Kuwelkar, 2023

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/368956159

RPL Protocol Enhancement using Artificial Neural Network (ANN) for IoT
Applications

Conference Paper · January 2023


DOI: 10.1109/IDCIoT56793.2023.10053540

CITATIONS READS

0 28

2 authors, including:

Sonia Kuwelkar
Goa College of Engineering
17 PUBLICATIONS 106 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Design of an Efficient RPL Objective Function for Internet of Things Applications View project

All content following this page was uploaded by Sonia Kuwelkar on 04 March 2023.

The user has requested enhancement of the downloaded file.


Proceedings of the International Conference on Intelligent Data Communication Technologies and Internet of Things (IDCIoT 2023)

RPL Protocol Enhancement using Artificial Neural


DVD Part Number: CFP23CV1-DVD; ISBN: 978-1-6654-7450-4

Network (ANN) for IoT Applications


Sonia Kuwelkar Hassanali G. Virani
E& TC Department, E& TC Department,
Goa College of Engineering Goa College of Engineering
Farmagudi, Goa, India Farmagudi, Goa, India
sonia@gec.ac.in virani@gec.ac.in

Abstract—In near future, IoT will revolutionize human the IETF is the 6LoWPAN specification drafted to provide
lifestyle. IoT is categorized as low power lossy network since it guidelines on carrying IPv6 datagrams over LLNs. The
employs devices with constrained power, memory and IETF constituted the Routing over Low Power Lossy
processing capability which are interconnected over lossy links. Networks (ROLL) working group and assigned task of
The efficiency of such networks largely depends on the design
of the routing protocol. To cater specific routing needs of such
devising a routing solution. The working group evaluated
networks, the IETF has proposed IPv6 routing protocol for the existing protocols like DSR, AODV, OLSR and OSPF.
LLNs (RPL) as a de facto routing standard. In RPL, routing They realized that these are inefficient for LLNs with
decision is based on a single parameter which leads to the regards to power, latency, overhead and reliability. After
selection of inefficient paths and affects network lifetime. This rigorous scrutiny, the IETF designed IPv6 routing protocol
work primarily focuses on improving the RPL protocol by for LLNs (RPL) and proposed it as de facto routing standard
overcoming the single metric limitation. In this work, a novel in such networks [5]. RPL provides IPv6 connectivity to
version of RPL is proposed which uses a Multilayer Feed battery operated wireless embedded devices which uses low
Forward Neural Network to make the routing decision based power radio to communicate. RPL builds upon popular
on multiple metrics. Four routing parameters namely, hop
count, delay, residual energy and link quality of candidate
routing protocols like CTP and Hydro used in the WSN
neighbors are fed as input to ANN in order to compute the domain, but is re-designed and extended to incorporate
fitness of each candidate and the one with highest value is IPv6. Over a period RPL has attained maturity and is
designated as the most suitable parent to route packets towards standardized as the routing protocol. The IP-stack of many
sink node. This technique lowers energy consumption by 15%, operating systems such as Contiki OS, TinyOS, T-Kernel,
improves Packet Delivery Ratio by 3%, lowers delay by 17% EyeOS, LiteOS and RIOT includes RPL as standard routing
and reduces the control overhead by 48% as compared to protocol.
standard RPL implementation. Although RPL was able to resolve major issues there
were still certain limitations which needed addressing. For
Keywords—IPv6, Low Power Lossy Networks, Routing protocol,
Artificial Neural Network (ANN), Internet of Things (IoT). instance, standard RPL makes the routing decision
considering a single metric which leads to selection of non-
I. INTRODUCTION optimal paths and thus reduces network lifetime [6]. In this
Billons of devices like energy and water meters, home paper, a novel version of RPL protocol is proposed, which
appliances, wearable devices, sensors and actuators are selects the best route based on multiple metrics. This
getting connected to the Internet to make a giant approach successfully overcomes the single metric
communications network termed as Internet of Things. It is limitation of standard RPL. The metrics Hop count (HC),
projected that by 2025 there will be more than 27 billion IoT ETX, Residual energy (RER) and End-to-end Delay are
connections [1]. IoT will transform the domain of weather chosen as the criteria for selecting the optimum parent. A
monitoring, smart cities, healthcare, logistics, industrial multilayer perceptron feed forward artificial neural network
automation, home automation and smart metering [23]. The is designed to make the multi criteria decision making. The
network here comprises of sensors which are battery four metrics of each candidate parent nodes are input to the
operated, use microcontrollers like the MSP430 and have feed forward neural network and the Quality values are
limited storage capacity. Unlike conventional networks, the computed by neural network. The candidate with peak
traffic in these networks can be one to one, one to many or quality value is chosen as the best parent in route towards
many to one. The communication network also has several the sink node. The paper is organized as follows: In section
drawbacks like instability, low data rates, high packet loss, 2 an overview of RPL protocol is provided. In section 3
limited transmission range, frame size limitation and related works done previously by researchers are discussed.
dynamically varying network topology. Besides these In section 4, the enhancements done to RPL using ANN are
aspects, the links connecting the devices can use discussed. Section 5 presents the results of simulations and
technologies like Bluetooth, Low-Power WIFI and wired detailed performance analysis. In last section, conclusions
PLC. Another aspect to be considered is that in large scale are discussed.
deployments, network can scale up to thousand nodes. Due II. RPL OVERVIEW
to these innumerable constraints on devices and underlying
communication technology, IoT is classified as a Low There are two types of routing protocols in Sensor
Power Lossy Network (LLN) [2][3][4]. All these unique networks: Reactive and Proactive. RPL belongs to the
requirements make the design of an efficient Routing proactive class of protocols. It is a distance-vector based
protocol a very challenging task. One major achievement by routing protocol. The sensor nodes direct their traffic
towards a single root or sink node. The root node is also

978-1-6654-7451-1/23/$31.00 ©2023 IEEE 51


Proceedings of the International Conference on Intelligent Data Communication Technologies and Internet of Things (IDCIoT 2023)
DVD Part Number: CFP23CV1-DVD; ISBN: 978-1-6654-7450-4

known as LLN Border Router (LBR) who acts as the 2. Destination Advertisement Object (DAO): They primarily
gateway for connecting the local network to the Internet. serve the purpose of propagating destination information in
Here, the network is constructed as a Destination Oriented up direction towards the DODAG root. This information in
Directed Acyclic Graph (DODAG). Whenever network is used for construction of downward routes between DODAG
being deployed, construction of DODAG is initiated by root and associated nodes.
Root node [8]. A node can join the DODAG by selecting a 3. DODAG Information Solicitation (DIS): Nodes willing to
preferred parent node. A DODAG is depicted in figure 1. join a DODAG but not received a DIO use these DIS
Each node is assigned a Rank, which is directly proportional messages to seek DIO’s from neighboring nodes.
to the distance from root. Within a DODAG, root node has 4. Destination Advertisement Object Acknowledgement
the lowest rank. A parent node will have a rank lower than (DAO-ACK): Upon receiving DAO messages, a node can
its children nodes. unicast a DAO-ACK to the DAO sender to acknowledge the
receipt of that DAO.
The root node initiates network formation by
broadcasting a DIO message. Whenever a node receives a
DIO, it can determine and join the RPL instance, add nodes
to the candidates list, choose best parent out of the
candidates and replace a previous parent. The DIO message
contains an Objective function along with other information.
The nodes select the most appropriate route by constraining
or optimizing routing metrics (parameters) specified by the
Objective function. Two Objective Functions, MRHOF and
OF0 are standardized in RPL. Both these are briefly
described below:
Fig 1: DODAG and Control Messages Minimum Rank Hysteresis Objective Function (MRHOF):
Uses the path cost or Expected transmission count (ETX) as
In a RPL based infrastructure, one or more RPL instances a routing metric to select the preferred parent in the
can exists, where each instance is identified by a unique ID DODAG [9]. ETX is the number of retransmissions required
(RPLInstance_ID). Each instance can have its own set of before a packet is successfully received at the destination.
metrics and routing policy independent of its neighboring MRHOF is designed with an aim to prevent frequent
instance. Multiple DODAGs can exists within a single RPL changes in the preferred parent.
instance. To distinguish the DODAG they are assigned a Objective Function Zero OF0 [10]. It depends on Hop Count
DODAG_ID. Every node can participate in multiple as a metric to make routing decisions. A candidate parent
DODAGs. In case of any inconsistency in the network, the having the lowest hop count, which implies that a node
sink can trigger DODAGs to reconstruct themselves. An closest to the root will be chosen as preferred parent.
incremental DODAG Version_Number is assigned to the Both these Objective functions have the shortcoming of
DODAG by the sink every time it reconstructs itself. The considering only single metric for routing decision. They
exact version of a particular DODAG can be identified by lack the holistic approach essential for providing QOS. To
the tuple of DODAG Version_Number, DODAG_ID and overcome this, RPL specification has provided the designer
RPLInstance_ID and. To make all this functional and share the flexibility to incorporate more metrics into the decision-
the routing sensitive information, RPL protocol specifies making process and we will see in later sections how this
four ICMPv6 control messages. These are: can be done.
1. DODAG Information Object (DIO): This message
comprises of crucial information like Objective function, III. RELATED WORK
rank, metrics, DODAG ID, trickle timer parameters and RPL
Researchers globally have taken keen interest in the RPL
Instance ID. This information helps a node to select a parent,
standard and have worked over the past years in upgrading
identify a RPL instance and join DODAG [5]. The DIO
its performance. Numerous shortcomings of the proposed
message body comprises of two parts, DODAG
standard have been identified and researchers have
configuration option and DAG metric container, where the
suggested routing improvisations by offering unique design
important information is embedded as shown below in fig 2.
approaches for load balancing, interference reduction [15],
These messages are essential for constructing upward routes
link estimation and neighbor table management [16]. In this
and maintaining the DODAG.
section, works which endorse evolutionary algorithms or
fuzzy logic to optimize RPL protocol are briefly described.
They employ multiple node and link metrics to overcome
the limitation of single metric in standard RPL. The authors
in [13] use firefly optimization algorithm to improve RPL.
This proposed version is termed as EEOPS-RPL. Here ETX
and Residual energy (RER) metrics are chosen as the
attractiveness parameters and the distance between nodes as
the movement parameter to select the optimal parent in the
DODAG. In [14] the authors have proposed a novel RPL-
Fig 2: DIO Message Format CGA which uses a chaotic genetic algorithm to search the

978-1-6654-7451-1/23/$31.00 ©2023 IEEE 52


Proceedings of the International Conference on Intelligent Data Communication Technologies and Internet of Things (IDCIoT 2023)
DVD Part Number: CFP23CV1-DVD; ISBN: 978-1-6654-7450-4

optimal solution for weighting factors of routing metrics. were tried and tested. But the one shown above gave better
This method gives low power consumption. accuracy and practicality of implementation. Table I shows
Kamgueu [17] proposed combining three routing metrics a list of metric values and the corresponding calculated
Delay, Energy and ETX with fuzzy logic technique to score by the ANN.
determine the best parent node. This is the pioneer work TABLE I: Metric values and computed Score by ANN
which suggested the use of Fuzzy logic for RPL ETX Delay HC RE Score- ANN
improvement since 2013. On similar lines, another work 3 200 3 225 87
[18] uses Fuzzy Logic model to include more than one 2 550 1 250 93
metric during routing and overcome the limitation of single 8 400 2 120 63
metric. Olfa Gaddour et al [19] have developed a holistic 3 400 1 205 86
objective function OF-FL which combines Hop count, End 2 600 1 220 88
to end Delay, Link Quality and Energy using fuzzy logic. 1 1500 3 190 77
Simulations are carried out comparing the performance of Several Operating Systems are proposed for motes to be
OF-FL with standard RPL objective functions for network deployed in IoT. A very popular and versatile is Contiki OS.
sizes up to 100 nodes. Better Network Lifetime, PDR and In this work the Contiki OS is used for implementation and
lower frequency of parent change is achieved. Hanane et al simulations are done with in-built simulator COOJA.
[20] have designed an objective function OF-EC, which The steps involved in best parent selection and routing
combines three metrics, namely ETX, Energy Consumption using the ANN based RPL implementation are as follows:
and Hop count using fuzzy logic and gives a composition 1. The values of metrics ETX, HC, RER and Delay are
metric based on which routing decisions are taken. This extracted from the received DIO message.
RPL implementation shows improvements in network 2. The DIO sender node is saved as candidate in the
convergence time, latency, routing overhead, energy candidate parent list.
consumption, Packet delivery ratio and network lifetime. In 3. If more than one parent exists in the list, then invoke the
[21] previous work, we have designed OF-FZ, wherein four neural network.
metrics are chosen and fuzzy logic is applied to improve the 4. Compute the scores for all the candidate parents.
routing. This work had design advantages over other fuzzy 5. Filter the candidate with maximum score and compare it
logic-based works and gave better performance. ETX, HC, to the score of existing parents.
RER and Delay are used to give higher PDR, lower latency, 6. If the candidate score is more than the score of the present
lower Power consumption and lower traffic overhead. In parent then select this candidate as the new best parent.
one study the authors [22] have projected a novel technique 7. Compute Rank of node proportional to the rank of the
to detect intrusion in RPL. Fuzzy logic is used to include new selected parent using Rank equation.
three metrics, namely Residual energy, distance, and 8. Send the DAO and DIO messages after expiry of
Expected Transmission Count (ETX). This method respective timers.
effectively detects local Repair attack with high rates of
True Positive and low False Positives.
IV. DESIGN OF IMPROVED RPL USING ARTIFICIAL NEURAL
NETWORK
This section describes the design steps of the proposed
novel Artificial Neural Network based RPL version. A
three-layer perceptron feed forward network is designed,
which will be applied to determine the optimal parent for a
node within the DODAG. The metrics Hop count (HC),
ETX, Residual energy (RER) and End-to-end Delay are
chosen as the criteria for selecting the optimum parent.
These four metrics are fed as input to the feed forward
neural network. The design consists of three layers of
neurons. In layer1 there are 10 neurons, layer2 constitutes of
Fig 3: Neural Network for Selecting Parent in RPL Network.
5 neurons and layer3 is the single output neuron. This is a
typical example of network for performing regression. The
network computes the SCORE for each candidate parent
depending on the values of ETX, HC, RER and Delay. This
network is trained using a database with 42,67,199 sample
values, where the dataset was split 70% and 30% for
training and validation. The gradient descent-based Delta
learning algorithm was used for training the network. This
network gives an accuracy of 96%. The network is shown
below in figure 3 and the scatter plot of actual vs predicted
is shown in figure 4. The design and training of the neural
network was done using tools like Google Collab and
Jupyter notebook. Several other neural network
configurations with different number of layers and neurons

978-1-6654-7451-1/23/$31.00 ©2023 IEEE 53


Proceedings of the International Conference on Intelligent Data Communication Technologies and Internet of Things (IDCIoT 2023)
DVD Part Number: CFP23CV1-DVD; ISBN: 978-1-6654-7450-4

Latency OF0 MRHOF ANN-RPL

1481
1479
2000

1374
1299

1254
1117

1107
1074

925
1000

Time ms
0
70 90 100
Node Density
a)
Packet Delivery Ratio
OF0 MRHOF ANN-RPL

98.97
Fig 4: Actual vs Predicted Scatter Plot for Trained NN

97.74
100

97.24

96.86
96.54

95.75
98
V. SIMULATION RESULTS AND ANALYSIS

94.37
94.24

94.21
PDR (%)
This novel ANN based objective function must be 96
incorporated in the operating systems meant for IoT motes. 94
Contiki is one such OS build for the wireless low power
sensor motes used in Internet of Things applications. 92
Contiki OS has an in-built TCP/IP stack with multitasking 90
flexibility. The memory requirement is around 30KB of 70 Node Density
90 100
ROM and 10KB of RAM. RPL is standardized as the
routing protocol in Contiki. The neural network b)
functionality is incorporated into Contiki OS for practical Average Power consumption (mW)
application of this version of RPL. A network with many OF0 MRHOF ANN-RPL

2.18
2.17

2.08
2.5

2.05
sender nodes and a single sink node in created in COOJA
1.87

1.782
1.90

1.87
1.674
environment and simulation are carried out. The test logs
2
generated during the simulation in COOJA are saved. The
Pavg (mW)

values of power consumption, packet delivery ratio, delay 1.5


and control overhead are obtained from the test logs by
writing PERL or Python scripts. The proposed ANN-RPL is 1
compared to standard RPL versions MRHOF and OF0
during simulations where number of nodes are varied from 0.5
70 to 100. A single sink node is used. Packets of length 128
0
bytes are transmitted at rate of 5 p/min. The simulation
70 90 100
environment used is shown below in table II. Node Density
c)
TABLE II : SIMULATION PARAMETERS
Control Overhead
Area of Deployment 400m * 400m OF0 MRHOF ANN-RPL
2000
1688

Simulator Contiki OS - COOJA


1463

Transmission Range 30m


1228
1121

1115
No of Packets

1500
1012

Number of Sender
70, 90, 100 static nodes
Nodes
827
743

1000
665

Number of Sink Nodes 1


MAC Layer Contiki MAC
500
Topology Random Deployment
Data Frequency 5 ppm and 10 ppm
0
Emulated Nodes WISMote 70 90 100
Transmission Ratio 100% Node Density
Simulation time 2 hours d)
Fig 5: Plots of a) Latency b) PDR c) Pavg d) Traffic Overhead for ANN-
RPL, OF0 and MRHOF at 5PPM

The simulations are run and graphs are plotted for


latency, PDR, Average power consumption and Control
Overhead at 5 ppm and 10 ppm are as shown in figure 5 and
6.
At density of 100 nodes and 5 ppm data rate, ANN-RPL has
a power consumption of 1.87 mw which is 15% lower as
compared to standard-RPL. The latency is 1254 ms which is

978-1-6654-7451-1/23/$31.00 ©2023 IEEE 54


Proceedings of the International Conference on Intelligent Data Communication Technologies and Internet of Things (IDCIoT 2023)
DVD Part Number: CFP23CV1-DVD; ISBN: 978-1-6654-7450-4

17% lower than standard- RPL. A PDR of 97% is achieved OF0 MRHOF ANN-RPL
Traffic Overhead

1359
which is 3% higher than standard RPL. This improvement in

1345
PDR, latency and power is due to the metrics RE, HC, Delay 1500

1091
1051

1035
which are used to make the routing decisions in ANN-RPL.

951
No of packets
Optimal routes are selected which have better link quality

805
1000

721
and lesser distance to root. Parent nodes with higher battery

613
levels are chosen. This ensures faster and reliable packet
delivery thus reducing number of retransmissions in turn 500
reducing power consumption. Traffic Overhead is 48% lower
as compared to Standard-RPL.
0
Latency (ms) OF0 MRHOF ANN-RPL 70 90 100
Nodes

1987
1840
1829

Fig 6: Plots of a) Latency b) PDR c) Pavg d) Traffic Overhead for

1744
1719
1714

1641
2000 ANN-RPL, OF0 and MRHOF at 10 ppm
1554

As the packet rates are increased to 10 ppm and network


1500 size to 100 nodes, the PDR decreases and delay increases.
Time (ms)

This is due to excess packets circulating in the network


1000 causing congestion and packet loss. The buffers at motes
can store up to 4 packets. Any extra packet coming at mote
500 will be dropped thus reducing the PDR. This will lead to
retransmission of packet thus reducing the energy
efficiency. At data rate of 10 ppm, ANN-RPL outperforms
0 standard OF0 and MRHOF versions.
70 90 100 A comparison Analysis of ANN-RPL with Other RPL
Node Density Implementations previously published by researchers is
a) done. ANN-RPL is compared with RPL-CGA [14],
Packet Delivery Ratio HOFESA [7], E-RPL [12], DN-RPL [11], OF-FZ [21] and
OF0 MRHOF ANN-RPL EEOPS-RPL [13]. ANN-RPL has higher PDR as compared
to Standard RPL, Chaotic genetic Algorithm, Firefly
91

100
87
86

81

Optimization Algorithm and Ant Colony Optimization


79
77
77

75
75

80 Algorithm as seen from figure 7 and Table III. ANN-RPL


60 reduces the number of advertisements overhead in the
PDR

network by transmitting DAO messages only when parent


40 change happens, thus ensuring lesser traffic, and congestion
20 in network. This ensures higher successful packet delivery
rate. DN-RPL has deep nets in the neural network which
0
gives slightly higher packet delivery but affects the delay.
70 90 100
Node Density HOFESA implementation is an empirical stability aware
b) technique which avoids frequent parent changes to maintain
network stability. It gives trivial higher PDR then ANN-
Average Power Consumption (mW) RPL but has higher power consumption. ANN-RPL has
2.938

OF0 MRHOF ANN-RPL lower power consumption as compared to most other


3 implementations. Genetic Algorithm technique and Deep
2.732

2.60348
2.676

Neural Net-based optimization has lower power then ANN-


2.8
2.50656

2.50135
Pavg (mW)

RPL. In deep nets more hidden layers are added to give


2.6
better performance. This helps to get lower Power
consumption but in compensation for higher delays.
2.4 Packet Delivery Ratio (%)
120
2.2 100
70 90 100 80
Node Density 60
c) 40
20
0

978-1-6654-7451-1/23/$31.00 ©2023 IEEE 55


Proceedings of the International Conference on Intelligent Data Communication Technologies and Internet of Things (IDCIoT 2023)
DVD Part Number: CFP23CV1-DVD; ISBN: 978-1-6654-7450-4

Control Traffic Overhead (No of packets)


Average Power Consumption (mW)
1500
2
1.5 1000
1
500
0.5
0 0

End-to End Delay (ms)


Fig 7: a) PDR b) Pavg, c) Latency, d) Overhead of ANN-RPL v/s other
1500
1400
1300
1200
1100

TABLE III : COMPARISON OF RELATED WORK

Packet Average
RPL Deliver Power Latency Control
Technique Used Metrics
Implementation y Ratio Consumption (mS) Overhead
(%) (mW)

Hybrid OF with three energy


HOFESA [7] metrics and Empirical consumption, 99 1.75 1400 1334
stability awareness RSSI, hop count

RPL-ETX [9] Single Metric ETX 95.28 1.895 1480.83 1121

RPL-HC [10] Single Metric Hop Count 95.5 1.873 1478.89 1012
Energy, ETX,
DN-RPL [11] Deep Neural Nets location of 96.56 1.42 1450 1050
nodes
Children count,
Ant colony
E-RPL [12] RER, ETX, 82 1.8 1456 810
optimization
rank value
EEOPS-RPL Firefly algorithm for RER, ETX and
80 1.80857 1475 807
[13] optimization distance
Genetic Algorithm
Queue length,
for optimizing
RPL-CGA [14] Delay, RER, 95 1.18 1350 1000
weighting factors in
HC, ETX
composition metric
Four metric
RER, ETX, HC
OF-FZ [21] combination using 96.5233 1.823 1353.82 629
and latency
Fuzzy Logic

ANN based Proposed ETX, HC, RE


ANN-RPL 97.8567 1.674 1253.82 665
work and latency

ANN-RPL uses the delay metrics to make the routing not consider delay metric in making
decision. It chooses the paths with lowest distance and delay routing decisions. RPL-CGA considers Queue length, RER,
to the root. This gives significant reduction in latency as HC, ETX and Delay. The computation time in CGA is
compared to all other techniques as shown in figure 8. E- significant due to the chaotic mechanism. The optimization
RPL, EEOPS-RPL, RPL-HC, RPL-ETX and DN-RPL do is more focused for Energy as compared to delay. When

978-1-6654-7451-1/23/$31.00 ©2023 IEEE 56


Proceedings of the International Conference on Intelligent Data Communication Technologies and Internet of Things (IDCIoT 2023)
DVD Part Number: CFP23CV1-DVD; ISBN: 978-1-6654-7450-4

compared with fuzzy based OF-FZ, the performance of [6] J. Tripathi, J. C. de Oliveira and J. P. Vasseur, "A performance
evaluation study of RPL: Routing Protocol for Low power and
ANN-RPL is seen to be superior. In fuzzy logic a range of
Lossy Networks," the 44th Annual Conference on Information
values within a fuzzy set are assigned the same value and Sciences and Systems (CISS), Princeton, NJ, pp. 1-6. (2010).
get evaluated without distinction. For instance, a candidate [7] Hassani, Abdelhadi Eloudrhiri, Aicha Sahel, Abdelmajid Badri, and
with RER value of 110 and 150 will get equal score since El Mourabit Ilham. "A hybrid objective function with empirical
stability aware to improve RPL for IoT applications." International
they will figure in same fuzzy set. But in this NN scheme,
Journal of Electrical & Computer Engineering (2088-8708) 11, no. 3
candidate with RER of 150 will clearly get higher weightage (2021).
over the 110 one. This technique has more sensitivity and [8] J. Vasseur et al., “RPL: The IP routing protocol designed for low
response to the metric values thus performs better. power and lossy networks,” Internet Protocol for Smart Objects
(IPSO) Alliance, San Jose, CA, USA, 2011.
VI. CONCLUSION [9] O. Gnawali, “The minimum rank with hysteresis objective
function,” Internet Eng. Task Force (IETF), Vol. 6719, pp. 1–13,
The ANN-RPL uses the Residual Energy metric for 2012.
choosing best parent, hence avoids using candidates with [10] E. P. Thubert, “Objective function zero for the routing protocol for
low-power and lossy networks (RPL),” Internet Eng. Task Force
lower energy levels as parents. This preserves the node (IETF), Vol. 6552, pp. 1–14, 2012.
energy and improves overall efficiency. It reduces the power [11] Shubhangi Kharche, Sanjay Pawar, “Optimizing network lifetime
consumption to the tune of 15%. The delay and HC metric and QoS in 6LoWPANs using deep neural networks”, Computers &
ensures that candidates with lower delay and shorter route Electrical Engineering, Volume 87, 2020, 106775, ISSN0045-7906.
https://doi.org/10.1016/j.compeleceng.2020.106775
are chosen as parents hence ensuring faster packet delivery. [12] Preeth, S.K.S.L., Dhanalakshmi, R., Kumar, R. et al. Efficient parent
The delay in ANN-RPL is lower by 17% as compared to selection for RPL using ACO and coverage based dynamic trickle
standard RPL. The ETX metric ensures better reliability and techniques. J Ambient Intell Human Comput, 11, 4377–4391
hence improves packet delivery. The PDR of ANN-RPL is (2020). https://doi.org/10.1007/s12652-019-01181-w
[13] Sennan S, Somula R, Luhach AK, et al. Energy efficient optimal
higher by 3%. There is significant amount of reduction in parent selection based routing protocol for Internet of Things using
the control overhead due to the mechanism adopted in firefly optimization algorithm. Trans Emerging Tel Tech. 2020;
transmitting DAO and DIO messages. The traffic reduction e4171. https://doi.org/10.1002/ett.4171
to the tune of 48% is achieved. This novel ANN-RPL [14] Cao, Y., & Wu, M. “A Novel RPL Algorithm Based on Chaotic
Genetic Algorithm. Sensors” (Basel, Switzerland), 18(11), 3647,
technique proves to be substantially better than standard Oct 2018. https://doi.org/10.3390/s18113647.
RPL. If we compare it with the Fuzzy Logic based RPL [15] Ajay Kumar, Narayanan Hariharan, “DCRL-RPL: Dual context-
versions, then it scores better due to precise computation of based routing and load balancing in RPL for IoT networks”, IET
neighbor node quality score by ANN. The evolutionary Communications, 2020, Vol. 14 Iss. 12, pp. 1869-1882.
[16] Emilio Ancillotti, Raffaele Bruno, Marco Conti, “Reliable Data
algorithm based RPL implementations require more Delivery with the IETF Routing Protocol for Low-Power and Lossy
computing time due to the large number of iterations Networks”, IEEE Transactions on Industrial Informatics, Vol. 10,
involved in the algorithms. On an average, the delay is seen No. 3, August 2014
to be around 0.4 mins. Also, they need more powerful [17] Patrick Olivier Kamgueu, Emmanuel Nataf, Thomas Djotio, Olivier
Festor. “Fuzzy-based routing metrics combination for RPL”. 2014,
microcontrollers with larger memory space. This pp1-.8. ffhal-01093965
considerably increases the deployment cost. These factors [18] Sonia Kuwelkar and H.G. Virani, “Design of an Efficient RPL
give ANN-RPL a leverage over the EA-based schemes. Objective Function for Internet of Things Applications”
ANN-RPL has much lower computation time since it International Journal of Advanced Computer Science and
Application,
involves only a series of addition and multiplication (IJACSA),12(6),2021.http://dx.doi.org/10.14569/IJACSA.2021.012
operations at the 3 layers of neurons. The delay is 0625
approximately 1.2 s and sizably lower than EA-based RPL. [19] Olfa Gaddour et al. “OF-FL: QoS-Aware Fuzzy Logic Objective
It can work well on most commercially available Function for the RPL Routing Protocol”. In: 12th International
Symposium on Modeling and Optimization in Mobile, Ad Hoc, and
microcontrollers like the MSP430. This extended and Wireless Networks (WiOpt) (2014).
improved RPL version is suitable for the various [20] Hanane Lamaazia, Nabil Benamarb, “OF-EC: A novel energy
applications of WSN and IoT domain consumption aware objective function for RPL based on fuzzy
logic” Journal of Network and Computer Applications, Elsevier,
REFERENCES May 2018
[21] Sonia Kuwelkar & Hassanali Gulamali Virani (2021) OF-FZ: An
[1] L. Atzori, A. Iera, and G. Morabito, “The Internet of Things: A Optimized Objective Function for the Ipv6 Routing Protocol for
survey,” in Computer Networks, vol. 54, no. 15, pp. 2787–2805, LLNs, IETE Journal of
Research, DOI: 10.1080/03772063.2021.1990139.
(2010).
[22] Behnam Farzaneh, Mohammad Koosha, Elahe Boochanpour,, Emad
[2] T. Clausen, U. Herberg and M. Philipp, "A critical evaluation of the Alizadeh, “A New Method for Intrusion Detection on RPL Routing
IPv6 Routing Protocol for Low Power and Lossy Networks (RPL)," Protocol Using Fuzzy Logic” IEEE 6th International Conference on
The IEEE 7th International Conference on Wireless and Mobile Web Research (ICWR), 2020.
[23] Pandian, Dr A. Pasumpon. "Novel Algorithm for Service
Computing, Networking and Communications (WiMob), Wuhan, Composition and QOS Enhancements in Composite Computing
pp. 365-372. (2011) Services." Journal of IoT in Social, Mobile, Analytics, and Cloud 2,
[3] Hui and D. Culler, "Extending IP to Low-Power, Wireless Personal no. 2 (2020): 118-128
Area Networks," in IEEE Internet Computing, vol. 12, no. 4, pp. 37-
45, July-Aug. (2008)
[4] J. Hui, JP. Vasseur, D. Culler, and V. Manral, "An IPv6 Routing
Header for Source Routes with the Routing Protocol for Low-Power
and Lossy Networks (RPL)", RFC 6554, Mar. 2012.
[5] T. Winter, P. Thubert, A. Brandt, J. Hui, R. Kelsey, P. Levis, K.
Pister, R. Struik, JP. Vasseur and R. Alexander, "RPL: IPv6 Routing
Protocol for Low-Power and Lossy Networks", RFC 6550, (2012).

978-1-6654-7451-1/23/$31.00 ©2023 IEEE 57

View publication stats

You might also like