Uneven clustering: An approach to maximize the lifetime of sensor nodes

Wireless sensor networks (WSNs) are made up of thousands of tiny electronic devices with the ability to monitor and communicate with one another. These tiny electronics devices are called as nodes normally placed in an operational environment with the purpose of sensing and gathering environmental data, like pollution levels, humidity, pressure, and temperature etc. The collected data can be sent to central place called base station or sink for further processing. The wireless sensor networks are characterised by certain constraints such as battery energy, storage space and computation ability due to the employment of midrange microprocessors/microcontrollers. The majority of the application involves the sensor nodes being placed haphazardly inside the targeted area, necessitating unsupervised operations. The typical design issue in the operation of WSN is the self-organization, since the WSNs are used in difficulty terrains in which human intervention is not possible. The basic function of sensor nodes is to sense, collect, process and forward the data to base station. These processes must be designed in such away that the energy consumption by the node must be very less, since the energy in the sensor node is limited. The majority of routing protocols work to extend the life of networks by lowering power usage. The literature review provides an evident that the clustering WSN for data forwarding is best way to maximize the lifetime of the networks. The purpose of this research paper is to maximize the lifetime of wireless nodes by proposing an energy-efficient uneven cluster-based routing protocol (UCR). In this protocol, the uneven clusters are created by dividing the network in such way that, the cluster size increases as we move away from the base station. The primary objective of the uneven clustering is that the load in terms of energy consumption is distributed almost uniformly among cluster heads (CHs). The proposed UCR take care of hot spot issues well with uneven cluster size. This approach pays significantly towards network longevity.


INTRODUCTION
The world has witnessed the revolution in data processing due to developments in the field of MEMS technology.This led to the development of cheaper, less power and a tiny embedded processor with suitable sensing ability namely motes/sensor nodes.For the given application, the wireless sensor networks are formed by group of large number of sensor nodes.This number may vary from thousand to tens of thousands depending on the requirement.The WSNs have gained attention of research community because, the nodes are inexpensive and could able to address cheaper solutions to diverse applications in real-time.The limited battery life, low processing power of processor, and low memory for storing data and application programs in sensor nodes limits the resiliency of a wireless sensor network.These small nodes are widely utilized in applications associated with disaster management, including detection of wildfires, flood relief, and battle zones etc.The battery power in the sensor nodes must be wisely managed to elongate the lifetime of the networks, since the nodes are built with limited battery power and not possible to recharge in the operational field.The primary design issue in the development of WSN is the energy efficiency which directly associated with the lifetime of sensor node.
The node with the highest energy in a clustered WSN is referred to as the cluster head (CH), and the other nodes are referred to as cluster members.The CHs transmitter must be ON till the new CH selection to collect sensed information from their member nodes (MNs) [1].The data is compiled by the CHs and sent to the base station, which may be situated far away.In clustered WSNs, the CHs communicate with the base station using either methods single hop and multi-hop communication [2].Regardless of how far away or close the base station is, the CHs transmit data directly to it when using a single hop technique.Using intermediary CHs, the data is transmitted to the base station in a multi-hop approach.
The CHs primarily performs additional activities as compared to member nodes.In one hop approach, the CHs farther from the base station loses their energy more rapidly than the ones nearest to base station, since they need more power to reach the base station.This approach led to an uneven energy load distribution among the CHs.In multi-hop approach, the CHs loses their energy almost evenly and is better compared to one hop approach.
Compared to the member nodes, the energy depletion in CHs is substantially larger [3].If the nodes stay as CHs for an extended length of time, they run out of energy more quickly and die sooner.Most of the cluster-based routing schemes rotate the CH function in the network on a regular basis to ensure a fair distribution of the energy consumption across the nodes.
Although the multi-hop technique is thought to be superior than the single hop approach, the energy load on CHs near the base station is larger than that of CHs farther away.Since these CHs required to receive and forward the information from the far away clusters in addition to their local information.The energy depletion in the nearest CHs is much more than the CHs farthest from the base station.Due to this non-uniform energy load distribution among CHs, the nearest CHs die out prematurely creating hotspot problems [4].Taking all these facts into account, we propose an uneven clustering which addresses the hotspot problem and also tries to distribute the energy load among all the CHs.

LITERATURE REVIEW
The LEACH [5] protocol proposed for the WSNs applications to collect the information from the member nodes of the cluster at regular interval.The protocol employs a probabilistic method to select a node as a CH and communicates aggregated data to the central collecting station through a one-hop.To balance the energy consumption, the CH's role switches between nodes on a regular schedule.The primary characteristics of LEACH protocol is that, it is simple and assures low communication overhead.In heterogeneous networks, however, LEACH performs poorly because it elects CHs without considering the nodes' residual energy.Several new protocols, including the multi-hop communication-based algorithm LEACH-M [6], the fuzzy based clustering protocol [7], the centralized algorithm LEACH-C [8], the multi-level clustering algorithm EEMLC [9], and an algorithm for heterogeneous networks [10], were proposed by researchers to fix these issues.
Using the cluster radius, HEED [11] determines the amount of power for transmission to be utilized in intra-cluster broadcast.Each node's residual energy determines the initial probability that it will become a tentative CH.The determination of final CHs is based on the intra-cluster communication cost.The goal of HEED is to disperse CHs fairly across the network and to conclude after a certain amount of iterations.
A far-off node sends data to a neighbouring node in the direction of the cluster head in a chain-based clustering protocol PEGASIS [12].After adding its data, the node sends it to the following neighbor.This proceeds until the cluster head receive the data.This protocol's primary flaws are chain formation and delay.
The hot spot problems are not dealt with by LEACH.Unequal clustering algorithms may be employed to handle hot spot problems.High inter-cluster traffic leads to unequal energy consumption, which was addressed by the unequal clustering algorithm based on unequal cluster size (LUCA) [13].The network is split up into various, varying-sized clusters.In order to act as the CH and balance energy consumption, a few high energy nodes are placed at specific locations.
An algorithm for energy-efficient unequal clustering (EEUCB) is presented in [14], wherein nodes are integrated into clusters of unequal sizes.Because EEUCB employs a probabilistic method for choosing CHs, it results in lone nodes.
In an UCLA [15], layers are constructed and a base station is positioned in the center of the grid for large-scale wireless sensor networks.Because the layers nearest the base station are smaller, the inner layers have more remaining energy for inter-cluster traffic.Because of the combined local broadcast and join mechanism, ULCA offers significantly reduced communication cost at the time of clustering the network and a longer network lifetime.
The WSNs are also used in specific applications, for example underwater acoustic sensor networks [16], traffic routing applications [17], vehicular sensor networks for analyzing the pollution level in real time [18] etc.

COMMUNICATION MODEL
The N number of tiny nodes are distributed haphazardly over a significant physical area to form the network of the proposed protocol.The sensor nodes that have been placed monitor the surrounding environment on frequent intervals and relay information gathered to the closest cluster heads.A few presumptions regarding deployment and the structure of the wireless sensor network are as follows: • The field of measurement is located at a considerable distance from the base station.• Once they are deployed, the base station and sensor nodes remain stationary.• Every sensor node is identical, and has a fixed amount of stored energy.• Every sensor can dynamically control the transmission power according to the distance from the receiving node.• Sensors have the ability to function in both low-power sleeping mode and active mode.• A distinct identifier (ID) is given to every node.
• The network's sensor nodes are cognizant of their position.
• The base station is aware of every sensor node's location within the network.The communication model described in [8] considered for the proposed protocol is depicted in Figure 1.In this model, the transmitter uses its ability to regulate power to send data to the receiver using the least amount of energy possible.
In order to forward data, the power amplifier and transmit electronics in the transmitter dissipates some amount of energy i.e. and ( ) respectively.The following formula determines the energy needed to send a k-bit of information to a given physical interval of d: The transmitter power consumption can be dynamically controlled by selecting the appropriate power amplifier settings based on the geographical separation between the transmitting and receiving devices.The multi-path (m p ) model is used if the physical interval is higher than a cut off distance (d 0 ), and the free space (f s ) model otherwise.The transmitter energy consumption model for multi-path fading channel (d 4 power loss) and free space (d 2 power loss) channels is provided below: The following represents the amount of energy consumed by the receiver circuitry to power the radio electronics to collect the information:

UNEVEN CLUSTERING PROTOCOL (UCR)
Single-hop communication happens between CHs and member nodes in cluster-based routing protocols, while multi-hop communication happens between CHs and base stations.The energy expended in the CHs is sum of the energy dissipated due to communication between member nodes and CHs (intra cluster) and communication between CHs of neighbouring cluster (inter-cluster).
The energy depletion of CHs for intra-cluster communication in an even-sized clustered network is nearly equal.However, CHs nearer the base station consume more energy because these must process additional information from the farther-off CHs.Because of the uneven energy consumption, the CHs nearer the base station die more quickly, causing hot spot issues.An innovative uneven cluster routing (UCR) algorithm is put forth in an effort to resolve this issue.The proposed novel routing algorithm partitions the network into uneven sized clusters.The size of the clusters is smaller in the vicinity of the base station and larger in the distant ones.The networks sensor nodes energy consumption is fairly balanced by the uneven clustering.The UCR protocol operations divided into cluster creation phase, steady state phase and routing phase as shown in Figure 2.

Cluster creation phase
During this phase, a random selection of FORWARDER nodes is made by the sink.These nodes gather initial data from nearby nodes, including initial energy, geographic position, and node unique IDs, and forward it to the sink node.The sink intelligently divides the sensor field into a number of uneven clusters based on the information collected.For every cluster, the sink generates a sequence table that includes the sequence number for every node in that cluster.The sequence number indicates the window of opportunity for the node to become a CH.Eventually, the sink node broadcasts into the network.The following sections describe the various stages that make up the cluster creation phase.

INFO gathering.
The sink notifies all nodes in the network to send information during this stage by broadcasting a message.The sink node location information is included in the broadcast message.Based on the broadcast message received from the sink, the nodes in the network are able to determine how far away the sink is and adjust the transmit power accordingly.A few randomly chosen nodes are designated as FORWARDER by the algorithm, and these nodes are able to transmit all information from their neighbouring nodes to the sink.Nodes' ID, location, and initial energy level are all included in the information.

Partitioning.
In this stage, the sensor field is judiciously divided into a number of uneven clusters by the sink.When it comes to an unevenly clustered network, the clusters that are closest to the sink have fewer members and are smaller in size than the clusters that are farther away.The uneven cluster size approach significantly contributes in optimizing the network lifetime and managing the energy of each CH.In order to create clusters of uneven size, base station calculate cluster radius RC i for each cluster using the maximum cluster radius R i [19].
where d(C i , S) is the distance between the cluster center C i and sink, C is a constant coefficient with a value ranging from 0 to 1, R i is the maximum cluster radius, and d max is the maximum distance from sink to the clusters and d min is the minimum distances from sink to the clusters.By changing the C values, the radius of the cluster can be adjusted.The cluster radius decreases if the C value ranges from 0 to 1 and the network is divided into required number of uneven clusters.After completion of the network partition, the base station creates a CH sequence table for each cluster.The main goal of using a CH sequence table is to reduce energy consumption by eliminating fresh negotiation among nodes to determine who gets to be the new CH after each round.The nodes ID, clusters ID and as well as the probable CH sequence numbers have been included in the sequence table.A node's eligibility to become a CH in subsequent round can be determined by its CH sequence number.While allocating a CH sequence number, the sink takes into account the position of the nodes and the separation between the CH of neighbouring clusters.This ensures the uniform distribution CHs over the network.Finally, the CH sequence table for all the clusters is broadcasted into the network.Whenever the node depletes it energy and about to die, it transmit a message to co-members of the cluster to update CH sequence table.

Steady state phase
During this phase, nodes get their CH sequence numbers and the cluster ID from the table and determine their turn to become a CH.For the first round, the node becomes CH if its CH sequence number is 1, for the second round, it becomes CH if it is 2, and so on.In general, if a cluster has n number of nodes and the node with sequence number i becomes a CH for the rounds i, n + i, 2n + i and so on.The next section describes the two stages of the steady state phase, which are the formulation of the schedule and the advertisement.

Advertisement stage.
Network-wide node grouping must begin simultaneously, which demands node synchronization.The sink sends a beacon signal into the network following each round to keep the process synchronized.When all the CH nodes according to the CH sequence number receive this beacon signal, starts grouping the nodes by broadcasting a CH message into the network.Nodes choose which CH to join after a predetermined amount of time.The choice is made in accordance with the received advertisement message's signal strength.The member node chooses the closest CH to join it since it needs less transmission power to get there and notifies the CH node that it is now a member by sending join message.

Schedule creation.
After getting the join message, the CH nodes count the number of member nodes and build the TDMA schedule accordingly.Every node has been given a distinct time slot, and nodes are free to communicate within that window.The CH nodes transmit this TDMA schedule to their member nodes.Once nodes have determined their TDMA time slots, they can switch off their transceiver and await their turn in sleep mode.This approach further extends the battery life by reducing the energy consumption in the nodes.

Routing phase
During this phase, the CH nodes need to keep their transceiver turned on in order to receive data messages coming from the member nodes.The member nodes provide data to the CH node during The processed data is sent to the sink by the CH via multi-hop communication.Since the CH are aware of the locations of nearby CHs.The aggregated data is forwarded in the direction of the sink via adjacent CHs.In doing so, the networks hot spots are reduced and the CH nodes energy consumption is balanced.
The protocol concludes a first round of operation when data from every CH node reaches the sink.To ensure that each node's energy consumption is balanced, the CH's role is alternated.This is achieved by re-grouping of sensor nodes by the new CH nodes.The sink initiates the next round of operation by transmitting node synchronization signal into the network.All the nodes, which are eligible to become CH based on the CH sequence number starts re-grouping of sensor nodes in distributed way by broadcasting an advertisement message as explained in the section 4.2.1 of steady state phase and the network operation continues.

RESULTS AND DISCUSSIONS
A simulation employing MATLAB is used to assess the UCR protocols efficiency.The energy costs associated with data transmission and aggregation for each round are considered while computing each node's energy consumption.To assess the proposed UCR protocol's energy efficiency, the LEACH protocol is employed.Table 1 displays the radio model's simulation parameters [8] are considered for simulating the proposed protocol.The performance is evaluated by deploying 200 nodes in 100mX100m square region.
In order to get to the sink, the CHs in LEACH use a one-hop method, which requires more energy.The communication overhead is more in LEACH, because the nodes make fresh negotiation to select the CH after completion of every round.In LEACH compared to UCR, the node energy quickly runs out and it dies faster.Figure 3(a) illustrates that while the FND (First Node Dies) in LEACH was recorded in round 820, it was reported in round 965 in UCR.In comparison to LEACH, the proposed UCR has a much slower nodes death rate.As can be seen from the result in Figure 3(b), the LND (Last Node Dies) is reported in round 1496 in LEACH, but it is reported in round 1267 in the proposed algorithm.The crucial factor in assessing WSN performance is lifetime, which is represented graphically in Figure 4.The number of nodes that survive over a simulation round is used to figure out the lifespan.
As illustrated in Figure 4, UCR considerably increases the network lifetime relative to LEACH in terms of both the duration till the first node and the last node dies.It can be seen from the outcome that the proposed approach has a slower node death rate.The protocol employs uneven clusters to maintain equilibrium in the energy spent on communication by the sensor nodes.The hot spot problem is effectively dealt by balancing energy consumption in nodes.Due to uneven clustering, larger clusters with many member nodes are found far from the sink, while smaller clusters with few member nodes are found close to the sink.
The clusters closer to sink are smaller in size and has subsequently fewer member nodes, the CHs closer to the sink require a moderate amount of energy for interaction within a cluster and a significant amount of energy for inter-cluster communication.The farthest clusters larger in size and can accommodate more member nodes, the heads use more power for intra-cluster interaction.The proposed protocol has the potential to extend the network lifetime by employing an uneven clustering technique.
An additional metric used to assess the protocol's performance is the network's total residual energy.Over the course of a simulation round, it is the total of all the nodes residual energy.Figure 5 illustrates that the overall energy dissipation in LEACH is greater than that of the proposed approach.Thus, the uneven clustering algorithm that is proposed outperforms LEACH and optimizes the lifetime of the network.

CONCLUSION
An inventive and energy-conserving uneven clustering protocol for WSNs is presented in this research work.The equal size clustering with multi-hop routing approach creates hot spot problems.Most research works present that the CH rotation and their residual energy are insufficient to balance the energy utilization throughout the network.An uneven clustering strategy is used to deal with the hot spot issue and balance energy consumption evenly among CHs.
The network is divided into clusters of varying sizes by the proposed UCR protocol, with the clusters closer to the base station being smaller and the clusters farther away being larger.The CHs further away from the sink require more energy for intra-cluster

Figure 3 :
Figure 3: Network lifetime in terms of FND and LND

Table 1 :
Simulation parameters related to the UCR protocol