1 Introduction

Mobile nodes in a MANET autonomous wireless ad hoc network start transmitting without taking into account base stations [1, 2]. Strategic networks, which are separate from other infrastructures involving the military and disaster relief organizations, depend heavily on networking [3, 4]. The MANET has a number of network protocol problems. In the absence of centralized control and access points, the protocols ought to provide dispersed solutions. The node mobility is superfluous in comparison to routing protocols [5], which can monitor the topology of the network. Changes in network topologies and power constraints are among the issues with network control that arise when creating higher level protocols that involve routing and executing applications using the Quality of service (QoS).

The MANET nodes wander aimlessly from one location to another. The network’s topology is fast and unpredictable changing. The nodes that are within the transmission range can communicate with one another directly. Dynamically discovering the path is the responsibility of each node. Despite numerous clustering strategies, the hierarchy in which MANET is organized needs to be enhanced in order to boost routing effectiveness as well as manage topology. The technique of clustering is used to reorganize all nodes into compact groups based on the CH’s geographic location and other nearby nodes [6,7,8,9].

The CH is in charge of overseeing the cluster’s operations, including controlling cluster processes, updating the routing table, and discovering new routes. Ordinary nodes are those in the cluster that are not the CH. Gateway nodes are nodes with inter-cluster connections that can communicate with one or more clusters. In comparison to a flat MANET, the clustering methods work better as the MANET size increases. With a high number of nodes and the nodes moving in different directions, the scalability problem in the flat MANET is extremely serious. Controlling congestion and making topology repairs quickly are two benefits of grouping mobile nodes of MANET. Clustered mobile nodes dividing are a multi-objective optimization problem when the MANET size is huge. MANET is partitioned into clusters using a variety of evolutionary methods, including genetic algorithms (GAs). GA has early convergence. Thus, to obtain the energy ware topology management, the following contributions are presented in this work.

  • For CH selection, IROA algorithm is presented.

  • In this algorithm, multi-objective functions such as power, connectivity, link lifetime, squared distance, and mobility are determined for selecting optimal CHs.

  • Besides, to manage the topology from node movement and clusters merging, cluster maintenance phase is included.

The following sections are sorted as follows. Section 2 reviews the related articles which presented efficient clustering mechanism in MANET. Topology control using optimized clustering scheme is proposed in Sect 3. Results of the proposed scheme are analyzed in Sect 4. The research paper is concluded in Sect 5.

2 Related Works

Several studies on this topic have been conducted and in this section, some of the recent works were discussed. C. Gopala Krishnan et al. [10] developed a self-configurable cluster mechanism with a k-means protocol technique to effectively designate CHs. This method was used to identify the unstable CHs that were later replaced by other nodes implementing the envisioned self-configurable cluster mechanism. Through the development of a power management system and the application of the trust management method, vulnerabilities in the MANET have been identified and addressed. For features like cluster-based algorithms and multipath routing protocols, this method used a network simulator. This suggested technique also reduced transmission failures by detecting CH failure earlier. Trust management needs to be enhanced in this method in order to address the real-time problems in MANET.

T. Maragatham, S. Karthik and R. M. Bhavadharini [11] presented the transmission and collusion aware clustering with enhanced weight clustering algorithm (TCAC-WCA)to decrease the risk of transferring hello messages to the target node. The transmission range clustering algorithm (TRCA) was implemented to get rid of unnecessary nodes that were not located inside the transmission range and to check the node's link consistency. To achieve the best routing performance, the given TCAC-WCA was tested based on two metrics named expected transmission count (ETX) or path encounter rate (PER) for each recognized condition. Finally, based on the simulation results, it was confirmed that the TCACWCA-ETX and TCACWCA-PER would produce better results when compared to the normal ***TCACWCA, but that they would suffer from a high packet delivery ratio.

An energy-efficient stable and secure clustering (EESSC) method is provided by Nagendranath M. V. S. and A. Ramesh Babu [12] to address the dual problems of security and energy efficiency in MANET design. The described EESSC technique used a fuzzy logic mechanism to properly choose CHs by utilizing five variables to accomplish stability and security. In this system, a standby CH (SBCH) was also developed to be beneficial in case of accidental occurrences such as CH dying, moving out of the cluster, or being comprised. In this circumstance, SBCH was called upon to act as CH, and another SBCH was chosen. The EESSC simulation takes place under various conditions, and the results confirmed the superiority of the proposed model over the compared ones; however, it still has to be improved in terms of security.

To overcome the issues of dynamic clustering in MANET, Fasee Ullah et al. [13] suggested the genetic Tabu Bee clustering algorithm (GTBC) approach. According to this method, the individual (bee) indicated a potential clustering framework, and its fitness was determined by its stability and balance of load. To generate stable and balanced clusters that lessen topology changes, increase network longevity, and lower clustering costs, this technique was developed by fusing honey bee traits with genetic algorithms. Performance was improved by using the tabu search features, genetic algorithm, and honey bee algorithm to create high-quality clusters. However, this method could be improved further by taking into account more matrices during the CH selection process, such as trust reputation and communication load.

B. Devika and P.N. Sudha [14] presented Chronological Earth Worm optimization Algorithm (C-EWA), for effective clustering. Graph constructing and graph clustering were the two key processes that this approach goes through. Utilizing the suggested C-EWA method for data transfer, the clustering of graphs was carried out and the best CHs were chosen. The Gabriel graph was built after the best CHs had been chosen in order to keep the MANET’s power and energy levels constant by reducing node transmission power. Although the suggested C-EWA was modified to transfer data effectively, this technique still uses more time and bandwidth and delays cluster formation.

The vulnerability of a Sybil attack on MOBIC clustering on the MANET was demonstrated by Amol Vasudeva and Manu Sood [15] with the aid of differences in the transmission capacities of Sybil nodes. To increase the possibility that a potential Sybil node would become the CH, this method made use of two theories related to the function of supporting Sybil nodes. The second theorem demonstrated that, for different MANET configurations, the suggested Sybil attack technique on MOBIC always enhanced the likelihood of a Sybil node becoming the CH during the process of election. Their simulation’s results showed that there was a considerably increased likelihood that the Sybil attacker node would become the cluster leader. This approach was limited to a certain kind of mobility-based clustering problematic behavior.

In order to effectively cluster data using cluster-head selection criteria in MANETs, Baolin Sun et al. [16] presented the Graph Kernel based Clustering Algorithm (GKCA). The shortest path (SP) was employed by this GKCA algorithm in conjunction with the d-hop graph kernel to connect various CH nodes for packet delivery. K-hop clusters were created using the cluster-head algorithm. The average end-to-end delay, the control packets ratio, and the packet loss ratio were combined to assess the stability and reliability of the MANET. The simulation results showed that the suggested strategy and parameters were a useful way to gauge and assess cluster-head stability in dynamic mobile networks. Improvements must be made to the algorithm's support for nodes with restricted mobility.

3 Topology Control Using Optimized Clustering Protocol in MANET

3.1 Overview

Figure 1 illustrates the overall diagram of the proposed scheme. To manage the topology of MANET nodes, an optimized clustering scheme is presented. At first, CHs are selected among the number of mobile nodes. For CH selection, an improved rabbit optimization algorithm (IROA) is presented. In this algorithm, multi-objective functions such as power, connectivity, link lifetime, squared distance, and mobility are considered for selecting optimal CHs. After selecting the CHs, each CH forwards the HELLO message to the neighbor nodes which are moving within its communication range. By receiving ACK from the neighbor nodes, each CH forms its cluster. Besides, cluster maintenance phase is added to preserve the cluster structure to all topological changes that occur as a result of node mobility. This phase manages the topology when the nodes leave or join to cluster and during the clusters merging. At final, the performance of the proposed scheme is analyzed terms of throughput, energy efficiency, delivery ratio and delay.

Fig. 1
figure 1

The overall diagram of the proposed scheme

3.2 CH selection

The number of mobile nodes in the network will be aggregated into various clusters in order to improve the network's energy efficiency. For efficient clustering, an improved rabbit optimization algorithm (IROA) is presented. Namely, using IROA, optimal CHs are selected from the number of mobile nodes.

Rabbit optimization algorithm [17] is a metaheuristic algorithm inspired by how rabbits survive. Detour foraging and random hiding is two laws of rabbit survival. Detour foraging is a method of exploration to avoid being discovered by natural predators by letting rabbits nibbles the grass close to the nest. In order to hide more effectively, rabbits often shift to different burrows in a random manner. Besides, to enhance the initial population quality and search ability of ROA, OBL method is added with ROA. OBL improves the searching ability and decreases the processing time. The primary steps in the IROA algorithm are,

  • Initialization

  • OBL

  • Fitness calculation

  • Compute energy factor

  • Exploration process

  • Exploitation process

  • Updation

  • Termination criteria

4 Step 1: Initialization

This algorithm’s main objective is to choose the optimal CHs. At first, initialize the artificial rabbit’s size N, the variables’ dimensionality D, the problem’s upper and lower bounds, and the maximum number of iterations. Here, the artificial rabbit is used to symbolise the CH. It is initially selected randomly. The following equation contains the initial solution format,

$$p = [\mathop {CH}\nolimits_{1} ,\,\mathop {CH}\nolimits_{2} ,....\mathop {CH}\nolimits_{n} ],$$
(1)
$$\vec{p}_{i,j} = s.(up_{j} - low_{j} ) + low_{j} \,\,\,\,\,j = 1,2,...D,$$
(2)

\(s\) defines the random number.

Step 2: OBL: The fundamental concept of OBL is that the current solution may not always be the best one; in such cases, the corresponding opposite should be considered. It seeks to increase the likelihood of identifying a solution \(\overline{p}\) from the existing solution p.


An opposite solution is estimated for each solution in this phase. It is defined as follows:

$$\overline{p} = a + b - p,$$
(3)

where, \(p \in \left[ {a,\,b} \right]\) is a real number.

Step 2: Fitness Calculation: To find the best solution, the fitness function is computed. Five factors such as mobility, power, squared distance, link lifetime and connectivity are used to calculate fitness for IROA. Since fitness is viewed as a maximisation function in this context, the best CH is chosen by clustering the solutions that produce the highest value of fitness. The IROA’s fitness is described as

$$Fit = \sum\limits_{m}^{z} {\frac{1}{5}\left[ {\frac{{\mathop P\nolimits_{m} }}{{\mathop H\nolimits_{F} }} + \mathop C\nolimits_{m} + \mathop {LT}\nolimits_{m} + \left( {1 - \frac{{\mathop M\nolimits_{m} }}{{\mathop H\nolimits_{F} }}} \right) + \left( {1 - \frac{{\mathop D\nolimits_{m} }}{{\mathop H\nolimits_{F} }}} \right)} \right]} ,$$
(4)

where m denotes the nodes, z denotes the total count of CHs, \(\mathop P\nolimits_{m}\) denotes the mth node’s power, \(\mathop H\nolimits_{F}\) denotes the normalization factor, \(\mathop C\nolimits_{m}\) denotes the connectivity, \(\mathop {LT}\nolimits_{m}\) denotes the lifetime of the link, \(\mathop M\nolimits_{m}\) represents the mobility and \(\mathop D\nolimits_{m}\) denotes the distance. These five factors are defined as follows:

(i) Power: Large transmission drains the power of nodes, which leads to a decline in node performance and, eventually, to the node's death. As a result, the node with the highest power is chosen for the clustering. Consequently, the formulation of the minimal power level for packet transmission from the node m to achieve the minimum received power is as follows:

$$\mathop P\nolimits_{m} = \sum\limits_{a = 1}^{b} {\mathop P\nolimits_{\max } \times \frac{{\mathop P\nolimits_{\min } }}{{\mathop P\nolimits_{a} }}} ,$$
(5)

where a represents the active node, b is the total count of active nodes, \(\mathop P\nolimits_{\max }\) and \(\mathop P\nolimits_{\min }\) denote the maximum and minimum power, respectively and \(\mathop P\nolimits_{a}\) denotes the received power of the ath active node.

(ii) Connectivity: The degree of connectivity is calculated using the bi-directional links used to connect two nodes, and is expressed as follows:

$$\mathop C\nolimits_{m} = \frac{1}{b}\left( {\sum\limits_{a = 1}^{b} {\frac{{\mathop C\nolimits_{a} }}{c}} } \right),$$
(6)

where \(\mathop C\nolimits_{a}\) represents the ath node’s connectivity which is calculated based on signal strength, data rate and network coverage. and c represents the total connections.

(iii) Life time of link: It connects two nodes so that information can be transmitted between them. In the network, messages are transmitted using the links. The primary issue that arises in MANET as a result of dynamic topologies is the breakage of links. As a result, the link lifetime is assessed beforehand to prevent network failure. The energy model is used to calculate the link lifetime, which is displayed as

$$\mathop {LT}\nolimits_{m} = \frac{1}{b}\left( {\sum\limits_{a = 1}^{b} {\mathop J\nolimits_{a} } } \right),$$
(7)

where \(\mathop J\nolimits_{a}\) denotes the ath node’s energy dissipation.

(iv) Mobility: According to their roles, each MANET node’s mobility is identified separately. The mobility factor is computed as follows:

$$\mathop M\nolimits_{m} = \frac{1}{{\left| {\mathop g\nolimits_{m} } \right|}}\sum\limits_{{a = \mathop g\nolimits_{m} }} {\mathop M\nolimits_{a} } ,$$
(8)

where \(\left| {\mathop g\nolimits_{m} } \right|\) denotes the neighbour nodes set of mth node, \(\mathop M\nolimits_{a}\) denotes the relative mobility and it is defined as follows:

$$\mathop M\nolimits_{a} = \frac{1}{l}\sum\limits_{i = 1}^{l} {\mathop R\nolimits_{a} } \left( i \right),$$
(9)

where l denotes the total time at which the node got HELLO packets within the time interval I, \(\mathop R\nolimits_{a}\) denotes the relative mobility and it is defined as follows,

$$\mathop R\nolimits_{a} \left( i \right) = \sqrt {\mathop r\nolimits_{a}^{2} \left( i \right) + \mathop r\nolimits_{m}^{2} \left( i \right) + \left( {2\mathop r\nolimits_{a} \left( i \right)\mathop r\nolimits_{m} \left( i \right){\text{Cos}} \left( {\mathop \theta \nolimits_{a} \left( i \right) - \mathop \theta \nolimits_{m} \left( i \right)} \right)} \right)} ,$$
(10)

where \(\mathop \theta \nolimits_{a} \left( i \right)\) and \(\mathop r\nolimits_{a} \left( i \right)\) denote the movement direction and mobility speed, respectively.

(v) Distance: The distance between the ath node and the mth CH is calculated using the links that are utilised to connect two nodes, and it is determined by

$$\mathop D\nolimits_{m} = \sum\limits_{\begin{subarray}{l} a = 1 \\ a \in m \end{subarray} }^{b} {\left( {\mathop u\nolimits_{a} ,\,\mathop g\nolimits_{m} } \right)} ,$$
(11)

where \(\mathop u\nolimits_{a} and\,\mathop g\nolimits_{m}\) denote the ath normal and mth CH, respectively.

According to (4), the solution with maximum fitness is taken as the optimal solution. Else, the solution is updated until finding the optimal solution.

5 Step 3: Compute Energy Factor

The energy of the rabbits is a key component of ROA since it changes over time, reflecting the move from exploration to exploitation. The rabbit algorithm’s energy factor is determined by

$$E(x) = 4 \cdot \left( {1 - \frac{x}{{X_{\max } }}} \right) \cdot \ln \frac{1}{s},$$
(12)

\(X_{\max }\) represents the maximum iteration and x shows the current iteration and s defines the random number in (0,1).

6 Step 4: Exploration Process

The exploration phase is mostly taken into account in detour foraging. Each rabbit’s propensity to move away from the food source and investigate a different rabbit spot randomly selected within the group is known as detour foraging. The most recent formula for detour foraging is

$$\vec{p}_{{new_{i} }} (x + 1) = \vec{p}_{j} (x) + Q \cdot (\vec{p}_{i} (x) - \vec{p}_{j} (x)) + round(0.5 \cdot (0.05 + s_{1} )) \cdot d_{1} ,$$
(13)
$$Q = l \cdot A,$$
(14)
$$l = \left( {e - e^{{\left( {\frac{x - 1}{{X_{\max } }}} \right)^{2} }} } \right) \cdot \sin (2\prod s_{2} ),$$
(15)
$$A(k) = \left\{ \begin{gathered} 1\,\,\,if\,\,k = = H(l) \hfill \\ 0\,\,\,else\,\,\,lk = 1,...,D\,{\text{and}}\,\,l = 1,...,[s_{3} ,D] \hfill \\ \end{gathered} \right.,$$
(16)
$$\begin{gathered} H = randp(D) \hfill \\ d_{1} \sim N(0,1), \hfill \\ \end{gathered}$$
(17)

\(\vec{p}_{{new_{i} }} (x + 1)\) represents the new position of rabbit and \(\vec{p}_{i}\), \(\vec{p}_{j}\) shows the ith and jth position of rabbit and \([ \cdot ]\) defines the ceiling function and r and p shows the stochastic arrangement from 1 to D and \(s_{1}\),\(s_{2}\),\(s_{3}\)\(\in\) (o,1).

7 Step 5: Exploitation Process

In the investigation phase of the algorithm, where rabbits often dig many burrows around their nests and choose one at random to hide in. Let start by defining how rabbits create burrows at random. The ith rabbit produces the jth burrow \(\vec{h}_{i,j} (x)\):

$$\vec{h}_{i,j} (x) = \vec{p}_{i} (x) + \zeta \cdot a \cdot \vec{p}_{i} (x),$$
(18)
$$\begin{gathered} \zeta = \frac{{X_{\max } - x + 1}}{{X_{\max } }} \cdot d_{2} \hfill \\ d_{2} \sim N(0,1), \hfill \\ \end{gathered}$$
(19)
$$a(k) = \left\{ \begin{gathered} 1\,\,\,if\,\,k = = j \hfill \\ 0\,\,\,else\,\,\,lk = 1,...,D\, \hfill \\ \end{gathered} \right.,$$
(20)

\(\zeta\) denotes the hidden parameter. The random hiding method’s updating formula is

$$\vec{p}_{{new_{i} }} (x + 1) = \vec{p}_{j} (x) + Q \cdot (s_{4} \cdot \vec{h}_{i,r} (x) - \vec{p}_{i} (x)),$$
(21)
$$a_{r} (k) = \left\{ \begin{gathered} 1\,\,\,if\,\,k = = [s_{5} \cdot D] \hfill \\ 0\,\,\,else\,\,\,lk = 1,...,D\, \hfill \\ \end{gathered} \right.,$$
(22)
$$\vec{h}_{i,r} (x) = \vec{p}_{i} (x) + \zeta \cdot a_{r} \cdot \vec{p}_{i} (x),$$
(23)

\(\vec{p}_{{new_{i} }} (x + 1)\) represents the new position of rabbit and \(\vec{h}_{i,r} (x)\) shows the randomly selected burrow and \(s_{4}\),\(s_{5}\)\(\in\) (o,1).

8 Step 6: Updation Process

Following the implementation of the two update procedures, the suggested model updates the new position of the rabbit by,

$$\vec{p}_{{new_{i} }} (x + 1) = \left\{ \begin{gathered} \vec{p}_{i} (x)\,\,\,\,if\,\,\,f(\vec{p}_{i} (x)) \le f(\vec{p}_{{new_{i} }} (x + 1)) \hfill \\ \vec{p}_{{new_{i} }} (x + 1)\,\,\,\,else\,\,\,f(\vec{p}_{i} (x)) > f(\vec{p}_{{new_{i} }} (x + 1)) \hfill \\ \end{gathered} \right..$$
(24)

The rabbit automatically chooses whether to stay put or relocate to a new location based on the updation value.

9 Step 7: Termination

The procedure is continued until the optimal CHs or rabbits are selected. The algorithm will be terminated after the best rabbit has been found.

9.1 Cluster Formation

After CH selection, each CH forwards HELLO message with its ID, status and weight value to its neighbor nodes. Based on the weight value of the CH, the mobile nodes joint with the CH and form a cluster. The weight value of CH is estimated as follows:

$$W\left( {\mathop {CH}\nolimits_{i} ,\,\mathop {MN}\nolimits_{j} } \right) = \beta \frac{{\mathop P\nolimits_{res} \left( {\mathop {CH}\nolimits_{i} } \right)}}{{D\left( {\mathop {CH}\nolimits_{i} ,\mathop {MN}\nolimits_{j} } \right)}},$$
(25)

where \(\mathop P\nolimits_{res} \left( {\mathop {CH}\nolimits_{i} } \right)\) denotes the residual power of \(\mathop {CH}\nolimits_{i}\), \(D\left( {\mathop {CH}\nolimits_{i} ,\mathop {MN}\nolimits_{j} } \right)\) distance between the \(\mathop {CH}\nolimits_{i} \,and\mathop {MN}\nolimits_{j}\) and \(\beta\) is the constant value.

According to (1), the \(\mathop {MN}\nolimits_{j}\) joints to the \(\mathop {CH}\nolimits_{i}\) with the maximum weight value. Namely, the \(\mathop {MN}\nolimits_{j}\) joints to nearest \(\mathop {CH}\nolimits_{i}\) which has maximum residual power.

9.2 Cluster Maintenance

The maintenance phase of clusters aims to keep the cluster structure despite all topology changes brought on by node mobility. For cluster maintenance, the following two scenarios are included in this phase: (1) node movement (2) clusters merging. The two instances below should be handled by invoking the maintenance procedure.

  1. (1)

    Node movement

    There are two main types of node motions. Node joining a cluster is step one. Node departing the Cluster is the second. If the migrating node is a Cluster member node, these movements will only have limited consequences on the structure of the clustered topology. Using the clustering Procedure, the cluster reformation must be carried out for the cluster’s nodes if the leaving node is the CH.

    Node joining: A new node initially uses a HELLO message to announce its presence. Neighbor node responds after getting the Hello message. In the absence of a CH, an ordinary node broadcasts the ID of its CH along with the status message if it is one. When the new node n receives the neighbours’ status updates, it transmits Join to the CH. The new node can transmit Join to any of the neighbours based on the strength of its signal if two of its neighbours are CHs with the same weight.

    Node leaving: A node might lose network connectivity for a number of reasons, including movement or battery depletion. A Dead message is sent to every neighbour when a node perishes. The neighbour node confirms if node n is its CH or not after getting the Dead message. Node n announces a new election and changes its reputation table if it is the CH. A CH refreshes its serving list if not.

  2. (2)

    Cluster merging

    When two nodes in the MANET are in each other’s transmission range and have the same node status as CH, re-clustering will not happen until the Cluster_Contention_Interval (CCI) timer has run out. This is done to permit unintentional interactions between passing mobile nodes. After the CCI timer timeout, if the nodes are still within transmission range of one another, the re-clustering procedure will be initiated, and the node with the greatest weight value will be chosen as CH.

10 Results and Discussion

NS2 simulation tool is used to simulate the proposed scheme. Table 1 illustrates simulation parameters and their assumptions. In this simulation, 100 numbers of mobile nodes are used. These nodes are moving within the simulation area 1000 m × 1000 m. Moving speed of the nodes is varied by 5–25 m/s. 802.11 MAC is used. Besides, the size of data packet and HELLO message are set to 3 kb and 1 kb, respectively. Maximum transmission range of each node is 250 m. AOMDV routing protocol is used for data routing. Besides, for optimal clustering, IROA algorithm is presented. Figure 2 illustrates the simulation output of the CH selection. Figure 3 illustrates the simulation output of cluster formation.

Table 1 Simulation parameters and their assumptions
Fig. 2
figure 2

The simulation output of the CH selection

Fig. 3
figure 3

The simulation output of cluster formation

Performance metrics

Delay: It is the time it takes for data to travel from a sensor node to its destination.

Delivery ratio: It is a metric that measures the percentage of data packets successfully delivered from the source node to the intended destination node within the network.

Drop: It refers to the situation where data packets transmitted by sensor nodes are not successfully received by their intended destination.

Energy Consumption: It refers to the amount of electrical energy that nodes utilize to perform their communication, and data transmission.

Network lifetime: It refers to the duration of time during which the network remains operational and capable of supporting communication among its nodes.

Throughput: It refers to the amount of data that can be successfully transmitted through the network within a given period of time.

11 Performance Analysis

The performance of the suggested clustering method is compared to clustering based on the ROA, particle swarm optimization (PSO), and GA in this section. As the ROA algorithm has good convergence rate than ROA, PSO and GA, CH selection is done with less processing time. Thus, delay of proposed scheme is reduced to 14%, 24% and 23% than that of ROA, PSO and GA, respectively, as shown in Fig. 4. The comparison of delivery ratio for various clustering strategies is shown in Fig. 5. The delivery ratio decreases as the number of nodes rises, as shown in the figure. The suggested scheme's delivery ratio is increased to 9%, 12%, and 13%, respectively, compared to ROA, PSO, and GA, though. After cluster formation, each cluster in the network is maintained by the pre-defined conditions so that packet drop and energy consumption of the network is reduced further. Namely, as shown in Fig. 6, packet drop of the proposed scheme is reduced to 25%, 55% and 65% than that of ROA, PSO and GA, respectively.

Fig. 4
figure 4

The comparison of delay of different clustering schemes

Fig. 5
figure 5

The comparison of delivery ratio of different clustering schemes

Fig. 6
figure 6

The comparison of drop of different clustering schemes

Figure 7 illustrates the comparison of energy consumption of different clustering schemes. As illustrated in the figure, energy consumption increases when the number of nodes increases. Namely, energy consumption of the proposed clustering scheme is reduced to 19%, 37% and 40% than that of ROA, PSO and GA, respectively, when the number of nodes is 20. Similarly, when there are 100 nodes, the suggested scheme’s energy usage is 29%, 35%, and 39% lower than that of ROA, PSO, and GA, respectively. Figure 8 shows a comparison of the proposed scheme's network lifetimes. The network lifetime of the suggested method is enhanced to 13%, 96%, and 98%, respectively, compared to ROA, PSO, and GA. Figure 9 illustrates the comparison of energy consumption of different clustering schemes. As illustrated in the figure, the proposed scheme achieved throughput 4000kbps than the existing methods when the number of nodes is 50 while it obtained throughput of 7400kbps the existing methods when the number of nodes is 100.

Fig. 7
figure 7

The comparison of energy consumption of different clustering schemes

Fig. 8
figure 8

The comparison of network lifetime of different clustering schemes

Fig. 9
figure 9

The comparison of throughput of different clustering schemes

12 Conclusion

To control topology changes due to mobile nodes, an optimized clustering scheme has been presented in this work. For CH selection, IROA algorithm has been presented. In this algorithm, multi-objective functions such as power, connectivity, link lifetime, squared distance, and mobility are considered for selecting optimal CHs. After CH selection, each CH forwards the HELLO message with its weight value to the neighbor nodes. Based on the weight value of the CH, the mobile nodes joint with the CH and form a cluster. To manage the topology changes due to node mobility, cluster maintenance phase has been included in this work. From the simulation results, the proposed scheme achieved better throughput and energy efficiency than ROA, PSO and GA based clustering schemes.