Routing in the Internet of Things

Nidal Nasser1, Lutful Karim2, Asmaa Ali3, Muhammad Anan1, Nesrine Khelifi4
1
College of Engineering, Alfaisal University, Saudi Arabia
2
School of ICT, Seneca College of Applied Arts and Technology, Toronto, Canada
3
School of Computing, Queen’s University, Kingstone, Canada
4
PRINCE Research Unit, ISITCom, University of Sousse, Tunisia
Abstract – Sensors, RFID, Wi-Fi, and other technologies embedded project [2] is an example of such framework that integrates a
with devices and items such as home appliances, vehicles, and grocery large number of geographically dispersed sensor networks to
items improve the quality of life by exchanging information among work as an IoT. However, achieving energy efficiency and fault
each other under a common network platform that defines the emerging tolerance of sensor nodes and transmitting data packets from the
future of the Internet, also known as Internet of Things (IoT). Sensors sensor-based IoT devices to the central server with a minimum
or Wireless Sensor Networks (WSNs) consist of an integral part of IoT
since sensors can be controlled by end users and data can be transmitted
possible end-to-end delay is a great challenge. Among many
to distant sites through Internet. Moreover, thousands of sensors and network design issues such as localization, packet scheduling
similar devices poses a great challenge in routing that raises the need and data aggregation, routing data packets among IoT devices or
for zone-based (or cluster) based routing protocol. Most existing sensors is highly important since it data delivery, energy
routing protocols are not designed considering the dense architecture of efficiency and fault tolerance (that ensures the transmission of
IoT. It is a great challenge to render these algorithms adaptive to the data through alternative paths in case any data transmission path
changing requirements of sensor-based IoT applications since their fails) of such large network.
routing policies are mostly predetermined. Thus, they are not energy
efficient and fault tolerant for such mobility centric IoT. In this article, Though extensive research on routing protocols has been
we provide a brief introduction to IoT with the current state-of-the-art conducted in sensor network, they cannot work in dense IoT
research and classify routing protocols based on several factors. We architecture due to hundreds of thousands of sensors and other
then introduce a Multiple Base station and Packet Priority-based devices. In existing schemes, a single base station (BS) exists
Clustering scheme (MBPP) for IoT and evaluate its performance and data from distant nodes are routed toward the single BS
through simulations.
through a large number of hops. They also do not differentiate
Keywords – Internet of Things (IoT); Clustering; Energy Efficiency; the packets types while transmitting large number of both real-
Sensor; Multiple base station; Data packet priority. time and non-real-time data. Thus, these schemes dissipate
energy faster for sensor nodes and incur a large end-to-end delay
I. INTRODUCTION for emergency real-time data. Thus, this paper introduces a
Multiple BS and Packet Priority-based Clustering scheme
Internet of Things (IoT) [1 - 2, 7, 13] is a network of things or
(MBPP) for IoT.
objects that will extend today’s Internet by allowing interaction
among objects, such as sensors, radio frequency identification
(RFID) tags, home appliances, and vehicles through a wireless
network. IoT provides object-object interaction and brings
thousands of everyday-use objects under the same network. For
instance, integrating Wi-Fi or short range radio technologies
(e.g., Bluetooth), a weight scale device can display weight on
iPhone, send weight information to doctor, subscribe to diet
services that get weight information and accordingly send diet
recommendations, and all community members having a similar
weigh scale machine can play a game where a person with the
minimum weight wins and be used for many other applications
[1]. If cloths have RFID tags a washing machine with a RFID
reader can read cloths information and adjust water requirement
and time accordingly. Similarly, any device can be made
workable integrating with other devices to exchange information
that forms IoT.

Figure 1 illustrates the integration of different wireless devices

and technologies to form IoT, where sensors in devices are the
integral part of IoT framework. Thus, integration of sensor
networks with Internet, as part of IoT, facilitates a large number Figure 1: A large scale architecture of IoT
of smart network applications such as health care support for
remote areas, air pollution monitoring, environmental The remainder of this paper is organized as follows. In section II
monitoring, traffic congestion control as is demonstrated in presents a classification of routing protocols that can be applied
Figure 1 and thus, improves the quality of life. The SENSEI in IoT based on different factors. Section III presents the

Authorized licensed use limited to: Indian Instt of Engg Science & Tech- SHIBPUR. Downloaded on October 09,2023 at 08:13:22 UTC from IEEE Xplore. Restrictions apply.
978-1-5090-5019-2/17/$31.00 ©2017 IEEE
 proposed MBPP routing protocol for large scale IoT by defining energy consumptions of these protocols could be further
some terminologies. Its performance evaluation is conducted in reduced, and none of these protocols consider transmitting data
Section IV. Finally, the paper concludes in Section V some based on priority.
future research directions in this area. B. Classification
Routing protocols for sensor-based IoT frameworks can be
II. RELATED WORKS
classified based on different factors as depicted in Figure 2.
This section presents routing protocols of sensor network since
we consider sensor-based IoT in designing the proposed Factor 1 – Based on When Routes are Created
Multiple BS and Packet Priority-based Clustering scheme 1. Proactive: In proactive routing protocols, routes are created
(MBPP) for IoT. Then we classify the protocols based on several before they are really needed. Thus these protocols require a
factors. large routing table in each node to store all possible paths to
transmit data to all possible destinations
A. Literature Review
APTEEN [11] is a proactive routing protocol.
Low Energy Adaptive Clustering Protocol (LEACH) [6] is a
well-known clustering protocol of sensor network, where a node 2. Reactive: In reactive routing protocols, routes are created on
is selected as a cluster head (CH) if a random number (between demand whenever a node requires transmitting data to a specific
0 and 1) chosen by the node is less than a threshold value and destination. TEEN [10] is an example of reactive routing
cluster member nodes transmit data using time division multiple protocol.
access (TDMA) scheme. In this protocol, a new cluster
formation is initiated in every round, which is not energy 3. Hybrid: This is a combination of the proactive and reactive
efficient. Moreover, occasionally all CHs exist in a close area (as routing protocols. Due to the resource limitations in terms of
CH rotates in a cluster). Then, non-CH nodes require more storage and the energy available for communication and due to
energy to communicate with CHs. The Dynamic Static the large number of nodes that may result in large routing tables,
Clustering protocol (DSC) [4] eliminates the problem by the proactive protocols are not good candidates for WSNs.
allowing base station (BS) to form clusters and select CH based Hence, reactive and hybrid protocols are good choices for
on the energy levels and positions of the sensors. In DSC, cluster WSNs. Examples of hybrid routing protocols is APTEEN [11].
formation is not done in each round, after a certain number of
rounds (i.e., 10) a new cluster formation (setup) phase is Factor 2 – How Source Node Reaches Destination
initiated. However, DSC and LEACH protocols transmit at each
timeslot (i.e., large number of data transmission), which is not Whenever a sensor node has a sensed data it transmits data to the
energy efficient. The Threshold Sensitive Energy Efficient desired destination node in different ways. These are:
Sensor Networks Protocol (TEEN) [10] reduces the number of Direct Communication: In this type of routing, source node
data transmission to BS by implementing oft and hard communicates directly with the destination. Direct
thresholds. Hard threshold is the minimum value of a sensed
communication protocols are not feasible for WSNs due to their
attribute. When a node senses an event (e.g., temperature),
whose value is equal to or greater than the hard threshold, the communication range needs, increase in the required energy as
node turns on its radio to transmit data to the BS. After hard the communication range increases and scalability issues which
threshold, a node will transmit only when the attribute value is result in increasing collisions as the number of nodes increases.
changed by the soft threshold. TEEN is reactive clustering Flat Routing: In flat routing protocol, the source reaches the
protocol. To reduce the limitation of responsive data collection
destination through multi-hop communications. Flat protocols
Adaptive TEEN (APTEEN) [11] is proposed, which works both
in periodic and responsive data collection mode. However, these are good but their problem is that nodes near the sink will
protocols do not provide any fault tolerance mechanisms. quickly run low in energy because all communications are done
Periodic, Event Driven and Query Based Routing (PEQ) through them. These sink's neighbors will be traffic bottlenecks.
Protocol [12] is a reliable, energy efficient and fault tolerant Also collisions and consequently latency will increase as the
clustering protocol. In PEQ, the network is organized as a hop network expands. Flooding, SPIN and Directed Diffusion [12]
tree, where nodes at different levels are determined by value of are flat routing protocols and are presented in detailed in Section
the hop count. PEQ uses subscription/publish paradigm for III.
periodic event notification. Fault-tolerant Energy-Efficient Data
Aggregation (FEDA) [12] is another fault tolerant protocol. If a Cluster-based Routing: In cluster-based routing protocols [6,
packet is lost in FEDA due to a link error or failure between two 11, 12], sensors are grouped into clusters each with its own
nodes, other nodes that have overheard the lost packet and have cluster head. Routing is done within the cluster in a flat way, and
not yet sent their data can aggregate the lost packet with their between the clusters through the cluster heads. Due to the
results and send. Moreover, FEDA uses tree structure and during limitations of direct and flat routing protocols, clustered routing
the tree building phase, each node chooses a node as primary is the choice of several currently operational WSNs. Cluster-
parent from which it receives a tree building message for the first based routing protocols are scalable and minimize the collisions
time and as a second/backup parent from which it receives tree problem in direct and flat routing. Moreover, they are energy
building message for the second time. However, tree topology in efficient. Routes are created in a way to save energy. Several
FEDA is not efficient. Geographical and Clustering-based clustering-based protocols, like LEACH [6], TEEN [10] and
Multi-Path Data Dissemination scheme (CMPDD) [15] is PEQ [12] will divide energy load among sensors in order to
another popular clustering scheme of sensor network. However, lengthen the whole network lifetime.

Authorized licensed use limited to: Indian Instt of Engg Science & Tech- SHIBPUR. Downloaded on October 09,2023 at 08:13:22 UTC from IEEE Xplore. Restrictions apply.
 Hybrid: Although cluster-based routing is a good candidate for of intermediate levels. Thus, nodes at the
several WSNs, but others like flat routing or even direct intermediate/upper levels will have more data and
communication have some characteristics that may be needed in processing requirements compared to lower level nodes.
certain applications. For this reason, sometimes a combination Considering this observation, the length of timeslots at the
of these categories may better fit certain application upper level nodes will be longer than the timeslot length of
requirements than any single routing class does. lower level nodes.
• Real-time and emergency data will have the highest
priority while routing different types of data. However,
lower priority data can preempt real-time if they cannot
transmit for a long period of time for the continuous arrival
of higher priority data.

B. Working Principle
Initially, more than one base station (BS), N BS ≥ 2 are
Figure 2: Classification of Routing Protocols of Sensor-based IoT deployed in the IoT network. A BS is deployed a location so that
the maximum number of IoT devices can directly communicate
Factor 3 – Mobility of Sensors with the BS. For instance, if a BS is required to be placed at any
Static Routing Protocols: In static routing protocols, all sensor of the two locations L1 and L2 and the number of IoT devices that
nodes are stationary. Routing protocols are designed considering can communicate directly with the BS if it is place at L1 and L2
the application requirements and mobility pattern of sensor are NIoT1 and NIoT2, respectively where NIoT1 > NIoT2, the BS will
nodes. be placed at NIoT1. Figure 3 illustrates the deployment of two BS.
Initially, all devices have the same energy and so, a number of
Mobile Routing Protocols: Routing protocols that work for
nodes NCH are randomly selected as cluster heads (CHs), which
WSN with mobile sensor nodes. These protocols are required for
is dependent on the CH probability, PCH. If the PCH is 0.05 and
WSN that are used in habitat monitoring, battlefield surveillance
number of IoT devices is 100 the number of CHs will be 5. A
etc.
non-CH IoT device joins to the cluster of a CH, if CH and the
Factor 4 – Number of Sink Nodes device are at the shortest Euclidean distance.
Routing protocols can be classified based on the number of
sinks.
Single Sink Routing Protocols Node: Most existing routing
protocols are designed for a WSN having a single sink node.
These protocols are simple but are not scalable. LEACH [6] is a
routing protocol designed for single sink.
Multiple Sink Routing Protocols: Multiple sink routing
protocols are complex but scalable and work for most WSN
applications. These routing protocols can again be classified as
(i) static sink and (ii) mobile sink. Mobility centric single and
multiple sink routing protocols are GBEER [14] and CMPDD
[15], respectively.

III. MULTIPLE BASE STATION AND PACKET

PRIORITY-BASED CLUSTERING SCHEME
(MBPP)
In this section, we present assumptions and working principle of
MBPP routing scheme.
Figure 3: Network topology of MBPP protocol
A. Assumptions Once clusters are formed, CHs select active IoT devices (AID)
We make a few assumptions to design the proposed MBPP from all member IoT devices of a cluster. The AIDs are selected
scheme for a large scale IoT. in such a way that they provide the network coverage of the
whole cluster. We follow the AID selection process that is
• Consists of a significantly large number of sensor-based
introduced in [8]. Figure 3 shows that two clusters are formed in
IoT devices for ease of design and implementation
the network with two IoT CHs. There are three active AIDs in
• All devices have the same initial energy. They know their cluster 1, which are AID11, A12, A13 (first digit represents
locations and send the location information through multi-
cluster id) and four active AIDs in cluster 2, which are AID21,
hop to base station (BS).
AID22, AID23, and AID24. The AIDs store the sensed data in
• Uses time division multiple access (TDMA) scheme with its ready queue (memory) and schedule to transmit to its CH at
variable length timeslots. Data packets are transmitted its allocated timeslot based on data subscription and priority.
from the lowest level nodes to gateway through the nodes Each data packet will be assigned a priority number. The packet

Authorized licensed use limited to: Indian Instt of Engg Science & Tech- SHIBPUR. Downloaded on October 09,2023 at 08:13:22 UTC from IEEE Xplore. Restrictions apply.
 with the highest priority will be transmitted first. For instance, if DistOfCH[i] is the shortest then
real-time, emergency data (e.g., forest fire) should have the Node[j] joins to cluster i
highest priority to transmit. If there is no such real-time, CH[i] subscribes to Node[j] for event of interest
emergency data packet available at the ready queue of an AID,
end if
it transmits non-real time data. We follow the data scheduling
scheme that is proposed by Karim et al. in [9]. end for
/*……. Active IoT Device (AID) Selection ………….*/
Thus, each AID of a cluster transmits either the data packet or a 6. for each cluster i do
special packet to the CH in its allocation timeslot using time do divide ClusterArea[i] into a square grid[i,j]
division multiple access (TDMA) scheme. The special packet is R
very small sized just to notify the CH that the node is still alive SideOfSquare ← s
in case the node does not have any data packet to send. If the CH 2 2
does not receive any data or special packet from an IoT device AID [i, j] ← Node[1]
in a timeslot, there is a high probability that the IoT device has for k ← 2 to NoOfNodesInGrid[i, j] do
failed. However, the CH excludes the IoT device from its if ResidualEnergyNode[k] > AID [i,j] then
timeslot and cluster only if it does not receive any data or special
AID [i, j] ← Node[k]
packet for a number of timeslots (fault detection). The size of the
special packet is much smaller than that of the sent event or data. else if ResidualEnergyNode[k] = ActiveNode[i,j]
Hence, sending a special packet consumes less energy as then
compared to that of a data packet. if bufferSpaceNode[k] >
bufferSpaceNode[k + 1] then
At the end of a round, a CH aggregates data packets that are AID [i,j] ← Node[k]
received from all active IoT devices (AID) to eliminate end if
redundant data transmission and reduce energy consumption.
end if
Each CH is again tied to a BS, which has the shortest Euclidean
distance to itself. So, once a CH has aggregated data it transmits end for
to the BS it is connected to. Figure 3 illustrates that the IoT while cluster is not fully divided
network has two clusters and two BS. The BS transmits data to end for
the central server through Internet using large range wireless 7. for each AID j in cluster i do //priority based routing
communication network (if necessary). if AID [i, j] senses a data event then
schedule data based on priority and transmit
MBPP Protocol
else
NIoT1: number of devices able to directly communicate with
AID [i, j] transmit small sized special packet
BS at location L1
end if
NIoT2: number of devices able to directly communicate with
end for
BS at location L2
NBS : number of BS, NIoT: number of IoT devices NCH:
IV. PERFORMANCE EVALUATION
number of cluster heads (CH)
PCH: probability of a node being CH, Rs: sensing range of This section presents energy model, simulation setup and results.
an IoT device
A. Energy Model
1. If NIoT1 > NIoT2, BS is placed at NIoT1
/*…….. Initial BS placement ……..*/ The Multiple BS and Packet Priority-based Clustering scheme
(MBPP) uses the energy model that is proposed in [6]. If a node
2. for i ← 1 to NBS do
transmits data of size ndata to another node at distance d
Follow step 1 and 2 to place BSi
ETX = ndata × ε data + ndata × d 2 × ε air (1)
end for
E RX = ndata × ε data (2)
3. N CH ← PCH × N IoT
4. for i ← 1 to NCH do In (1) and (2), ETX and ERX represent energy consumptions for
BS randomly select CH[i] and broadcasts it ID
transmitting and receiving data respectively. ε data and ε air
for each node j that receives the message do
represent the energy spent in transmitter electronics circuitry and
if Node[j] is not a CH then
RF amplifiers for propagation loss respectively. This energy
Node[j] estimates DistOfCH[i] model sets the transmitting energy consumption of a sensor node
end if to be directly proportional to d2 where d is the distance between
end for the sending and the receiving nodes (a node here might be a BS).
end for
5. for each non-CH node j in the network do

Authorized licensed use limited to: Indian Instt of Engg Science & Tech- SHIBPUR. Downloaded on October 09,2023 at 08:13:22 UTC from IEEE Xplore. Restrictions apply.
 Table I. Simulation Parameters and Their Values
Parameter Value
Network area 100 m X 100 m
Number of cluster Max. 8
Number of gases sensors Max 200
Data packet size 24 Bytes
Energy consumptions for sending data packets 50 nJoule/bit
Energy consumptions in free space/air 0.01 nJoule/bit/m2
Initial node energy 2000 Joules

B. Simulation Setup and Results

In this simulation, we use a randomly connected Unit Disk
Graphs (UDGs) on an area of 100 meters x 100 meters as a
network simulation model. Nodes are distributed uniformly over
the clusters. Simulation parameters and their respective values
are presented in Table I. We perform the simulation for a fixed
Figure 6: Number of transmissions over a number of rounds.
number of clusters, and nodes by varying the number of rounds.
A round comprises network setup phase and steady phase (e.g., Figures 4 – 6 illustrate the performance of the proposed MBPP
routing) where steady phase has a number of frames. Each frame protocol in terms of network energy consumption, network
has a number of timeslots and a timeslot is allocated to a node in lifetime and number of data transmission and compare it with
each frame. We set the number of clusters, and nodes to 4 and existing LEACH and DSC protocols.
100, respectively.
Figure 4 demonstrates that the MBPP clustering scheme much
lower network energy as compared to the existing DSC and
LEACH clustering protocols. This is because the MBPP
protocol selects a minimum number of active IoT devices in each
cluster that provide network coverage and uses multiple base
stations (BS) so that a cluster head (CH) can choose the closest
BS to transmit the aggregated data. Moreover, in MBPP
protocol, an active IoT member device transmits a very small
sized special packet whenever it does not have any data packet
(of event of interest) to transmit. This also results in shorter
network lifetime of the MBPP protocol, as is illustrated in Figure
5. Figure 6 shows that the number of message transmitted in the
MBPP protocol is much lower than that in DSC and LEACH
protocols since a subset of IoT devices are selected to work as
active devices for a number of timeslots (or period) in MBPP
protocol whereas all devices are active in DSC and LEACH
protocols. The MBPP protocol is also scalable since multiple BS
Figure 4: Network energy consumptions of over a number of rounds allows a CH to transmit data with the minimum possible delay
and reduces energy consumptions of CHs

V. CONCLUSION
Internet of Things (IoT) opens up the door for next generation
networks and ubiquitous computing by embedding sensors,
radio frequency identification (RFID) tags, and other
technologies in any device or thing used in our everyday life to
exchange information among these devices. Thus, prolonging
the lifetime of IoT devices and fast delivery of data from sensor-
based IoT devices to the Internet or end users is of vital
importance to achieve the objectives of IoT. Among many
network design issues, efficient clustering protocol at sensor
nodes can reduce energy consumptions. In this paper, we
discussed the importance of clustering protocol in sensor-based
IoT and classified them according to a set of factors. We also
Figure 5: Network lifetime over a number of rounds. proposed Multiple BS and Packet Priority-based Clustering
scheme (MBPP) for IoT that uses more than one base stations
(BS) and prioritize the transmission of real-time emergency data
packets. The number of BS is estimated based on the network

Authorized licensed use limited to: Indian Instt of Engg Science & Tech- SHIBPUR. Downloaded on October 09,2023 at 08:13:22 UTC from IEEE Xplore. Restrictions apply.
 size. The performance of the proposed MBPP protocol is
evaluated and compared against some existing solutions. In
future, we plan to compare the clustering scheme with other
approaches.

REFERENCES
[1] Invited Speech, Rafi Haladjian, “Internet of Things”, in Lift Asia 09,
Korea, Sept 2009, http://www.youtube.com/watch?v=CeyWSlGLknY&NR=1
[2] SENSEI: Integrating the physical with the digital world of the network,
http://www.sensei-project.eu/
[3] S. A. B. Awwad, C. K. Ng, N. K. Noordin and M. F. A. Rasid, "Cluster based
routing protocol for mobile nodes in wireless sensor network," in Collaborative
Technologies and Systems, 2009. CTS '09. International Symposium on, 2009,
pp. 233-241.
[4] F. Bajaber and I. Awan, "Dynamic/Static Clustering Protocol for Wireless
Sensor Network," Computer Modeling and Simulation, 2008. EMS '08. Second
UKSIM European Symposium on, pp. 524-529, 2008.
[5] L. Catarinucci, R. Colella and L. Tarricone, "Integration of RFID and sensors
for remote healthcare," in Applied Sciences in Biomedical and Communication.
[6] Heinzelman W., Chandrakasan A., Balakrishnan H., “Energy-Efficient
Communication Protocol for Wireless Microsensor Networks”, Proceedings of
the 33rd Hawaii International Conf on System Sciences, January 2000.
[7] Lumpkins W., “The Internet of Things”, The IEEE Consumer Electronic
Magazine April 2013
[8] L. Karim, N. Nasser, and Sheltami T, “A fault-tolerant energy-efficient
clustering protocol of a wireless sensor network”, Wireless Communication and
Mobile Computing, 2011.
[9] N. Nasser, L. Karim, T. Taleb, “Dynamic Multilevel Priority Packet
Scheduling Scheme for Wireless Sensor Network”, IEEE Transactions on
Wireless Communications, 12(4): 1448-1459 (2013).
[10] Manjeshwar A., Agrawal D.P., “TEEN: A Routing Protocol for Enhanced
Efficiency in Wireless Sensor Networks”, Proceedings of the 15th International
and Distributed Processing Symposium, 2001.
[11] Manjeshwar A., Agrawal D.P., “APTEEN: A Hybrid Protocol for Efficient
Routing and Comprehensive Information Retrieval in Wireless Sensor
Networks”, Proceedings of the International and Distributed Processing
Symposium, 2002.
[12] A. Martirosyan, A. Boukerche and R. W. N. Pazzi, “A Taxonomy of
Cluster-Based Routing Protocols for Wireless Sensor Networks,” Parallel
Architectures, Algorithms, and Networks, I-SPAN 2008. International
Symposium on, pp. 247-253, 2008.
[13] Tarkoma S. and Ailisto H., “The Internet of Things Program: The Finnish
Perspective”, the EEE Communications Magazine, pp. 11-12, March 2013
[14] Kisuk Kweon, Hojin Ghim, Jaeyoung Hong, Hyunsoo Yoon, "Grid-Based
Energy-Efficient Routing from Multiple Sources to Multiple Mobile Sinks in
Wireless Sensor Networks," Wireless Pervasive Computing, 2009. ISWPC 2009.
4th International Symposium on, pp.1-5, 11-13 Feb. 2009.
[15] Shihan Li; Depei Qian; Yi Liu; Jie Tong, "Cluster-Based Multi-Path Data
Dissemination Scheme for Large Scale Wireless Sensor Networks," Wireless
Communications, Networking and Mobile Computing, 2007. WiCom 2007.
International Conference on, pp.2715,2718, 21-25 Sept. 2007.

Authorized licensed use limited to: Indian Instt of Engg Science & Tech- SHIBPUR. Downloaded on October 09,2023 at 08:13:22 UTC from IEEE Xplore. Restrictions apply.