. Kang et al. [8] proposed a Topology Aware
packet loss [16]. However, existing congestion control Resource Adaptation (TARA) scheme to alleviate congestion
mechanize does not perform well in the health care application in WSNs. This approach focuses on the adaptation of the extra
because they are unable to reduced transmission delay during resources to control congestion within the network. This
vital signs transmission from the monitoring patients to
approach activates a particular sensor node whose mode is
medical staff. Also, existing approaches used huge message
sleeping to form a new topology. A new topology of the
overhead to control burst traffic within the network that
network can control the increased traffic of the network. It
significantly reduces network lifetime. Moreover, many
approaches are already available in the literature on puts a large overhead for the large scale WSNs. On the other
congestion control routing for WSNs. However, the consumer- hand, Sergiou et al. [17] proposed a Hierarchical Tree
centric IoT environment includes heterogeneous sensor Alternative path (HTAP) algorithm for congestion control in
devices. Hence, the traditional congestion control approaches WSN. It builds a source-based tree to find all possible routes
for WSNs must be enhanced to suit the WSNs in the to the sink. HTAP selects the least buffer node to forward the
consumer-centric IoT environment. Thus, congestion control excess packets. This algorithm also suffers from large message
is one of the most critical problems in the designing and overhead and a huge time delay. Misra et al. [11] proposed a
implementation of WSNs, especially for the consumers-centric Reliable and Energy Efficient Protocol (REEP) for reliable
e-health or consumer’ wellness care services where a vital and energy-efficient data routing in WSNs. It is a data-centric
event may occur any time and on arbitrary nodes within the and on-demand routing protocol where the routing phase
network. selects based on local communication. This approach puts
To address these issues, this paper proposes a distributed huge message overhead on the network. It also includes high
traffic-aware congestion control scheme for IoT-enabled transmission costs.
WSNs. In the proposed scheme, deployed sensor nodes are Sharma et al. [18] presented a congestion control protocol
divided into different levels and transmit their sensing for the WSNs called Bidirectional Reliable and Congestion
information through the congestion-free path. The proposed Control Transport Protocol (BRCCTP). In this approach, a
algorithm minimizes the total energy consumption of the rate adjustment based scheme is used to manage congestion
network and improves the Quality of Services (QoS). For
within the network. It provides bidirectional reliability during
efficient and reliable communication in IoT networks, we aim
data transmission. The proportion of average packet service
to frame a cross-layer design that spans different layers, i.e.,
time over average packet inter-arrival time is used to detect
stable congestion-free routing in the network layer, effective
access control and transmission power control in MAC sub- the congestion level in the nodes. In this approach, all data
layer, reliable and light-weight transmission control at the sources are assigned by an equal priority index that helps to
transport layer, and a middle-ware technique to integrate the mitigate congestion. In this approach, if congestion is detected
wireless personal area network application subnet with the within the network then it significantly increases data
Internet backbone. In this paper, we propose a congestion-free transmission latency.
data forwarding method at the network layer. Zhuang et al. [19] proposed a congestion control mechanize
The rest of the paper is organized as follows. Section II for data gathering in sensor network called Congestion-
introduces related works. Section III presents the network Adaptive Data Collection scheme (CADC). In this paper, an
model and problem formulation. Section IV gives an adaptive lossy compression based mechanism is developed to
analytical description of the proposed congestion control mitigate congestion. Furthermore, a weighted CADC scheme
mechanism. Theoretical analysis has presented in Section V. is applied to manage cyber-physical applications congestion.
Section VI presents experimental results and analysis and In CADC, data items are defined by the different priorities.
Section VII finally concludes this paper highlighting the main However, CADC does not focus on the energy consumption of
conclusions of the study. the deployed sensor nodes. Therefore, this approach suffers
from a poor network lifetime.
II. RELATED WORKS It has been seen from the literature that the existing
Congestion control is a primary objective in several multi- congestion control method leads to a large number of message
hop WSNs especially in healthcare applications where exchanges over the network for data and status exchange. It
deployed sensor nodes continually sense data from the increases data transmission delay within the network and
monitoring patients and forward this information to the Base energy consumption of the deployed sensor nodes. Also, the
Station (BS) to satisfy certain requirements of the medical quality of services (QoS) of the existing approaches is very
staff. This section provides an overview of the most recent poor. Due to poor performance and high transmission delay of
existing research works related to congestion control in the existing set of approaches, it is very challenging to
WSNs. implement in IoT-based healthcare applications. In this paper,
a distributed level based congestion control mechanism has

0098-3063 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Otago. Downloaded on May 31,2020 at 10:01:34 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCE.2020.2987433, IEEE
Transactions on Consumer Electronics
3

been proposed for effectively classifying the data packets X=(x1, x2,…., xn)T. X presents the sending rate of all deployed
which help in congestion avoidance. Furthermore, a level- nodes, which varies with time. In the IoT network, when
based data routing scheme is used for making a more suitable traffic demand exceeds the network’s bandwidth, data packets
decision to implement an optimal value in the route will accumulate in the cache of nodes, which will cause
identification process. The existing congestion control routing congestion. This means that when node vi sends data to vj, if
protocols which were proposed for routing in WSNs are not xi  x j , then the queue length of cache in the node vj will
able to provide optimal energy consumption solution.
gradually increase, the network congestion occurs. QH, QL,
Therefore, the proposed model considered the sensors present
and QC are the specific queues used for storing different types
in the IoT scenario. Hence, the proposed congestion control
of medical data in relay nodes.
data routing scheme provides optimal results by reducing the
The congestion index helps in identifying the level of
energy consumption and delay. The proposed scheme is
congestion in the IoT network and is given by the following
increasing the packet delivery ratio by avoiding congestion in
equation:
the network. This is achieved by the effective monitoring of
Ci  i, jn,vi , jV ( x j  xi )  Qi (3)
different queues. It helps to avoid packet drops occurring due
to congestion. Finally, it is noted that the proposed scheme Where,
gives optimal results in terms of QoS in IoT-based healthcare 1 Length (Q)  threshold
applications. Qi  
0 Length (Q)  threshold
III. NETWORK MODEL AND PROBLEM FORMULATION The objective of this paper is to estimate the congestion
during data gathering process, energy consumption and
A. Network Model routing delay for an IoT-based healthcare network, to provide
We consider an IoT-based healthcare network consisting of an important guideline for network optimization, such as
N numbers of static sensor nodes arbitrarily located in the routing design and network lifetime. The entire network
hospital environment. In the initial network topology, each lifetime is divided into p+1 stages [S0 , S1, S 2 ,....,S p1 , S p ] ,
node has transmission range Rmax and deployed network where Si represents the ith network stage, e.g. the first sensor
topology is represented with an undirected weighted graph G= node of the IoT network dies at the end of stage S0 and the
(V, E), where V  {v1 , v2 ,...., vN } is the set of nodes and network is totally out of work at the stage Sp. The duration at
E  {(vi , v j ) dis(vi , v j )  Rmax } is the set of edges, where each stage, namely the number of data phases at each stage Si,
is denoted by [a0 , a1 , a2 ,.....,a p1 , a p ] . Thus, is the IoT
dis(vi , v j ) shows the distance between nodes vi and vj. The
network lifetime from network initialization until the first
data transmission links between any two nodes do not have the node dies. The average traffic load of node vi in a data round
same capacity and edge (vi , v j )  E is expected to have a
of each stage is denoted by [ti(0) , ti(1) , ti(2) ,......,ti( p 1) , tip ] .
stipulated capacity Cvi ,v j to transmit packets from node vi to Specifically, we present our objectives as follows.
node vj. Each deployed node (vi V ) periodically sends the 1. For an IoT based healthcare network, each deployed sensor
node effectively estimates congestion, energy consumption
sensed data to the BS through multi-hop communication. The
and routing delay according to Ci and selects an energy-
first order radio model is used for energy consumption [20-
efficient query processing/data gathering path.
22]. So the energy consumed while transmitting a -bit 2. The average traffic load and energy consumption of the
message to a receiver node located d distance apart can be deployed sensor nodes at each stage, i.e., for each 0  i  n,
formulated as the equation,
ti( j ) , ei( j ) should be minimum. Then, we can mitigate energy

 Eelec   fs d
2 d  do
Et   (1) consumption problems in high traffic congestion.

 elec
E   d 2 d  do
 amp The average duration vector of the network stages
where Eelec denotes transmitting circuit loss and d0 is the [a0 , a1, a2 ,......,a p 1, a p ] should be minimum based on Ci.
threshold distance.  fs and  amp denote the energy for power Then, we can reduce packet loss in high traffic congestion.
amplification in the models, respectively. The energy
consumption for receiving an –bit message depends on the IV. PROPOSED CONGESTION CONTROL MECHANISM
dissipation in the circuit. Therefore, The main objective of the proposed scheme is to control
Er  Eelec (2) congestion in IoT-based healthcare aware WSNs which can
For idle listening, the energy consumption rate of the sensor potentially increase the network’s lifetime. The proposed
scheme also considers the fairer delivery of sensing
nodes is denoted by Eidle .
information to the gateway. To achieve these objectives, the
B. Problem Formulation proposed scheme considers two main parameters energy and
Let xi is the packet sending rate of vi and if xi is detected as delay. Furthermore, it considers two types of traffic: sensitive
zero then vi is out of work. Therefore, Gx=(V, E, X) graph and non-sensitive. Sensitive traffic is designed to transfer high
simulate a topology of sensor-based IoT network, in which priority data and non-sensitive traffic is designed to transfer

0098-3063 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Otago. Downloaded on May 31,2020 at 10:01:34 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCE.2020.2987433, IEEE
Transactions on Consumer Electronics
4

normal data. sensor nodes also identify their single-hop child nodes and
The proposed congestion control scheme consists of three their current energy condition. After level detection, all sensor
phases: 1) setup phase, 2) request distribution phase and 3) nodes transmit their update level value, ID, location
data routing and event occurrence report phase. In healthcare information and current energy condition to gateway node
applications, due to different stages of vital signs and different through their intermediate parent nodes. Due to mobility,
priorities of several IoT based medical devices, the proposed sensor nodes may change their current locations. If the sensor
scheme forwards sensed information to the gateway node with node moves from one level to another level, the mobile node
different priorities. Before packets transmission, the source updates its current level value with the help of neighbor nodes.
node sets a priority according to the importance of the data. The detailed description of the process is summarized in
Intermediate nodes route these packets according to the algorithm 1.
priority of the data packets.
B. Query Distribution Phase
A. Setup Phase In this phase, when the gateway node receives a query from
The setup phase runs only once, during the initialization of the medical staff (doctor/nurse), the gateway node distributes
the network. In this phase, after deployment each node detects received requirements among deployed nodes to fulfill the
its single-hop neighbor nodes and deployed nodes are also need of the medical staff. In healthcare applications, the type
divided into different levels. For level detection, initially of data is very important during the data gathering process. In
gateway sets its level value once i.e. L(Gateway)=1 and sends some situations, health parameters may include highly
a request message LEVEL to the sensor nodes in the range sensitive information. During the query distribution phase,
Rmax. gateway identifies level and location information of the sensor
Algorithm 1: Level Detection Algorithm node from the earlier record and calculates the lifetime of a
query packet within the network. The lifetime of the query
Input: Deployed sensor nodes
Output: A level value has allotted to all sensor nodes packet can potentially increase the reliability of data
1. Initialize: L(Gateway)=1;// L(Gateway) initialized to one transmission and reduces message overhead within the
2. Gateway broadcasts a LEVEL msg. in the range Rmax. network. The lifetime of the query packet is calculated as
3. for each sensor node vi
4. if (D(Gateway,vi) ≤ 2Rmax) then follows
5. L (s)= L(Gateway)+1; d (level )
6.
7.
PN(vi)=Gateway;
end if
LLife _ Packet   1
(TT  TR ) * Pt (4)
8. end for
9. Similarly, node vi broadcasts a MOD_LEVEL msg. in the range 2Rmax where d(level) is the level value of the destination node. TT is
10. if (node vj receives the msg. and L(vj)>L(vi)) then the transmission delay and TR is the receiving delay. Pt is the
11. L(vj)=L(vi)+1; processing delay.
12. PN(vj)=vi
13. CN(vj) =vi. Then the gateway broadcasts an RREQ message within the
14. else level one sensor nodes. The message contains a query for
15. Discord the msg. radio links capacity, ID of the destination node, level (L) value
16. end if
17. for each sensor node vi of the destination node, lifetime of the request (LLife_Packet), and
18. vi sends its current level value, energy condition, ID, Location location of the destination node. To obtain an estimation of
information; these capacities, every node, vi, assesses radio link related it to
19. end for
its parent, vp (v p  PN (vi ) , by sending a burst of packets
The message contains its ID, level (L) and location during a determined amount of time, T. Every packet is sent
information. When a sensor node vi receives the LEVEL after the reception of the previous packet’s acknowledgment
message then vi sets its level value to one higher than the or after a time-out from the former submission. The link
gateway, i.e. L(vi)= L(Gateway)+1 and also sets the gateway capacity, Cvi ,v p , is then approximated by the total number of
as its parent node (PN) i.e. PN(vi)=Gateway. Similarly, all
acknowledged packets divided by the period T. When a node
sensor nodes within the range 2Rmax to the gateway node
vi receives RREQ message, it checks destination node ID. If
increase their level value to one higher than the gateway node
request node ID is the same with vi, vi sends its update
and set the gateway as their parent node. Recursively, node vi
information to the gateway through the highest capacity link.
broadcasts a MOD_LEVEL message to the deployed sensor
Otherwise, it checks the level value of the destination nodes
nodes in the range 2Rmax. The message contains its ID, L(v1),
and lifetime of the request packet. If the level of the node vi is
current energy condition (Ecurrent), and location information. If
equal or less than the level of the destination node and lifetime
a sensor node vj receives the MOD_LEVEL message and its
of the request packet is expired, it simply discards the
level is less than the level value of the node vi, it sets as one of
message. Otherwise, it forwards the RREQ message within the
the child nodes i.e. CN(vj)=vi. Otherwise, it updates its level to
child nodes. In healthcare applications, the gateway may
one more than the level value of the node vi and sets it as one
receive multiple requests at the same time for the same or
of the parent nodes, i.e., PN(vj)= vi. In the setup phase, all
multiple IoT based medical devices. If gateway executes these
deployed nodes detect their single hop parent node-set and
requests at the same time then congestion may occur within
their current energy condition. Recursively, all deployed

0098-3063 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Otago. Downloaded on May 31,2020 at 10:01:34 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCE.2020.2987433, IEEE
Transactions on Consumer Electronics
5

the network. the average residual energy of the parent sensor nodes, using
Algorithm 2: Request Distribution Algorithm the formula
Input: Query from the doctors and nurses
p
Output: Respond of the query
1. Gateway receives a query from the doctors and nurses
 (vk )   ECurrent(vi ) p (5)
i 1
2. Start answering process according to request priority
3. if required data is not available then where p is the number of nodes in the PN(vk) and Ecurrent is
4. Gateway identifies level and node ID of the vj node
current energy condition. Let S={s1, s2, …, sp} be the parent
5. Calculated LLife_Packet using eqn. 1
6. Set life time of the query packet nodes that residual energy is greater than or equal to  (vk ) .
7. Broadcast RREQ msg. within the single-hop one node Then vk sends all data packets to the parent node concerning
8. for each node vi
9. if node ID(vi) = ID(vj) then the buffer status of the parent sensor node (which is specified
10. Send requested data to the gateway by congestion index Ci) and distance of the parent sensor
11. end if node. If the buffer status of the nearest parent node is
12. if L(vi) ≠ L(vj) then overflow, vk selects the next nearest parent from S. The
13. Node vi forwards RREQ msg. within the child nodes
14. end if
pseudo-code of the routing algorithm is described as follows.
15. if LLife_Packrt=0 and L(vi) < L(vj) then
Algorithm 3: Event Distribution Algorithm
16. Discord the message
Input: Event identifies by the sensor node
17. end if
Output: Event report receive at BS/Gateway
18. end for
1. for each sensor node vk
2. for each sensor vi belongs to PN(vk)
3. sum=sum + Ecurrent (vi)
Therefore, the proposed scheme identifies the priority of the
4. end for
query according to the priority of the patient. The proposed
5.  (vk )  sum / PN (vk )
scheme also considers the request time to specify the
6. for each sensor node vi belongs to PN(vk)
execution of the query for a part node. For example, in
7. if (Ecurrent (vi) ≥  (vk )
healthcare applications, vital signs related to sensitive
information (such as the heart rate, breathing condition, and 8. S=union (S,v1)
9. end if
blood sugar) have high priority; therefore this type of request 10. end for
is assigned a high priority. Also proposed scheme can consider 11. for each sensor node vl belongs to S
other types of traffics for data related to non-sensitive (such as 12. if (MAX> Ecurrent (vk) && min D(vl,vk) then
legs sensors) which has a lower priority. In the proposed 13. MAX=Ecurrent(vk)
14. Node vl selects data routing.
scheme, when the gateway node receives a query from the 15. end if
medical staff, gateway sets query dissemination time 16. end for
concerning the current time. On the other hand, each sensor
node attached to the patient’s body should report to the D. Congestion Control Mechanism
gateway when it detects vital signs out of the normal range. In
The main objective of this work is to provide a routing and
that situation, the source node sets high priority during the
congestion management mechanism for IoT-based healthcare
vital sign transmission. The detail description of the query
network. In this work, deployed nodes select their data routing
distribution process is summarized in the algorithm 2.
path in such a way that the congestion occurrence probability
C. Data Routing and Event Report Phase will be reduced. The congestion control phase is executed with
After the request dissemination phase, the destination node the data routing algorithm.
sends its update information to the BS/gateway node through The proposed congestion control scheme uses a classifier to
the energy-efficient highest capacity link. Similarly, if a node classify different types of data and route them into different
senses a medical emergency based on its duty, it will report to queues. All data packets are classified into three categories: a)
the gateway according to the specifications. The reports must high priority data packets, b) low priority data packets, and c)
have the required parameter values so that the gateway can control data packets. Each data packet contains its type in the
show the proper reaction and sends a medical emergency packet header. Classifier classifies received data packets and
report to the doctor/nurse. sends them into different queues concerning their class. The
In this phase, the information related to the occurring proposed scheme uses a priority queue based scheduler to
medical emergency is sent to the gateway. For this purpose, transmit different data. If class 1 data is found within the
the node creates a packet containing the information related to queue, scheduler stop class 2, class 3 data transmission and
the sensed event and sends it to the nearest parent node start class 1 data transmission. After transmission of all class 1
through the highest capacity link. According to the setup data packets, the scheduler starts class 2 and class 3 data
phase, a sensor node may have multiple parent nodes and packets transmission.
multiple paths to the gateway. Let {v1, v2, v3,…..vp} are the set If the data transmission rate Txmax is less than the data
of nodes belonging to set PN(vk). For sending a medical
receiving rate ic1 Rxi , the buffer of a node will overflow after
emergency packet or a normal data packet, node vk calculates
some interval time. In this situation, the node does not receive

0098-3063 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Otago. Downloaded on May 31,2020 at 10:01:34 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCE.2020.2987433, IEEE
Transactions on Consumer Electronics
6

any data from its child node and the network suffers from the Proof: Since node vi is the small region of Ad, its traffic load
packet loss problem. To solve this problem, the proposed can be calculated as the average traffic load in Ad according to
scheme selects an alternative path when the buffer level our analytic model. Therefore, we first calculate the average
reaches the threshold value (which is identified by Ci). When a traffic load in Ad. σ is the width of Ad and σ denote the angle
node receives a data packet from the child node then it sends formed by Ad and the sink/base station; thus, we can obtain the
buffer status with the ACK message. If a node v1 detects that number of nodes in Ad. The number of nodes in Ad is
class 1 queue has reached the threshold value, it sends d , if d 
N Ad   (7)
  ,
H_AlterPath message to all child nodes so that the high 2
otherwise
priority data is sent to another immediate energy-efficient As these nodes receive and forward the data from the lower
shortest path. When a node vk receives an alternative path level regions, the number of nodes in the upper level Ad+ir, is
selection message from the parent node then vk selects the next  (d  ir ) | 0  i  z1, if d  
nearest energy level node from the parent node set PN(vk) and N Ad ir  
sends all high priority data from this parent node. Similarly, ( / 2  ir ) | 0  i  z 2 , otherwise (8)
when node vk detects it is class 2 queue and class 3 queue have where z1  ( R  r ) / r and z 2  ( R   ) / r
overflows then it sends L_AlterPath and C_AlterPath message Since each node only generates a data packet per round with
to all child nodes. Child nodes also select an alternative path respect to one query, the number of data packets equals to the
for class 2 and class 3 data packet transmission. The threshold number of the involved nodes. Thus, the number of data
value for the alternative path selection process depends on the packets sent by Ad is
number of child nodes hold by a parent node and receives the
data rate of the node. In this work, when a parent node detects DAd  N Ad  N Ad r  ..... N Ad  zr
(9)
that its queue is 95% full by received data packets, then it
sends an alternative path selection request to the child nodes. Based on the above equation, we have the average traffic load
After some time when node vk again senses that its buffer is of Ad as D Ad N Ad . Since the traffic load of the node vi
less than the threshold value then also it sends REQ 1, REQ 2, approximately equals the average traffic load of the sensor
and REQ 3 message to all child nodes for resending class 1, nodes in Ad for data packet transmission, the traffic load of the
class 2, and class 3 data packets. node vi at S0 should be ti(0)  DAd N Ad . With some simple
calculation, we have ti(0) as (6).
V. THEORETICAL ANALYSIS OF PROPOSED CONGESTION
CONTROL MECHANISM Theorem 2: Assume that vi is in the small region of Ad with
the width of σ. Denote b0 as the distance between Ad and the
This section examines the complexity of the proposed
boundary node. If the sink node forwards one query packet for
congestion control algorithm to provide a theoretical analysis
each node per round, the average query sent by vi in a round at
of whether the proposed solutions are applicable in real-world
S0 is
healthcare practice.
 z (1  z1 )r
Theorem 1: The message complexity of the proposed  ( z1  1)  1 if b0  
 2b0
congestion control algorithm in O (N) where N is the number ti( 0)   otherwise
 1 ( z  2) 2  1 z ( z  1)r
q

deployed IoT-enabled sensor nodes. (10)



2
2
2
2 1

Proof: The total number of messages exchanged by the


where z1  ( R  r ) / r and z2  ( R   ) / r
deployed IoT based sensor nodes over the network for
executing the proposed congestion control scheme is known as Proof: Similar to the proof of Theorem 1.
message complexity. In the setup phase, the proposed The average traffic load of node vi due to query distribution
congestion control scheme needs one message exchange over and data gathering in a data round of each stage is
the network that means each node vi broadcast a single ti( 0)  (ti(d0)  ti(q0) ).
message to their neighbors. Based on the received data each
sensor node vi calculates its level value and parent node. Theorem 3: Denote  r as the time period for one query
Therefore, the proposed congestion control algorithm needs processing round i.e. query transmission and data gathering
maximum O (N) message exchange over the network. round. Node vi is in the region Ad, where d is the distance
Theorem 1: Assume that vi is in the small region of Ad with between Ad and the sink. If the data transmission rate of the
the width of . Denote d as the distance between Ad and the sensor node is G bits/s, in a data round, the average energy
base station. If each node receives one query packet per round, consumption ei0 of vi is ei0  ei(,0r)  ei(,0t )  ei(,0j)  ei(,0q) , where
the average data sent by vi in a round at S0 is
ti( 0) e (0) t
(0)
Eelec  (t d(0)  1)( Eelec   k k  )
D  i,q d
 z1 (1  z1 )r  (0) (0)
if d   e  (t  1)Eelec
( z1  1)  ,  i,r d
 2 d e (0)  t  ( Eelec   k k  )
(0)
1 1 otherwise (6)  i,t d
 ( z 2  2) 2  z 2 ( z1  1)r ,  (0)
 2 2 2t  
e (0)  Eidlem
(0)
 Eidle (m  d  )
where z1  ( R  r ) / r and z2  ( R   ) / r 
 i, j d, j B B

0098-3063 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Otago. Downloaded on May 31,2020 at 10:01:34 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCE.2020.2987433, IEEE
Transactions on Consumer Electronics
7

And if d  r , k  r; otherwise, k  d , and if k  k0 , ki  E0  Euse


i ,k
 E0  ic0 (ek(i ) .t (i ) ) and the remaining energy
 k   fs and α=2; otherwise,  k   amp and α=4.
of the network is  (i )  nE0  Euse
i
 nE0  ic 0 nk 0 (ek(i ) .t (i ) ) .
Proof: In a data round, the energy consumption of node vi Theorem 5: The estimated Query Processing Delay (QPDi)
consists of the following four parts.
for sensor node vi is 2  dist(vi  v j ) c  2 i where,  i is the
1) Energy consumption for query distribution: Since node vi processing time.
in the region Ad, the received queries in a round is tj(0) and
send queries (ti(0)  1) . Thus, the energy consumption for Proof: Estimated query processing delay is the approximate
time required by a sensor node vi to transmit its query as well
query distribution is as data to all its surrounding neighboring nodes which are
ei(,0q)  td(0)Eelec  (td(0)  1)( Eelec   k k  ) . coming under its transmission range r. The estimated query
2) Energy consumption for data receiving: Since node vi is processing delay QPDi can be defined as
in the region Ad, the received data amount in a round is 
QPDi  2  max(QPDi, j   i ), vi  Negi }
(ti(0)  1) , according to theorem 1. Thus, the energy
Where QPDi,j i.e. estimated transmission time between the
consumption for receiving is ei(,0r)  (t (j0)  1)Eelec .
sensor vi and vj and is the processing delay of vi. The QPDi,j
3) Energy consumption for data transmitting: Since the data can be calculated as given in
amount sent by vi in a round is tj(0), the energy
consumption for data transmitting is QPDi, j  dist(vi , v j ) c
 ei(,0t )  ti(0) ( Eelec   fs k 2 ),
 k  k0 where c is the speed of light.
 ( 0)
ei , t  ti  ( Eelec   ampk ),
( 0) 4
 otherwise
We define the estimation error Ek at node vi equals the
If d  r, k  r; otherwise k=d. That is because the accumulated errors from each of its children
transmission distance is r if d≥r ; otherwise, it is d. v1 , v2 ,....,vk ,...,vN A :
D
4) Energy consumption for idle listening: According to the
network model, the duty cycle is γ. Thus, the active time N AD

per round is ba  Br  . The energy consumption for idle Ev   (  iv   i ) 2   (  iv   i ) 2  ......  (  iv   i ) 2


i 1 i v1 i v N A
D
listening is the multiplication of Eidle and the duration in
idle listening. Since the duration in idle listening, denoted where  i denotes the data measured by node i and  iv
by bi(,0j) , is the active time excluding the time for data denotes the value of  i received by node vi. For theorem 1 we
transmitting and receiving, we have can easily calculate the number of nodes in Ad is N AD . The
(ti(0) 1) estimated error Ev at node vi equals the accumulated errors
bi(,0j)  ba   ti(0) / G . Therefore, we derive the from each of its children. We define the error contribution of
G
energy consumption for idle listening as vi’s child node vk as
N AD N AD
2t  
( 0)
Cvk   (  iv   i )   ((  iv   ivk )  (  ivk   i )) 2
ei(,0j)  Eidlebi(,0j)  Eidle (ba  d  To summarize, in i 1 i 1
G G
(0)
a round, the energy consumption ej of node vi is N AD N AD N AD
ei(0)  ei(,0r)  ei(,0t )  ei(,0q)  ei(,0j) .   (  iv   ivk ) 2   (  ivk   i ) 2  2  (  iv   ivk )(  ivk   i )
i 1 i 1 i 1
N AD N AD
Theorem 4: For each network stage Si, the reaming energy of   (  iv   ivk ) 2   (  ivk   i ) 2
i 1 i 1
any node vk after Si is ki  E0  ic 0 (ek(i ) .t (i ) ) , and the
remaining energy of the network after Si is N AD N AD
 2  (  iu   ivk ) 2 .  (  ivk   i ) 2  Dvk  Evk  2 Dvk .Evk
ki  n( E0  ic  0 (ek(i ) .t (i ) )) . i 1 i 1
(i)
Proof: Since ek is the average energy consumption of node vk where Ev k is the estimation error at child vk and Dv k is data
in a data round (combination of query distribution and data
distortion due to data compression of vk.
collection) at Si and t(i) is the duration of Si, we can determine
the energy consumption of node vk during Si as Ek ,i  ek(i ) .t (i ) . VI. RESULTS AND DISCUSSION
From S0 to Si, the total energy consumption of node vk is
A. Simulation Environments and Simulation Results
i, k
Euse  ic  0 Ei , k and the total energy consumption of the
We performed extensive experiments on the proposed
i
network is Euse  ic  0 nk  0 Ei, k . As the initial energy of the scheme using NS-2 [23]. The medical sensor nodes are
network in nE0, after Si, the remaining energy of node vk is randomly distributed in 2D space. An IoT-enabled
gateway/BS was used for data gathering and queries
processing. The proposed congestion control mechanism for

0098-3063 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Otago. Downloaded on May 31,2020 at 10:01:34 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCE.2020.2987433, IEEE
Transactions on Consumer Electronics
8

100
IoT-based healthcare aware network is evaluated and 100
Proposed Scheme Proposed Scheme
95
95 BRCCTP
compared with the existing well-known TARA [8], REEP BRCCTP
90 CADC
90 CADC

Average Throughput(pkts/s)
[11], HTAP [17], BRCCTP [18], and CADC [19] schemes in

Average Throughput(pkts/s)
TARA TARA
85
85
HTAP HTAP
the literature in terms of percentage of the successfully 80 REEP 80 REEP
received packet, average throughput, average hop-by-hop 75
75

delay. Furthermore, the energy-saving performance of the 70


70

proposed algorithm and the network lifetime was evaluated 65


65

60
where the initial energy of each sensor node is E0=0.5 60
(b)
(a) 55
[Joules]. All the simulations are based on collision-free MAC 55
50
protocol. Simulation parameters are listed in Table 1. 10 20 30 40 50 60 10 20 30 40 50 60
Sources Data Rate (pkts/s) Sources Data Rate (pkts/s)
TABLE I
SIMULATION PARAMETERS Fig.2. Comparison of average throughput. a) 20 nodes, b) 100 nodes.
Parameter Value
Fig. 2a and 2b present the throughput of the proposed
Deployment area 100×100 m2
scheme using different percentages of sensor nodes in the
Number of nodes 20-100
Data packet size 500 bits
network. This is measured with the rate of data that reaches
Initial energy of each node (E0) 0.5 Joules the gateway node. Fig. 2a indicates that the proposed scheme
50nJ/bit improves throughput up to 29% comparing with BRCCTP, up
0.0013 pJ/bit/m4 to 31% comparing with CADC, up to 33% comparing with
10pJ/bit/m2
REEP, up to 34.33% comparing with TARA and up to 36%
Control message size 100 bits comparing with HTAP. Fig. 2b shows that the proposed
Duty cycle 10% scheme improves throughput up to up to 26.5% comparing
Duration of a data period 10s with BRCCTP, up to 29% comparing with CADC, up to
Energy consumption rate for 0.88 mJ/s 31.21% comparing with REEP, up to 32% comparing with
ideal listening TARA and up to 33.45 % comparing with HTAP. It is due to
Data transmission rate 512 Kb/s the level-based congestion control mechanism where data
Sensing radius 20m transmission congestion is controlled through the effective
Antenna type Omni Antenna alternative data routing path selection process.
Max packet in queue 50 3.8 4.2
3.6 Proposed Scheme 4.0 Proposed Scheme
3.4 BRCCTP 3.8
BRCCTP
3.2 3.6

Average hop-by-hop delay (ms)


CADC CADC
Average hop-by-hop delay (ms)

3.0 3.4
100 2.8
TARA 3.2 TARA
100
2.6 HTAP 3.0 HTAP
2.8
90 90 2.4 REEP REEP
2.6
Average Received Pkts Ratio (%)

2.2
Average Received Pkts Ratio (%)

2.4
80 2.0
80 2.2
1.8 2.0
70 1.6 1.8
70 1.4 1.6
60 1.2 1.4
60 1.0 1.2
Proposed Scheme
BRCCTP
50
Proposed Scheme 0.8
0.6
(a) 1.0
0.8
(b)
50 CADC 40
BRCCTP 0.6
0.4
TARA CADC 10 20 30 40 50 60 10 20 30 40 50 60
40 HTAP 30 TARA
Sources Data Rate (pkts/s) Sources Data Rate (pkts/s)
REEP (a) 20
HTAP (b)
30 REEP
10 20 30 40 50 60
10 20 30 40 50 60
Fig.3. Distribution of average hop-by-hop delay according to the sources data
Sources Dta Rate (pkts/s)
Sources Dta Rate (pkts/s) rate. a) 20 nodes, and b) 100 nodes.
Fig.1. Percentage of successfully received packets. a) 20 nodes, b) 100 nodes.
Fig. 3a and 3b present average hop-by-hop delay results
using the tested algorithms in scenario 1 (20 nodes) and
Fig. 1a and 1b present the percentage of successfully scenario 2 (100 nodes). This performance metric is important
received packets of the proposed scheme. This is measured for how the proposed scheme controls inter-path interferences,
with the number of data packets received at the gateway. Fig. link-layer retransmissions, and overhead introduced by control
1a shows that the percentage of successfully received packets packet exchanges. In Fig. 3a, it is seen that the proposed
of the proposed scheme is more by 37% as compared with scheme has a less average hop-by-hop delay of 31% as
BRCCTP, 39.4% as compared with CADC, 41.3% as compared to BRCCTP, 33.6% as compared to CADC, 35.4%
compared with the REEP scheme, 42.5% as compared with as compared to REEP, 38.5 % as compared with HTAP, and
the HTAP scheme, and 57% as compared with the TARA 41% as compared with TARA algorithm. Fig. 3b indicates that
algorithm. Also, Fig. 1b indicates that the proposed scheme the proposed scheme reduces average hop-by-hop delay up to
increases the percentage of successfully received packets up to 28% comparing with BRCCTP, up to 31.3% comparing with
38% comparing with BRCCTP, up to 42% comparing with CADC, up to 32% comparing with REEP, up to 35%
CADC, up to 44 % comparing with REEP, up to 45.7% comparing with HTAP, and up to 37% comparing with TARA
comparing with HTAP and up to 59% comparing with TARA. algorithm. It is since the proposed scheme efficiently selects
This causes due to the level-wise elimination of data the data routing path with the minimum number of control
congestion during the data routing process and priority-based packet exchanges. Therefore, the average data transmission
congestion control mechanism. delay is less in the proposed scheme.

0098-3063 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Otago. Downloaded on May 31,2020 at 10:01:34 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCE.2020.2987433, IEEE
Transactions on Consumer Electronics
9

100 100 100


Proposed Scheme 100
95 95 Proposed Scheme Proposed Scheme
95 95
90 BRCCTP 90 BRCCTP BRCCTP
90

Average Received Pkts Ratio (%)


85 CADC 85 CADC 90 CADC

Average Throughput(pkts/s)
80 85
80 TARA TARA 85 TARA
75 80
75 HTAP HTAP HTAP
70 75 80
REEP REEP REEP
Life Time(s)

Life Time(s)
70
65 70
65 75
60 65
60 55 60 70
Proposed Scheme
55 50 55 BRCCTP 65
50 45 50 CADC
45 40 45 60
TARA
35
40
30
40 HTAP (a) 55 (b)
35
35
(a) 25
(b) 30
REEP
50
30 10 20 30 40 50 60
100 200 300 400 10 20 30 40 50 60
100 200 300 400
Sources Data Rate (pkts/s) Sources Data Rate (pkts/s)
Traffic Load (kb/s) Traffic Load (kb/s)
Fig.4. Comparison of Lifetime over traffic load. a) 20 nodes and b) 100 nodes. Fig.5. Comparison of real testbed results. a) Percentage of successfully
received packets and b) Average throughput in different data rate.
In Fig.4a and 4b, the energy performance of the proposed
scheme is measured compared with BRCCTP, CADC, TARA, Fig. 5b illustrates that our proposed scheme has an
HTAP, and REEP scheme from the view of traffic load. In increased average throughput compared to BRCCTP, CADC,
Fig. 8a, it is seen that the proposed scheme has more network TARA, HTAP, REEP schemes, verifying the theoretical case
lifetime of 37% as compared to BRCCTP, 40% as compared in Fig. 2. From the figures, it is clear that the proposed scheme
to CADC, 43% as compared to TARA, 42.3% as compared to demonstrates desirable performance for the congestion control
HTAP, and 45 % as compared to REEP. Also, Fig. 4b mechanism in an IoT-based healthcare system.
indicates that the proposed scheme increases network lifetime 4.0 100
3.8 Proposed Scheme
up to 38.5% comparing with BRCCTP, up to 42% comparing 3.6
BRCCTP
95
90
3.4
with CADC, 44.91% comparing with TARA, up to 43.78% Average hop-by-hop delay (ms) 3.2 CADC 85
3.0 TARA
comparing with HTAP, and up to 48% comparing with REEP. 2.8 HTAP
80
75
2.6
The lifetime of the network is measured with the first dead REEP 70

Life Time(s)
2.4
2.2 65
node. The lifetime of the proposed scheme is more since it 2.0
60
1.8
energy efficient as compared to BRCCTP, CADC, TARA, 1.6 55 Proposed Scheme
1.4 50 BRCCTP
HTAP, REEP scheme. 1.2 45
CADC
1.0
TABLE II 0.8
40
TARA
PARAMETERS VALUES USED IN THE INDOOR TESTED EXPERIMENT 0.6 (a) 35
HTAP
0.4
Parameter Value
30
REEP (b)
10 20 30 40 50 60 25
Number of sensor nodes 20 Sources Data Rate (pkts/s)
100 200 300 400

Deployment area Traffic Load (kb/s)


15 15
Base Station/Gateway 1 Fig.6. Comparison of real testbed results. a) Average hop-by-hop delay and b)
Sending rate 1 packet/s Lifetime over traffic load.
Packet size 500 bits
Initial voltage of battery 1.6.0V Fig. 6a shows the average hop-by-hop delay in the practical
test. It can be seen that the average hop-by-hop delay has been
B. System Implementation and Testing reduced compared to BRCCTP, CADC, TARA, HTAP, and
REEP schemes. As described in the theoretical case, our
A testbed of the proposed congestion control scheme has been proposed scheme shows an improved result because of the
created for demonstration where twenty medical sensor nodes class-based congestion management mechanism.
are used for data collection. In this experimental setup, we Fig. 6b shows the lifetime over traffic load. As predicted by
used a blood pressure sensor, pulse oximetry sensor, ECG the simulation tests, it can be seen that the lifetime of the
sensor, EMG sensor for data collection. The testbed consists proposed scheme was increased when the traffic load was
of a gateway for query processing and data collection from the increased within the network. From this result, it can be
deployed medical sensor nodes. Twenty medical sensors are observed that our proposed scheme shows improved
deployed into two different laboratory rooms where each performance in an IoT-based healthcare system compared to
person is monitored by four medical sensors. The detailed the state-of-the-art congestion control approaches such as
parameters and corresponding values in real experiments are BRCCTP, CADC, TARA, HTAP, and REEP in the real
reviewed in Table 2. hardware experiments.
Fig. 5a shows that the result of the practical test is almost
similar to that of the simulation case in Fig. 1. As predicted by VII. CONCLUSION
the theoretical case in Fig.1, it can be seen that the percentage
This paper proposed a congestion control mechanism for
of successfully received packets of the proposed scheme was
wireless sensor networks applied in IoT-based healthcare
decreased when sources data rate increased. From these
scenarios. The proposed scheme used a priority-based data
results, it can be observed that our proposed scheme shows
routing strategy to control congestion within the network. It
improved performance compared to the existing congestion
classifies data packets into three different categories according
control approaches in the real IoT-based healthcare
to the priority of the data packets. Furthermore, the proposed
experiments.
scheme used a priority queue based scheduler to transmit

0098-3063 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Otago. Downloaded on May 31,2020 at 10:01:34 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCE.2020.2987433, IEEE
Transactions on Consumer Electronics

0098-3063 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Otago. Downloaded on May 31,2020 at 10:01:34 UTC from IEEE Xplore. Restrictions apply.

You might also like