International Journal of Internet of Things 2022, 10(1): 22-31

DOI: 10.5923/j.ijit.20221001.02

A Combined Metric Objective Function for RPL

Load Balancing in Internet of Things
Kala Venugopal*, T. G. Basavaraju

Department of Computer Science and Engineering, Government Engineering College, Hassan, Karnataka, India

Abstract In an IoT-oriented intelligent world, everything around us is interconnected and integrated. The IoT ecosystem
comprises a network of constrained devices called Low-Power Lossy Networks, where RPL (IPv6 Routing Protocol over
Low power lossy networks) is recognized as the standard routing protocol for these networks. Though RPL projects several
distinctive features, Load Balancing is identified as one of the significant inadequacies unaddressed by this protocol. In this
paper, a novel Combined Metric Objective Function (COM-OF) is proposed that considers a combination of significant
metrics to design a load-balanced DODAG, that distributes the traffic load equally amongst the different nodes within the
network and maximizes the network lifetime. The metrics of our interest incorporates the expected lifetime of the node, the
child count, the energy consumption, and link reliability for the nodes. The performance of COM-OF is compared with the
standard RPL objective functions OF0 and MRHOF using the Cooja simulator of Contiki OS. The results prove that
COM-OF exhibits enhanced performance than OF0 and MRHOF with 33% reduction in power consumption, 97% packet
delivery ratio, up to 45% improvement in network lifetime and 36% reduction in the average number of child nodes.
Keywords Internet of Things, Load balancing, Low power Lossy Networks, Objective Function, RPL

power. The interconnects are constrained, providing low

1. Introduction data rates, stability, and high loss rates. LLNs support
point-to-point, point-to-multipoint, and multipoint-to-point
IoT envisages the idea of worldwide connectivity. Here
traffic flow between thousands of nodes.
we can connect a network of dedicated devices or objects
These constrained devices can be connected to the IPv6
called “things” that are entrenched with sensors, actuators,
network using the 6LoWPAN (IPv6 over Low-power
software and, electronics to the Internet so that they can
Wireless Personal Area Network) standard. Here, an
gather and exchange real-world data without human
adaptation layer is compressed between the IP and 802.15.4
intervention. Post data collection, the sensors share their
standard protocol layers in the protocol stack [3]. This
data through other IoT devices or the gateway, which is
adaptation layer compresses and fragments the IP protocol
later analyzed locally or in the cloud. It is anticipated that
stack 1280 octets MTU (Maximum Transmission Unit) into
the volume of data created by IoT will reach a massive total
packets of 127 bytes, attuned to the 802.15.4 standard. It also
of 79.4 zettabytes, with around 75 billion devices connected
lets routing protocols be implemented at the network layer to
by 2025 [1]. IoT has shaped multiple opportunities to link
provide end-to-end connectivity for various applications [4].
the natural world with computer-aided systems leading to
Routing data for these heterogeneous low power lossy
automation in numerous fields and empowering innovative
networks is a unique unresolved challenge due to the
applications. When combined with the latest technologies
intrinsic properties of IoT like inadequate resources,
like 5G, IoT helps improve the operational efficiency
fluctuating link quality, and regular topological changes.
of organizations with minimal cost and enhanced
Therefore, routing protocols that can acclimate to these
decision-making and user experience.
low-power lossy links play a vital role in reliable end-to-end
IoT typically deals with a distinct category of a network
data delivery. Here we look for reliable routing protocols that
of embedded devices termed Low-power Lossy Networks
can handle critical and massive data generated by the
(LLNs), where both the routers and interconnects are
tremendous traffic of numerous IoT applications.
constrained [2]. The routers are constrained in resources like
Unfortunately, with the evolution of the Internet and IoT,
low memory, battery power, bandwidth, and processing
conventional routing protocols like OSPF, OLSR, AODV, or
* Corresponding author:
DSR cannot lodge in IoT, overcoming the limitations of
[email protected] (Kala Venugopal) LLNs. Therefore, IPv6 Routing Protocol for Low power and
Received: Apr. 14, 2022; Accepted: May 6, 2022; Published: May 12, 2022 Lossy Networks (RPL) was developed as the routing
Published online at http://journal.sapub.org/ijit protocol used in 6LoWPAN for LLNs and was standardized
 International Journal of Internet of Things 2022, 10(1): 22-31 23

for IoT [2]. RPL is proposed to aid MultiPoint-to-Point on source routing developed by IETF for LLNs and designed
(MP2P), Point-to-MultiPoint (P2MP), and Point-to-Point for IEEE 802.15.4 PHY and MAC layers [2].
(P2P) traffic. MP2P traffic provides connectivity to the
core private IP network or the Internet. A destination 2.1. Topology and Control Messages of RPL
advertisement technique is used by P2MP traffic to provide RPL is constructed based on Directed Acyclic Graphs
paths toward the downward destinations. P2P traffic routes (DAGs) that calculates routes between LLN nodes. A node
the packet from the root to the destination with the help of can have multiple parents in the DAG structure, unlike trees
routing tables at the nodes [2]. with only one parent. In RPL, nodes are organized as
RPL has many incomparable features like nominal battery DODAGs, i.e., Destination-Oriented DAGs. The network's
usage, quick topology creation, self-healing nature, and a root node also referred to as the Sink or Border Routers (BR)
loop-free structure. Still, one of the significant limitations of or LLN Border Routers (LBR) or Gateways, connects the
this protocol is that it cannot handle Load balancing, which is 6LoWPAN RPL network to the IPv6 Internet. Also, the RPL
required for the fair dispersal of traffic in the network. RPL network can run multiple RPL instances, each instance
Objective Functions create volatile load traffic for the nodes identified by a unique RPLInstanceID and having single or
significantly closer to the root. The preferred parent node several DODAGs in it.
through which most of the traffic is diverted becomes a The Network topology in RPL is formed by exchanging
fragile node serving all the traffic from the child nodes four RPL control messages that are different types of
compared to other parent nodes. This drainage of energy ICMPv6 control messages [8]. These messages are identified
from the preferred parent node causes network breakage, by their code field
affecting network lifetime and reliability [5].
This paper intends to equalize the network load and • A DODAG Information Solicitation (DIS) message is
increase the network's lifetime. We present a novel Objective issued by a new node trying to reach and attach to a
Function named Combined Metric Objective Function network and explore the nodes nearby.
(COM-OF), which considers the ETX metric and energy • A DODAG Information Object (DIO) message is
consumption of the nodes and a combination of the multicast by a node, usually the root node, in the
Estimated Lifetime (ELT) metric and the Child count metric. network, to let the nearby nodes determine the RPL
We compare the COM-OF with the RPL’s default objective Instance and let them choose a parent based on the Rank
functions, Minimum Rank with Hysteresis OF (MRHOF) [6] and other design specifications. The rank of a node
and Objective Function Zero (OF0) [7]. The performance identifies the node's location in the network from the
evaluation of COM-OF is done by conducting simulations DODAG root.
using the Cooja simulator of Contiki OS, with a network of • A node replies with a Destination Advertisement Object
10, 30, and 50 nodes, and the results are compared. (DAO) message to its parent, broadcasting the route
The key features of the proposed objective function are as information in the upward path and informing its
follows: participation.
• A node receiving the DAO message responds by
• First, the Child Count metric is calculated based on the sending back a unicast Destination Advertisement
parent ID in the DIO message. Object Acknowledgement (DAO-ACK).
• Then, the Estimated Lifetime (ELT) metric is
calculated based on the total traffic generated, ETX, 2.2. DODAG Construction
and energy consumption.
• A Combined Metric Objective Function (COM-OF) is DODAG formation is centered on the Neighbour
projected that uses a combination of metrics mentioned Discovery process, where DIO messages are directed from
above to construct a load balancing DODAG. COM-OF the root to all the other nodes in the network, and DAO
minimizes the energy consumption, advances the messages are sent from the other nodes to create a path
network lifetime, and increases the network reliability towards the root [9]. A node issues a DIS message probing
in comparison to the standard objective functions. for the topology information from its neighbors to join a
The paper's structuring takes the following layout: Section network. Based on the Trickle timer, DIO messages are
2 outlines the RPL protocol functioning. Section 3 reviews broadcast from the root to the child nodes in the network.
the related works about load balancing. The proposed work This DIO message is accepted by a node wanting to join the
is briefed in Section 4, and the performance evaluation is network or a node already existing in the network. The DIO
presented in Section 5. Ultimately, Section 6 concludes the message carries information regarding the RPLInstanceID,
research work. the details of the node’s rank, and the Objective function.
The node, based on (1) and (2), calculates the Rank and
updates the parent list with the sender of the DIO message
2. Outline of RPL Protocol [10].
The IPv6 Routing Protocol over Low power lossy Rank(NewN) = Rank(PParentN) + Incr_Rank (1)
networks (RPL) is a Distance vector routing protocol based Incr_Rank = Step X minHopRankIncrement (2)
24 Kala Venugopal and T. G. Basavaraju: A Combined Metric Objective
Function for RPL Load Balancing in Internet of Things

Here, Rank(NewN) indicates New Node Rank, recent years.

Rank(PParentN) indicates the Preferred Parent Rank, Authors in [10] use an amalgamation of three metrics,
Step signifies a scalar value, and minHopRankIncrement is ETX, Content, and Energy, to fulfill the needs of different
the smallest increment in node’s rank when compared to its IoT applications using the Cooja simulator of Contiki OS.
parents. After the rank calculation, the updated rank is They use the enhanced triggering technique to remove the
forwarded with the DIO message. trickle timer’s accumulative effect on the short-listen
On accepting the DIO message, a new node adds the problem. It was observed that when compared to default OF,
sender address to the parent list, uses the objective function the results for PDR and Latency delay were better for Energy
to calculate its rank, and multicasts this rank in the DIO combined with Content (EC) and aggregation with Enhanced
message. An existing node receiving the DIO message can timer (EC_En_Timer) design, and the results for energy
reject it or use the objective function to calculate the rank. If consumption was better for Residual Energy (RE) merged
the new computed rank is lesser than its current rank, the with ETX (EE) and combination with Enhanced timer
existing parent is discarded, and a better location in the (EE_En_Timer) design.
DODAG is accepted. If not, it can maintain its prevailing To handle the energy imbalance problem and improve the
position in the DODAG. Thus, by considering the fact that network lifetime, the authors in [11] propose a Time OF
the parents' rank must be lesser than the child nodes’ rank, (T-OF) where ETX, a weighted combination of energy, and
routing loops can also be avoided in the DODAG [9]. the Number of Children and Siblings (NOCS) metrics is used.
Compared to MRHOF, OF0, and Energy Objective Function
2.3. RPL Objective Functions (EN-OF), T-OF shows a PDR of 95%, an improvement in the
RPL uses the Objective function (OF) to construct routes, lifetime of the network to 54 minutes, and an advancement in
defining how the nodes choose the best-optimized route in an balancing of energy due to lesser latency of 11.35 seconds.
RPL instance [2]. Here, the nodes support and advertise one In [12], the authors compare the diverse OFs where
or more metrics and constraints through their DIO messages OF0, MRHOF (ETX), MRHOF (NONE), MRHOF (ETX
used for the parent selection in DODAG. An Objective + ENERGY), MRHOF (ENERGY) and is compared
function specifies how these metrics or constraints are for enhanced RPL performance. It was observed that
converted into a rank value, defining the location of a node MRHOF(ETX) could be favored for bigger networks and
from its DODAG root and specifying the parent selection that Hybrid OF is best suited for small networks compared to
mechanism for a node in DODAG. Thus, referring to the larger networks. Simulations with the Cooja simulator in
DAG container within the DIO messages, the objective Contiki OS show an increase of 24% in PDR, a decrease of
function decides on the rules for using metrics and 28% in power consumption, and a lesser inter-packet time of
constraints, parent selection mechanism, load balancing, etc. 39% compared to single metric RPL.
[2]. Unfair traffic distribution due to a lack of load balancing
The two Objective Functions that are defined in the in the network also leads to a decline in network efficiency.
RPL protocol are Objective Function Zero (OF0) [7] and A new method based on cluster ranking called C-Balance is
Minimum Rank with Hysteresis OF (MRHOF) [6]. proposed in [13] to improve the network lifetime. Here,
OF0 is proposed to find the closest Grounded root. The cluster heads are identified for clusters using a first rank that
OF0 finds the shortest path to the sink node based on the hop is calculated. Then, parents are determined for these cluster
count metric, where a node’s rank is calculated based on the heads for packet forwarding using a second rank. The
least number of hops to the sink. Here, load balancing is not metrics ETX, hop count, the number of children, and the
considered while routing the traffic towards the root through residual energy were used to calculate the Ranks. The
the preferred parent. C-balance functioning is equated with OF0, MRHOF, and
The Minimum Rank with Hysteresis OF (MRHOF) QU-RPL using the Cooja simulator of Contiki OS with a
discovers the routes with minimum path cost avoiding network topology of 20, 40, and 60 nodes. It was observed
unnecessary churns in the network. The term ‘hysteresis’ that the energy consumption and network lifetime in
confirms that the path with minimum rank or cost selected is C-Balance are improved by 30 to 40% and 28 to 48%,
less than the existing one by a threshold value. MRHOF respectively, and the average number of child nodes is
utilizes the Expected Transmission Count (ETX) metric that reduced by 15 to 23%. However, the end-to-end delay is
computes the quality of the link, thereby considering the more in comparison with OF0, MRHOF, and QU-RPL.
congestion at the link level rather than the node level. To extend reliability, the lifetime of the network, and
stability of the network, two novel objective functions
called the weighted combined metrics objective function
3. Related Survey (WCM-OF) along with non-weighted combined metrics
objective function (NWCM-OF) is projected in [14]. They
As already mentioned, Load balancing in RPL networks are centered on the ETX, the quality of the link, the node
is one of the critical issues that need to be addressed. energy consumption, and the additive combination with
Researchers have proposed several methods to alleviate the equal and non-equal weights. Simulations on the Cooja
load balancing issue and improve the network lifetime in emulator of Contiki OS show that the network performance
 International Journal of Internet of Things 2022, 10(1): 22-31 25

based on the proposed objective function is improved not meet all the performance requirements of a network that
compared to the default OFs. include network lifetime, stability of the network, end-to-end
A new enhancement of the RPL protocol called Energy delay, PDR, control overhead, reliability of the network,
and Load aware RPL (EL-RPL) protocol is projected in [15] energy imbalance issues, power consumption etc. It is
to maximize the network lifetime. The rank of DODAG is therefore quite evident that a single objective function may
calculated based on a composite metric based on Load and not be able to achieve all the needs and requirements of the
battery depletion index and ETX for route selection. The different IoT applications. In [16], though the objective
performance of EL-RPL is approximated to RER(BDI) RPL function confirms better load balancing and network lifetime,
and fuzzy logic-based RPL (OF-FL RPL) protocols using the it also leads to several DIO and DAO messages. While
Cooja simulator. EL-RPL shows improved performance in research work in [5] is yet to consider a greater number of
network lifetime and PDR by 8-12% and 2-4%, respectively. metrics for their future work, resource-based load balancing
To balance the Load in terms of traffic in the network and is the future deliberation in [18]. Also, validation of the
extend the lifetime of overburdened nodes, the authors in [5] efficiency of the different approaches on real test beds [17] is
propose the Load balanced objective function (LB-OF) that still an anticipated work for most of the current research
considers the child count for each candidate parent node. The works.
different steps put forward include amending the DIO
message, including a new utilization technique for this DIO,
and finally balancing the nodes using a new RPL metric. The 4. The Proposed Work
simulations were conducted using the Cooja simulator and 4.1. RPL Load Balancing Problems
Contiki OS for up to 100 nodes and compared with MRHOF
and OF0 OFs. It is observed that LB-OF displays better Unequal traffic or load affects the connectivity of LLNs,
performance for power consumption, PDR, and network leading to various issues like congestion in the network,
lifetime. excessive utilization of the energy of the nodes, reduced
Authors in [16] propose the Parent Aware Objective network lifetime, reliability, and stability of the network.
Function (PAOF) for RPL, where the ETX metric and parent As RPL is conceived for heterogeneous networks of
count metric are used to calculate the optimal route to the heterogeneous devices, it must deal with unequal and
sink. Here, a more significant number of intermediate nodes heavy traffic. RPL objective functions and metrics lead to
are preferred parents to route the packets to the destination. congestion as it does not consider Load Balancing.
The Contiki OS and the Cooja simulator are used to compare Several reasons can induce load imbalance in a network.
the performance of PAOF with MRHOF. It was observed One such instance is the Hotspot problem [19], where a
that PAOF showed improvements in parent diversity and preferred parent node is overloaded with traffic during
load density ensuring better load balancing and network congestion in the network and increases the data relay to
lifetime. manage the traffic. Such a node becomes a hotspot in a
To enhance the reliability and power consumption of RPL, network, leading to energy depletion and reduced network
LBSR, which is a new Load Balancing Objective Function, lifetime. Another instance is the Bottleneck problem [20],
is projected in [17]. Here, a new routing primitive to where the hotspots are the first-hop node to the root called
calculate the number of children, a new load balancing the bottleneck nodes. As most of the traffic is directed
mechanism to fairly distribute energy among the nodes, a through the bottleneck nodes, the energy depletion of such
new path selection mechanism to eliminate instability issues, nodes occurs must faster. The Thundering herd or herding
and a new routing propagation mechanism is developed. effect problem [21] [17] is another reason for load imbalance
Cooja simulator and Contiki OS are used to simulate in a network during congestion. Here, a set of nodes keeps
the network, demonstrating improvements in energy switching between the preferred parent nodes in a pointless
consumption, PDR, and load distribution compared to RPL. attempt to balance the load. A similar problem is the
To detect and correct the load imbalance, bottleneck, and Instability problem [22], where the nodes keep changing
network congestion in large networks at the different levels their preferred parent based on the updated rank value and
in DODAG, a new child count method called Child Count link metric value of the parent node leading to instability of
based Load Balancing in RPL (Ch-LBRPL) that enhances the network. Another reason to cause load imbalance, totally
[5] [17] is proposed in [18]. The simulations were conducted by chance, is the “Randomly unbalanced network” problem
for 31 nodes using the Cooja simulator of Contiki OS and [23]. Here, a parent node can be repeatedly and randomly
compared to RPL and Load balanced RPL (LBRPL). The selected between two parent nodes having the same rank.
number of DIO messages were observed to be reduced with
4.2. RPL Load Balancing Metrics
minimal parental switching. It also exhibited better control
overhead and energy usage. The mathematical model The selection of a path is based on the metric of an
proposed also suits well for scalable network size too. objective function and is measured as the path cost [24].
However, from the aforesaid review, it is observed that Metrics could be node-based or link-based, dependant on
research works in [10-12] and [14–15] concentrate on the where it was developed in a network. Metrics like Energy
usage of one or more metrics that are quite complex but does and hop count are examples of node-based metrics, whereas
26 Kala Venugopal and T. G. Basavaraju: A Combined Metric Objective
Function for RPL Load Balancing in Internet of Things

metrics concerned with link quality like ETX, throughput, Thus, using the Expected Lifetime (ELT) metric, we can
latency, etc., are link-based metrics [25]. We can also use a find the bottleneck nodes that will run out of energy first.
single metric or multiple metrics for routing in a network. ELT metric can be used to consider the data traffic and
The metrics of interest deliberated for our proposed work are energy of the nodes to balance the overburdened nodes.
given below. 2) Child Count Metric: The leaf nodes of RPL networks
1) Expected Lifetime Metric: A node's Expected Lifetime always tend to select a preferred parent node with a lower
(ELT) is the residual time until a node gets exhausted rank when compared to parent nodes with fewer children but
of its energy [26]. In our work, we not only consider higher ranks. A lower rank parent node entices more leaf
the link quality variations and maximization of reliable child nodes, thereby draining its energy quicker than the
energy-efficient routes but also reducing the consumption of other nodes leading to network breakage [20]. Thus, in our
energy by the most energy-consuming nodes or bottleneck proposed work, to achieve the node lifetime maximization
nodes. We take into consideration a node’s Expected lifetime and handle the overloading issue, we keep track of the
for parent selection. ELT metric is used to maximize the number of child nodes for a given parent node during the
lifetime of bottleneck nodes in a network. It calculates the parent selection mechanism of DODAG construction in RPL
time for a node to die if it keeps forwarding traffic at a given network using the child count metric [20]. Here we consider
rate. selecting a parent with a fewer number of child nodes when
The ELT calculation for a Node (NN) is done as follows compared to other parent nodes to achieve load balancing.
[26]. During RPL DODAG construction, when a parent node in
• Total traffic to transmit T(NN) for a given new node the DODAG receives the DIO message from the child node,
NN is calculated as the traffic generated by the node it usually discards the message. In our proposed work, the
TG(NN) along with the traffic generated from its child parent node is made to utilize the DIO message coming from
nodes TC(NN) as given in (3). the child node to buffer it and calculate the number of child
nodes for it [20]. For this, we amend the DIO message with
T(NN) = TG(NN) + TC(NN) (3)
its Parent ID along with the RPL InstanceID, details of the
where TC(NN) = Σ C ε Children (NN) TC node’s rank, and OF. When a parent node receives the DIO
• Initially, The Average number of retransmissions RTavg message from the child node, it adds it to a buffer set
is calculated using (4). maintained and compares its ID with the parent node ID. If it
RTavg = Σ P ε Parents (NN) αPN Χ ETX(NN,PN) (4) is the same, then the parent node adds one to the child set,
thereby keeping a count of the child nodes during the
where ETX (NN, PN) is the reliability of the link from Node
construction of the DODAG. A node then uses this child
NN to its parent node PN, PN∈Parents(NN), αPN is the traffic
count metric in the objective function to calculate its updated
ratio to the parent node and Σ P ε Parents(NN) αPN = 1. ETX metric
rank.
is explained further in the forthcoming subtopic. Then,
the Average number of MAC transmissions (TMACavg)
is determined by (5), where T(NN) is the total traffic
transmitted and RTavg is an average of the number of
retransmissions.
TMACavg = T(NN) x RTavg (5)
• Then, the ratio of medium usage time for transmission
is calculated using the DataRate as given in (6)

TMACavg
(6)
DataRate
• The energy to transmit the total traffic is calculated
using (7), where the product of the ratio of medium
usage time for transmission and the Radio Transmission Figure 1. Parent selection using child count metric for load balancing
power of the node PTX(NN) is calculated.
TMACavg
Figure. 1 shows the parent selection mechanism based on
X PTX (NN) (7)
child count metric. Here a new node 10 wants to join the
DataRate
DODAG consisting of an LBR (LLN Border router) and nine
• Lastly, the Expected lifetime of the node NN (ELT(NN))
other nodes. It must choose between the two parent nodes,
is computed using (8) by finding the ratio between the
Node 2 with three children and Node 3 with one child node.
node’s Residual energy Eres and the energy for total
As Node 3 has a lesser number of children when compared to
traffic transmission. The residual energy metric is also
Node 2, the new Node 10, based on the implementation of
explained in the forthcoming subtopic.
child count metric, selects Node 3 as its parent node. This
E
ELT(NN)= TMACavg res (8) helps equalize the load in the network.
X PTX (NN)
DataRate 3) Expected Transmission Count Metric: The Expected
 International Journal of Internet of Things 2022, 10(1): 22-31 27

transmission count (ETX) metric is the expected number of load balancing by considering significant metrics like the
transmissions required to successfully transmit the packets number of child nodes and the Expected lifetime of the nodes
from the sender node to the receiver node without any errors in addition to the ETX and residual energy of the nodes in the
[25]. It measures the path quality between two nodes in the network. During DODAG construction, the DIO message is
network and provides high-throughput reliable paths for a amended with the parent ID. Once a node accepts the DIO
network. A low value of ETX represents a reliable link. message, it compares the parent ID with its own ID to keep
The value of ETX [15] is calculated using (9). track of the child count if the message is from its child node.
ETXnew = β . ETXold + (1 - β). ETXpacket (9) The traffic to and from the node is calculated using the
1
throughput metric [26]. The ELT value for the node is
where ETXpacket = , calculated based on the traffic estimated, ETX value, and
PF X PR
which is the probability that the packet is sent in the residual energy of the node, as explained in the earlier
forward direction and reached the receiver (PF) and the sections. The ELT and the child count values are used to
acknowledgment has reached back to the sender in the calculate the new rank of the node. Here, the objective
reverse direction (PR) [25] and β represents the learning ratio function selects a preferred parent with a minimal number of
which in ContikiRPL is given a value of 0.9 [5]. children, a maximum network lifetime with minimal energy
In RPL, the OF uses the ETX metric to calculate the rank. usage, and a reliable link to the node. This preferred parent
During DODAG construction, on the reception of the DIO selection helps equalize the load in the network.
message, the node uses the rank value with minimum ETX We understand that the metric ELT is maximizable,
value to choose the parent node. Though ETX may provide whereas the metric CC is minimizable. To transform the
a reliable network, it may not provide an optimal path derived metrics in our proposed work (ELT and CC) to
considering the load in the network. the same domain [27], we express the network lifetime
4) Residual Energy Metric: The residual energy of the maximization metric as inverse lifetime minimization [28].
Therefore, to make the ELT metric minimizable, we use the
nodes in a network plays a substantial role in balancing the
inverse lifetime value of the node N, ELTinv (N) given as
load, increasing the network lifetime, and calculating an 1
optimal path in LLNs. As shown in (8), the Residual Energy ELTinv (N) = to calculate the rank. Also, we use two
ELT (N)
metric is used to calculate a node's Expected lifetime (ELT) constant parameters, α = 0.5 and β = 0.5, whose values
and eventually improve the network lifetime. should be between 0 and 1, with a total value of 1. These
The Residual Energy (Eres) of a node N is calculated using parameters are used to balance each metric’s influence on the
(10). network and are given by (13).
Eres (N) =
Eini (N)
(10) MinOF (ELTinv, CC) = α * ELTinv (N) + β * CC (N) (13)
Ecurr (N)
Algorithm 1. The COM-OF Algorithm
Where, Eini (N) is the node’s initial energy and Ecurr (N) is
the node’s current energy. The node’s Current Energy Ecurr Input: CurrentNode (CN), NodeID, ParentNode (PN), ParentID,
(N) is calculated using (11), where Econs (N) is a node’s NodeSet (NS)
consumed energy. Output: Load balanced preferred parent selection
Ecurr (N) = Eini (N) – Econs (N) (11) Begin
CN ← DIO
A node’s Energy consumption (Econs) is calculated using if (NodeID == ParentID) then
(12). CC ← CC + 1;
EnergestV X Curr X Volt
Econs = (12) end if
RTimerSec
//Calculate the rank based on the ELT and Child Count
Here, EnergestV represents the time in ticks that a node is for all CN in NS do
in a particular mode or state, that is, CPU Idle, Low-Power Calculate and update the ELTinv value and CC for every node
(LPM), Transmission (Tx), or Reception (Rx) state. Curr in NS
and Volt are the CPU consumed Current and Voltage, Rank (CN) ← Rank (PN) + Rank_Increment;
respectively, in accordance with the Tmote sky motes, and Rank_Increment ← Step + MinHopRankIncrement
RTimerSec is the platform-dependent constant number of Step ← α * ELTinv (CN)+ β * CC (CN);
ticks, basically used to convert time from ticks to seconds, MinHopRankIncrement ← 256;
given as 32768 ticks per second. if SenderNodeRank < NodeRank then
Thus, during the RPL DODAG construction, the Residual Add the SenderNode as parent and discard the
energy metric can be used in the objective function for parent current parent;
selection by a node. else
Maintain the current location in the DODAG
4.3. The Proposed Combined Metric Objective Function
end if
The projected Combined Metric Objective Function end for
(COM-OF) provides a holistic approach towards network End
28 Kala Venugopal and T. G. Basavaraju: A Combined Metric Objective
Function for RPL Load Balancing in Internet of Things

Now, the ELTinv and Child Count (CC) values help The performance parameters considered for the comparison
calculate the rank of the node. include Power consumption, Packet Delivery Ratio (PDR),
The Proposed Combined Metric Objective Function Network Lifetime, and the average number of child nodes.
(COM-OF) algorithm is given in Algorithm 1. The simulations are carried out multiple times, and the
The flowchart for the parent selection process is explained average of the performance parameters is considered. The
in Figure. 2. simulation parameters considered for the evaluation are
given in Table 1.
Table 1. Simulation Parameters

Parameter Value
Operating System Contiki 3.0
Node Type Tmote sky
Routing Protocol RPL
MAC/ Adaptation Layer ContikiMAC/ 6LowPAN
Radio Environment Unit Disk Graph Medium (UDGM)
Number of Nodes 10, 30, 50
Simulation Duration 60 min
Full Battery 3000 mJ
Transmission Range 100 m
Data Packet Timer 60 sec

One of the essential parameters to be considered for Load

balancing in RPL is the power consumption of the nodes.
Power consumption is given in milliwatts (mW), which is the
average energy consumption of all the nodes in the network
during its lifetime. The Network Lifetime is dependent on
the energy consumption of the nodes as batteries power
the nodes. The energy consumption of the nodes can be
calculated using the equation already given in (12).
Figure 2. Flowchart for the parent selection process
Therefore, based on (12), the power consumption of a node
(PCons) is provided by (14).
5. Performance Evaluation PCons =
ECons
Run_Time
(14)

To gauge the performance of our proposed objective where ECons is the Energy consumption of a node as
function, we have carried out our simulation experiments explained in Equation (12) and Run_Time is the time
using the Cooja simulator of the Contiki Operating system interval between the different EnergestV modes of a node.
[29]. Contiki is a prevalent, accessible, and open-source Here, the power consumption of the nodes using the
operating system specially designed for IoT, targeting small, proposed COM-OF is compared to OF0 and MRHOF for a
constrained devices with insufficient memory, bandwidth, network consisting of 10, 30, and 50 nodes, respectively. It is
battery power, and processing power. It is compatible with observed that the power consumption of the nodes rises with
the IP network stack where the existing protocols like UDP, the density in the network. MRHOF consumes more power
TCP, and HTTP coexist with protocols designed explicitly when compared to the other two OFs. Figure. 3, Figure. 4,
for constrained networks like CoAP, 6LowPAN, and and Figure. 5 show the power consumption of nodes using
RPL. Cooja is a powerful network simulation software OF0, MRHOF, and COM-OF for 10, 30, and 50 nodes,
tool developed by Contiki-OS developers to simulate respectively. There is around a 33% reduction in power
Contiki-based applications. Contiki motes of small and huge consumption using the proposed COM-OF compared to OF0
networks can be simulated or emulated using the simulator. and MRHOF in smaller networks. For medium-sized
We use the Tmote sky motes for our simulation, which are networks, the power consumption using COM-OF is reduced
low-power, high data rate wireless sensor module that by up to 10% compared to OF0 and MRHOF. Therefore, on
supports interoperability with IEEE 802.15.4 devices [30]. average, there is around a 20% reduction in power
The simulation is carried out with 10, 30, and 50 nodes consumption of the nodes using the proposed COM-OF
spread over an area of 100 X 100 meters for one hour each. compared to OF0 and MRHOF. Thus, a load-balanced
In all the scenarios, the proposed COM-OF is compared with network leads to a decrease in the consumption of energy of
the standard RPL Objective Functions, OF0 and MRHOF. the nodes in a network.
 International Journal of Internet of Things 2022, 10(1): 22-31 29

Figure 3. Power consumption for 10 nodes Figure 5. Power consumption for 50 nodes

Figure 4. Power consumption for 30 nodes

Figure 6. PDR comparison of OFs for 10, 30, and 50 nodes
Packet delivery ratio (PDR) is another important metric Another significant evaluation metric to consider to check
that measures the successful transmission of packets in a the performance of a load-balanced network is the Network
network. It is figured as the ratio of the total number of Lifetime. Network Lifetime is defined as the time up to the
packets received at the destination node (Σ Preceived) by the point where the first node in the network dies or uses up its
total number of packets sent by the sender nodes (Σ Psent) as energy. As already seen, an improvement in the energy
given in (15). consumption of the nodes for a load-balanced network would
∑ Preceived
PDR = (15) extend the lifetime of the nodes and, in turn, improve the
∑ Psent
Network Lifetime. We estimated the lifetime of the nodes
Figure. 6 compares the PDR of OF0, MRHOF, and using (8) to share and balance the Load in the network
COM-OF for 10, 30, and 50 nodes, respectively. It is equally. Figure. 7 shows the graph of Network Lifetime
observed that COM-OF gives up to an 8% increase in the using OF0, MRHOF, and COM-OF for 10, 30, and 50 nodes,
packet delivery ratio with a maximum value of 97% respectively. It is noticed that there is considerable
compared to OF0 and MRHOF. For medium-sized networks, improvement in the network lifetime from 23% to 33% and
the packet delivery ratio of COM-OF equals that of MRHOF, 20% to 45% using COM-OF when compared to OF0 and
providing a reliable link for load balancing in the network. MRHOF, respectively. We can see that as the density of the
For a load-balanced network, it is therefore seen that the network increases, the network lifetime decreases. This
packet delivery ratio increases with the number of nodes in a decrease is because the additional traffic from a more
network. significant number of nodes to the bottleneck nodes depletes
30 Kala Venugopal and T. G. Basavaraju: A Combined Metric Objective
Function for RPL Load Balancing in Internet of Things

their energy much faster to run out of energy finally. Thus, energy consumption and increasing the network's lifetime.
extending the network lifetime involves balancing the Load
or traffic among the nodes in a network.
6. Conclusions
In this paper, a novel Combined Metric Objective
Function (COM-OF) is recommended to address one of the
prime issues in RPL called Load balancing. COM-OF
balances the load in a network by considering the Child
count, the Estimated lifetime, the energy consumption of the
nodes, and the reliability of the links in the network. Our
projected objective function was simulated using the Cooja
simulator of Contiki OS for small and medium-sized
networks, and the performance was compared with the
standard RPL objective functions OF0 and MRHOF.
COM-OF exhibits better performance when compared to the
single metric objective functions OFO and MRHOF in terms
of power consumption, PDR, network Lifetime, and the
average number of child nodes. On average, COM-OF
exhibits a 10 to 33% reduction in power consumption,
36% reduction in the average number of child nodes, 18 to
33% improvement in network lifetime, and around 8%
Figure 7. Network Lifetime comparison of OFs for 10, 30, and 50 nodes improvement in the packet delivery ratio. In our future work,
we plan to consider other composite metrics and methods
that also assess the stability and scalability of the network.

REFERENCES
[1] https://www.statista.com/statistics/471264/iot-number-of-co
nnected-devices-worldwide/.
[2] T. Winter et al., “RPL: IPv6 routing protocol for low-power
and lossy networks”, IETF RFC 6550, March 2012.
[3] Rahul Sharma and T. Jayavignesh, “Quantitative analysis and
evaluation of RPL with various objective functions for
6LoWPAN”, Indian Journal of Science and Technology, Vol
8(19), 2015, DOI: 10.17485/ijst/2015/v8i19/76696.
[4] H. Kharrufa, H. A. A. Al-Kashoash and A. H. Kemp,
“RPL-based routing protocols in IoT applications: a review”,
in IEEE Sensors Journal, Vol. 19, no. 15, pp. 5952-5967,
Figure 8. Comparison of Average number of child nodes for 10, 30, and August 2019, DOI: 10.1109/JSEN.2019.2910881.
50 nodes
[5] Mamoun Qasem, Ahmed Al-Dubai, Imed Romdhani, Baraq
As Load balancing in COM-OF is also based on the child Ghaleb and Wajeb Gharibi, “A new efficient objective
function for routing in internet of things paradigm”, 2016
count metric and ELT metric, we estimated the average IEEE Conference on Standards for Communications and
number of child nodes in the network using COM-OF and Networking (CSCN), pp.1 – 6, 2016,
compared them to OF0 and MRHOF for a network with 10, DOI: 10.1109/CSCN.2016.7785168.
30, and 50 nodes. Figure. 8 shows that the average number
[6] Gnawali O and P Levis, (2012): The Minimum Rank with
of child nodes for a node is comparatively reduced by an Hysteresis Objective Function, RFC 6719.
average of 36% in COM-OF compared to OF0 and MRHOF.
The number of child nodes is reduced by 25% to 50% using [7] P. Thubert, “Objective function zero for the routing protocol
COM-OF in dense and small networks. However, we notice for low power and lossy networks (RPL)”, RFC 6552, March
2012.
that the average number of child nodes for COM-OF and
MRHOF is the same for a small-sized network. By reducing [8] A. Conta, S. Deering and E.M. Gupta, “Internet control
the average number of child nodes for a node, we equally message protocol (ICMPv6) for the internet protocol version
distribute the load among the nodes in a network, reducing 6 (IPv6) Specification”, 2006, RFC: 4443.
 International Journal of Internet of Things 2022, 10(1): 22-31 31

[9] Olfa Gaddour and Anis Koubaa, “RPL in a nutshell: a survey”, [19] Wakatsuki, Shun, Nobuyoshi Komuro, Hiroo Sekiya and
Elsevier, Computer Networks, Volume 56, Issue 14, Pages Shiro Sakata, “Prolonging network lifetime for 6LoWPAN /
3163-3178, 2012, DOI: 10.1016/j.comnet.2012.06.016. RPL wireless sensor network using mobile sink with dynamic
sojourn time”, 2014.
[10] Solapure, S.S. and Kenchannavar, H.H, “Design and analysis
of RPL objective functions using variant routing metrics for [20] B. G. Mamoun Qasem, Ahmed Al-Dubai and Imed Romdhani,
IoT applications”, Wireless Networks 26, 4637–4656, 2020, “Load balancing objective function in RPL”, ROLL – WG
https://doi.org/10.1007/s11276-020-02348-6. INTERNET DRAFT, pp. 1–10, 2017.
[11] Seywan Moradi and Reza Javidan, “A new objective function [21] Hyung-Sin Kim, Hongchan Kim, Jeongyeup Paek and
for RPL routing protocol in IoT to increase network lifetime”, Saewoong Bahk, “Load balancing under heavy traffic in RPL
International Journal of Wireless and Mobile Computing, routing protocol for low power and lossy networks”, IEEE
2020 Vol.19, No.1, pp.73 – 79, 2010, Transactions on Mobile Computing, Volume.16, No.4, pp.
DOI: 10.1504/IJWMC.2020.109270. 964-979, 2017, DOI: 10.1109/TMC.2016.2585107.
[12] Gupta, N., Pughat, A. and Sharma, V, “A critical analysis [22] M. Aboubakar, M. Kellil, A. Bouabdallah and P. Roux,
of RPL objective functions in internet of things paradigm”, “Toward intelligent reconfiguration of RPL networks using
Peer-to-Peer Networks. Appl. 14, 2187–2208, 2021, supervised learning”, 2019 Wireless Days (WD), Manchester,
https://doi.org/10.1007/s12083-02101180-9. United Kingdom, pp. 1-4, 2019,
DOI: 10.1109/WD.2019.8734236.
[13] Seyed Ali Fatemifar and Reza Javidan, “A new load
balancing clustering method for the RPL protocol”, Springer, [23] Doruk Pancaroglu and Sevil Sen, “Load balancing for
Telecommunication Systems, Issue 2/2021, RPL-based internet of things: A review”, Ad Hoc Networks,
DOI: 10.1007/s11235-021-00760-7. Volume 116, 102491, ISSN 1570-8705, 2021,
https://doi.org/10.1016/j.adhoc.2021.102491.
[14] Abdelhadi Eloudrhiri Hassani, Aicha Sahel and Abdelmajid
Badri, “A new objective function based on additive [24] Mah Zaib Jamil, Danista Khan, Adeel Saleem, Kashif
combination of node and link metrics as a mechanism Mehmood and Atif Iqbal, “Comparative performance
path selection for RPL protocol”, International Journal of analysis of RPL for low power and lossy networks based on
Communication Networks and Information Security different objective functions”, International Journal of
(IJCNIS)”, Vol. 12, No. 1, April 2020, Advanced Computer Science and Applications, Vol. 10, No.
DOI: 10.54039/ijcnis.v12i1.4446. 5, 2019, DOI: 10.14569/IJACSA.2019.0100524.
[15] Sankar, S., and Srinivasan, P, “Energy and load aware [25] Vasseur, J., Kim, M., Pister, K.S., Dejean, N., and Barthel, D.,
routing protocol for internet of things”, International Journal “Routing metrics used for path calculation in low-power and
of Advances in Applied Sciences, Vol. 7, No. 3, pp. 255~264, lossy networks”, RFC 6551, 1-30, 2012.
2018, ISSN: 2252-8814,
DOI: 10.11591/ijaas.v7.i3.pp255-264. [26] Oana Iova, Fabrice Theoleyre and Thomas Noel, “Using
multiparent routing in RPL to increase the stability and the
[16] Necip Gozuacik and Sema Oktug, “Parent-aware routing for lifetime of the network”, Ad Hoc Networks, Elsevier, Volume
IoT networks, Internet of Things, Smart Spaces, and Next 29, pp. 45-62, 2015,
Generation Networks and Systems”, Springer International https://doi.org/10.1016/j.adhoc.2015.01.020.
Publishing, Switzerland, pp. 23-33, 2015,
DOI: 10.1007/978-3-319-23126-63. [27] Th. Zahariadis, Ed. and P. Trakadas, Ed., “Design guidelines
for routing metrics composition in LLN”, ROLL Internet
[17] Baraq Ghaleb, Ahmed Al-Dubai, Elias Ekonomou, Wajeb Draft, 2022.
Gharibi, Lewis Mackenzie and Mustafa Bani Khalaf, “A new
load-balancing aware objective function for RPL’s IoT [28] Nesrine Khernane, Jean Couchot and Ahmed Mostefaoui,
networks”, 2018 IEEE 20th International Conference on High “Maximum network lifetime with optimal power/rate and
Performance Computing and Communications, pp. 909-914, routing trade-off for wireless multimedia sensor networks”,
2018, DOI: 10.1109/HPCC/SmartCity/DSS.2018.00151. Computer Communications, Elsevier, 124, pp.1 – 16, 2018,
hal-02182832.
[18] A. Sebastian, “Child count based load balancing in routing
protocol for low power and lossy networks (Ch-LBRPL)", [29] Contiki O.S and Cooja simulator, http://www.contiki-os.org/.
Springer Nature Singapore Pte Ltd, Smart Systems and IoT:
Innovations in Computing, Smart Innovation, Systems and [30] Moteiv Corporation. Tmote sky: Datasheet (2006): https://in
Technologies, Volume 141, pp. 141-157, 2020, sense.cs.st-andrews.ac.uk/files/2013/04/tmote-sky-datasheet.
DOI:10.1007/978-981-13-8406-6_15. pdf, Nov 13, 2006.

Copyright © 2022 The Author(s). Published by Scientific & Academic Publishing

This work is licensed under the Creative Commons Attribution International License (CC BY). http://creativecommons.org/licenses/by/4.0/