3.1 Introduction

Advances made in the wireless sensor networks (WSNs) have attracted and gained momentum in the various fields of real-time applications like medicine, military, monitoring systems, and tracking systems. The different aspects and requirements of the applications have led to the development of low-cost and power consumption wireless devices. Decision-making process is one of the supporting tools which help to take action by analyzing several parameters. The incorporation of different devices exposes a different form of information, and thus, the decision-making process becomes quite complex tasks [1]. In specific, information obtained from the WSNs has the potentiality of associating toward different devices which can lead to delayed information transfer processes. In spite of those unique features rendered by WSNs, despite its application in real world is limited. In general, the radio range is used to connect the different devices. Relied upon the requirement of application, the sensor nodes are deployed and communicated. Nodes form networks by organizing among themselves to reachable and great value information from the physical environment [2].

The nodes are managed by the clustering approach. It is performed in two ways, namely, centralized and distributed [3]. In the viewpoint of centralized clustering approach, the sink node takes charge of collecting the information from wireless networks. Each sensor node is provided with the global knowledge since the sink node is limited from the aspects like energy constraints and storage constraints. Finally, the sink node estimates the cluster heads (CHs) and also its members. However, it is not suitable for the optimal-based large-scale environment [4]. The distributed clustering approach makes use of local knowledge wherein each sensor node is capable of electing the CHs on the basis of requirements. In the case of heterogeneous sensor networks, the deployment of static and dynamic clustering approach has brought the challenges like network congestion, heavy traffic rate, and undersampling and oversampling of the cluster centers.

3.2 Related Study

This segment discusses the reviews of existing techniques. Different sensor nodes are placed on the wireless environment under multi-hop communications [5]. Sink nodes have consumed an additional energy to transmit/receive the data packets. Energy hole issue is one of the vital concepts which was evaluated under AODV, DSR, and TORA protocols. The results have stated that the AODV and DSR protocols performed better than the TORA from the aspects of packet delivery ratio, throughput, and overheads. Though it has improved the network lifetime, the increased sensors node in the topology has lowered the performance. Reaching the base station (BS) has become quite complex in large-scale networks. This leads to the routing problem which was resolved by the protocol, named On-Hole Children Reconnection (OHCR) and On-Hole Alert (OHA) [6]. The connectivity factor between the sensor nodes was efficiently handled under energy metrics. Compared to Shortest-Path Tree (SPT) and Degree-Constrained Tree (DCT), the suggested technique has achieved 75% increased network lifetime. Further, an energy-efficient LEACH protocol [7] was studied to improve the residual energy. Each deployed sensor node was projected into a clustering process. Depending on the instruction given by CHs, the resources are optimized. Though it has reduced the energy-consumption rate, the traffic flow between sensors under the clustering process is not explored.

Quality-of-Service (QoS)–based routing protocols [8] were introduced under multi-objective functions. Here, heuristic-based neighbor selection models were formulated under geographic-based routing models. It has followed the distance, delay, and path-based metrics to obtain optimal routing path. It has significantly reduced the network consumption, yet the network congestion rate becomes increased. Several offloading computational algorithms [9] were introduced to effectively utilize the route-discovery mechanisms for building topologies, CH formation, and CH selection. The time taken for CH selection is higher while discovering the routes. It develops an overhead over the protocol. Owing to it, several clustering-based routing protocols were developed using non-deterministic approaches [10]. PSO protocol for Hierarchical Clustering (PSO-HC) was designed to improve the lifetime of the CHs as well as network scalability. System has reduced the CH, and the link quality of networks cannot be studied.

With the baseline of PSO, multiple-sink placement algorithm [11] was suggested for encoding the particles and evaluating the fitness. Depending on the hop count, the multiple hop count was employed for energy-efficient systems. The position of the sink has significantly depleted the energy and thus, heuristics models were used for finding the minimum sink utilization. Then, Enhanced PSO-Based Clustering Energy Optimization (EPSO-CEO) algorithm [12] was suggested to improve the local searching algorithms for CH selection. With the help of multi-hop routing protocols, the network lifetime and the consumed energy were enhanced. The data collection and aggregation methods have to be enhanced. Delay-sensitive–based multihop routing protocols were established by [13]. Here, an end-to-end delay was improvised using probability blocking mechanisms. Different numerical simulations were formulated for relay nodes and thus reduced the end-to-end delay. Though QoS models are improved, the probability of the hop count was increased in small-scale networks.

Different optimization techniques were studied by different researchers. Since the energy transmission/consumption [14] remains to be challenged, a different numerical solver was suggested. CH selection was done by the clusters rate, hop-count rate, and the relay nodes. Cluster-based aggregation mechanisms were designed for inter-cluster and intra-cluster communication. Network overloads [15] were re-formulated to reduce the congestion near cluster nodes. However, extracting the CH nodes is not properly defined. Energy consumption rate was improvised the position of the sink nodes. Ring routing protocols were introduced to reduce the overhead of the mobile sinks. System has increased the packet delivery rate on small-scale networks. Along with the similar objectives, genetic algorithm [16] was studied for network lifetime enhancement and the cost minimization. A protocol, named Genetic Algorithm–based Energy-efficient Clustering Hierarchy (GAECH) was designed from three aspects, viz., First Node Die (FND), Half Node Die (HND), and Last Node Die (LND). The estimated fitness function ensured a well-balanced cluster formation which increased the lifetime and stability of the node. While, the degree of a node might affect the performance of the network lifetime, which is not focused.

In the case of virtual networks, wireless connectivity plays a key role in CH formation. The deployment of Weakly Connected Dominating Set (WCDS) [17] has explored the proper utilization of the CH selection. The network edge and balancing of loads were improved in this study. The system has obtained better utilization of CHs, irrespective of the node size. Similar approaches were explored in the game theory applications [18]. Most of the routing protocols were designed to overcome the issues of energy consumption, delay, and throughput rate. In coordination to this, game theory was used to enhance the packet delivery rate. By optimizing the route establishment time incessantly using hierarchical routing protocols. Though it improved packet delivery, multi-objective–based decision-making process is not focused. Wireless network application is widely inclined in the field of Internet of Things (IoT). Again, the clustering algorithm has been stimulated in the IoT applications. It was explored in the LEACH protocol [19] and obtained 60% development in throughput, network span and residual energy. Similarly, game theory–based energy-efficient clustering routing protocol (GEEC) [20] was developed to balance the energy efficiency and the lifespan without compromising the QoS of wireless networks. Topology formation has resolved the energy repletion issue with better cluster forming. Evolutionary game models have significantly detected the behavior of the nodes that has resulted in better control messages.

Owing to it, an energy-efficient data aggregation scheme for clustered WSN (EEDAC-WSN) [21] was established for small-sized control frames of cluster member nodes. By monitoring the stability of the node, the member nodes are efficiently communicated in the networks. Compared to the LEACH protocol, the suggested protocols have lowered the delay rate even for small-scale networks as well as large scale networks. Induction trees in hierarchical-based clustering nodes [22] have been studied to resolve the inference problems during induced tree formation. Induced tree of the crossed cube (ITCC) was designed on the basis of the degree of the graph nodes. It was understood that the Voronoi cell has monitored the node’s behavior. Though it claims for reduced delay rate, the use of relay nodes is higher. Since the WSNs are decentralized, the information transformation takes via a clustering algorithm. During clustering-based communication process, the sensor node takes maximized energy which was resolved by WEMER protocol [23]. Though it was concentrated on improvising the gateway nodes, the congestion between those nodes are not concentrated. To makes the other pervasive more effective WSN technologies growth perceived a demand for IoT applications. Over world-wide IoT [19] is utilizing for providing the consistently to network. In addition, for guaranteeing the ubiquitous communications, the IoT follows the principle [20]. Moreover, the IoT engaged with applications, which includes smart home, cities, and agriculture military [21]. In various IoT applications, for improving the data transmission, scalability, and the energy efficiency in the WSNs, the BS received the sensed data that sensor node passes to that makes network more possible. Based on the above-mentioned requirement, that may guarantee latency, loss rate, the minimal energy consumption, and delay in IoT with the development of multipath-aware routing protocol is the essential prerequisite [4].

For the IoT application, the WSN is responsible to collect sensed data for the IoT application [12]. The sensor nodes are here, initially scattered in the IoT, and then sensed data to the sink (gateway) node of the Wireless Sensor Node, the sensor nodes periodically forward back this. After that, the sink node utilizes gathered data for producing fruitful information. Therefore, those routing protocols are classified into two categories, namely, flat-based and cluster routing protocols. While comparing the energy consumption on both, the cluster-enabled protocols have big advantage [14]. In the sensor nodes cluster, one node is selected as the CH, and the remaining are cluster members. To forward the data directly to the sink node, these cluster members not able [13].

Wireless Sensor Node consists of these constraints, such as power, processing memory, transmission range, and bandwidth availability. To improve nodes lifetime as well as energy is the challenge [16]. In this environment, the resources should be follow by the routing protocol, such as manage enable fast convergence, lossy links, energy saving, and prevent routing loops. The secure routing protocol is introduced for effective data routing more, to improve the clustering performance and sensors positioning [15, 18]. In addition, due to adaptive communication patterns, sensors, and lightweight routing protocols, the existing routing protocols failed to apply the WSN directly. As ad hoc topology used by WSNs without identifying path, infrastructure, and then forwarding data to the sink is the quite difficult and interesting task [7]. Moreover, for extending heterogeneous WSNs, lifetime effective device placement, routing protocol, and the topology management techniques are introduced [17]. Nowadays, various protocols, namely, gradient-based routing, rumor, and energy, are introduced, termed as the data center–oriented protocol [18].

3.3 Clustering in WSN

Clustering technologies play a major role in the WSNs which have assisted in improving the performance of the network. The WSNs are classified into two networks, viz., flat networks and clustered networks. The clustered networks are widely adopted to enhance the scalability, reliability, and availability of the wireless networks. Figure 3.1 presents the clustered WSNs [24].

Initially, the sensor nodes are distributed arbitrarily in the wireless environment. CH is selected by the deployed sensor nodes based on the energy parameter. It helps to collect, aggregate, and transmit the data to the sink nodes. This takes the hierarchy structure of the sensor nodes. It also diversifies the member nodes and the CHs. The other parameters in clustering are the cluster number/size, complexity of algorithms, overlapping, relay-overhead in inter-cluster, and routing policies. The clustering algorithms were divided into two types: probabilistic and non-probabilistic algorithms. The CHs are elected by its co-sensor nodes of assigned probabilities known as probabilistic algorithms. Whereas, CHs are selected under a certain criteria as well as proximity of sensor nodes known as non-probabilistic algorithms. In the perspective of a collaborative environment, with the help of nearest cluster nodes, the data is collected, aggregated, and transmitted to its sink node.

3.4 Research Methodology

This section presents the methodology of the research study. A novel routing protocol, using Artificial Intelligence (AI) uses with the objectives of enhancement of network lifetime and reduced usage of cluster number/size. The proposed phases are presented as follows:

3.4.2 Clustering Approach

This phase aim is to increase the network lifetime of the deployed sensor nodes. It follows the static as well as dynamic wireless environment. Here, the clustering has been done on demand basis, i.e., whenever a specific sensor area initiates for transmission purpose, at that time, the CH has elected. Let us assume, network lifetime of a sensor node is denoted as, N_Lifetime(i). The elapsed time taken by nodes until the first node depletes its energy is given as N_ElapTime. Collectively, it is represented as follows:

(3.1)

The AI-based clustering eliminates the demerits of static and dynamic clustering approaches, by not performing clustering at each round. Based on the sync_pulse window, the clustering process is initiated for the particular sensor area. The below pseudo-code depicts the working of AI-based clustering approach. Algorithm 3.1 explains how clustering is performed using AI technique.

Once the lifetime of a network is preserved, a CH node has been selected. Here, AI-based clustering approach is employed as an energy-efficient routing process in WSNs. The objective of the AI approach is to elect the CH from the set of sensor nodes. In association with the above process, the nodes are aware of when to initiate the clustering process. Let us assume that set of clusters is represented as C = {c₁, c₂ c_n}. The deployment of AI-based clustering is to efficiently utilize the energy of the sensor nodes. Thus, objective function of energy minimization is given as follows:

(3.2)

where

is the j’s degree of a node on cluster i.

is the distance between node j and the midpoint of cluster i.

Once the distance between nodes is minimized, then the energy consumption is also minimized. Since the CH election is done by rounds, each round performs data transmission operation. Initially, the nodes are assigned with the degree, and thus, it is used for the formation of a clustering process. The proximity value of each node is estimated. If the value of a node is closer to the proximity value, then it is labeled as cluster C₁. Likewise, all sensor nodes in the application area are examined. Once the clusters are formed, the sink node selects the nearest cluster center C_i to become CH and then the information of CH node was broadcasted. The total number of clusters C is computed as follows:

(3.3)

In some cases, the non-CH at particular time T may attempt to transmit the data packets to sink nodes. Thus, it can deplete the energy. In order to eliminate this scenario, an optimal CH has been elected by the present CH at each round. At the initial round, the sink node selects the CH. The CH is selected, and then, the data transmission process is scheduled by TDMA.

3.5 Conclusion

Rapid innovations of wireless technologies have impressed the researchers to delve into the study of wireless-based IoT systems. A unique feature is the reliable monitoring services, increased network lifetime, and minimized energy consumption rate. However, a complete solution is possible due to the issues like congestion and overload of the network scenarios. In this study, an energy-efficient hybrid hierarchical clustering algorithm for wireless sensor devices in IoT is designed. It is explored by two phases, namely, CH selection using AI approach and shortest route path finding using AI-based energy-aware routing protocol. Our main novelty is the clustering process is initiated on the received request from the sensor nodes. It eliminates the traffic analysis caused during clustering analysis.