Optimal cluster-based energy efficient routing scheme for QoS aware IoT-enabled wireless body area network

Abstract

Wireless Body Area Networks (WBANs) are an integral component of contemporary IoT-driven healthcare and can enable wearable sensors to continuously monitor a patient’s health. Despite their usefulness, routing data in WBANs is challenging because of factors such as tight energy restrictions, frequent node movement, congestion and unreliable trust between nodes. These problems often lead to lower network performance and reduced system life. Clustering techniques may be useful in enhancing energy consumption and scalability, however there are numerous pre-existing approaches, yet they still face issues such as premature convergence, unbalanced workloads and poor selection of cluster heads (CHs). To overcome those limitations, this work proposes a new QoS-aware, energy-efficient clustering-based routing scheme (QEEC-Routing) which combines three new algorithms. The Modified Raccoon Optimization (MRO) algorithm forms well-balanced clusters to distribute the energy usage more evenly. A Two-level Quaternion-Valued Recurrent Neural Network (TQV-RNN) is used to calculate adaptive trust levels to obtain more accurate CH selection. The Improved Hypercube Natural Aggregation (IHNA) algorithm then finds the most reliable routing paths, even with nodes in motion or congestion in the network. Tests conducted in NS3 simulator indicate that QEEC-Routing reduces energy consumption by 51.5%, improves packet delivery by 6.5% and increases the overall network lifetime by 14.9% as compared to current approaches. Altogether, the proposed design proposes a more reliable, energy-aware, and trust-conscious communication strategy that can be used for real-time IoT healthcare applications.

Introduction

The healthcare sector is currently experiencing big problems due to the sudden increase in the number of elderly people, prevalence of chronic diseases, and high medical costs¹. For example, healthcare expenditure in the United States announced around $3.6 trillion in 2019 and is still on the rise². In many developed countries, people grow old in special places of care rather than at home with their families and this is forming an increasing need for continuous and remote health monitoring systems. In that context, the QoS in the transmission of medical data has become very important to ensure the delivery of patient information in a timely and reliable manner³. Because medical data may vary in degree of urgency, it is important to prioritize medical data that may be imperative, such as electrocardiogram signals, over less time-sensitive parameters such as body temperature^4,5. Any delay or loss in transmission can have a major impact on the accuracy of diagnosis and the safety of the patient. Therefore, intelligent planning of data transmission and priority-aware transmission are indispensable, especially in WBANs⁶. Gateways in WBANs are intelligent intermediaries in the network with data processing capacities capable of guaranteeing prioritized and secure transmission of data.

There is a huge need for systems that allow to continuously monitor the health status of older persons and patients and to exchange health information easily with distant caregivers or healthcare institutions⁷. Consequently, more and more attention has been given to professionals finding ways to make remote medical care and real-time patient monitoring economically viable and efficient. To meet these demands, WBANs have become a promising technology in next generation healthcare systems⁸. A WBAN is a network of low-power, intelligent and miniaturized nodes of biosensors that wirelessly collect and transmit physiological data within or around the human body^9,10. These sensors can be used to measure various physical parameters, such as heart rate, body temperature and blood pressure, allowing for the continuous monitoring of critical and abnormal health conditions.

The confluence of WBANs with the IoT further improves the automation of healthcare, enabling the intelligent acquisition, analysis, and communication of data^11,12. The IoT is a world research border that enables seamless interconnectivity between smart devices which are able to exchange data in real time via cyberspace. With the advent of the 5G technology and future 6G communication technologies, IoT-enabled WBANs are expected to provide high-speed, low-latency and energy-efficient data transmission^13,14. The IoT ecosystem continues to grow at an exponential rate with an expected growth rate of more than 50 billion interconnected smart devices by 2022, having a transformative effect on healthcare as well as the global economy¹⁵.

The IoT-WBAN architecture will generally contain multiple nodes of biosensors, which will be used to monitor vital signs and then send the collected data to a remote medical server or cloud which can further analyze the data¹⁶. Given the critical nature of medical applications, these operations need to be performed smartly to make them energy efficient and reliable. Routing is at the heart of this process, finding the most energy efficient and congestion free communication paths between the sensor nodes and gateway^17,18. Numerous routing protocols have been proposed, e.g. temperature-aware, delay-tolerant and QoS based protocols¹⁹. However, existing protocols still suffer from major limitations, e.g. uneven distribution of energy, frequent re-clustering, higher delay under mobility, and unreliable trust management. These shortcomings often lead to network congestion, packet loss and lower QoS, especially in heterogeneous traffic environments²⁰.

In order to address the above-mentioned challenges, this work proposes a QEEC-Routing scheme designed specifically for IoT-enabled WBANs. The proposed framework combines three intelligent algorithms to improve the clustering stability, trust reliability and routing performance. The proposed QEEC-Routing scheme combines three intelligent components that significantly improve the clustering stability, trust reliability and routing efficiency in IoT-enabled WBANs. First, the MRO algorithm achieves adaptive cluster formation by dynamically organizing the sensor nodes into balanced clusters according to the energy levels and the proximity, which aims to minimize the energy consumption and prolong the overall network lifetime. Next, the TQV-RNN model measures the trustworthiness of each node based on several behavioural and contextual parameters, which can be used to correctly identify the most reliable CH for secure data aggregation. Finally, the IHNA algorithm optimizes the data routing paths, to find the best energy-efficient and congestion-free path between the data sources and the mobile gateway, ensuring low-latency and high-reliability communication. Together these three algorithms constitute a coherent hybrid framework for adaptive clustering, intelligent trust management and QOS-aware routing in energy constrained WBAN environments.

The reason for the adoption of the MRO algorithm is that the nodes of WBAN have the characteristics of frequent change of topology, energy imbalance, and nonlinear fitness relationship among distance, residual energy, and delay. Traditional algorithms, such as PSO, GA or CSO usually converge too early or do not find a good balance between exploration and exploitation under the dynamic limitations of WBANs. The MRO proposes a dual zone search mechanism (potential and visible regions) with adaptive memory updating which dynamically adjusts the cluster boundaries depending on the node mobility and link fluctuations. This mechanism provides load balanced clustering and avoids the problem of local minima stagnation, which leads to a direct improvement of energy uniformity and network lifetime compared to about 41–55% of other existing cluster based WBAN routing approaches. Unlike the Improved Pelican Optimization implemented in LEOC-MP that has no adaptability to dynamic mobility and congestion scenarios, MRO adaptively reconfigures clusters according to changing energy and link quality metrics. This allows the continuous self-organization to be ensured, and QoS aware stability under the circumstance of mobile WBAN, which makes this available to be well suited for real-time heterogeneous medical care monitoring environments.

The TQV-RNN was designed to address the shortcomings of scalar trust models and linear aggregation methods that cannot successfully model multi-dimensional node behavior, such as mobility, signal strength, congestion, and packet success rate in real time. The quaternion representation allows the four correlated trust parameters to be processed as a single hypercomplex unit, retaining the relationships between them in space and time when the parameters are computationally processed. The recurrent structure further models the temporal dependencies and captures the nonlinear variations of the node reliability, achieving stable and accurate CH selection, even when the sensor states are fluctuating in a fast manner. Unlike conventional fuzzy or weighted-sum models of trust computation which involve manual tuning of their weights and are unable to learn temporal correlations, the TQV-RNN learns dynamically changing variations in trust and learns from historical network states. Consequently, it helps improve the accuracy of trust by about 23–28% and also helps to greatly reduce the CH misclassification which occurs compared with conventional fuzzy or probabilistic methods, which achieves reliable data aggregation and transmission within the network.

The choice of IHNA algorithm is to solve the multi-objective routing optimization problem with delay, energy, link reliability, and congestion in consideration which is not well addressed by the traditional routing algorithms such as AODV or DSR based on a single metric (hop count) based decision. IHNA uses the hypercube-based search space formulation which enables the parallel exploration of multiple path regions, effectively reducing the redundancy of routes and minimizing the route convergence delay. It uses dynamic scaling and centroid adjustment to preserve routing stability even under high mobility, under which reactive protocols like AODV and DSR will cause excessive re-routing overhead and packet loss. In contrast to the moderated puffer-fish optimization used in LEOC-MP, which does not have adaptive congestion control, IHNA incorporates self-adaptive scaling to minimize the routing overhead by about 35–40% and the latency by over 50%. This guarantees sustainable QoS and reliable multipath routing in dense and mobile WBAN environments with a focus on high traffic conditions.

Together, the MRO, TQV-RNN and IHNA algorithms constitute a hierarchically integrated optimization framework to perform the adaptive clustering under high mobility, intelligent CH selection based on multidimensional trust evaluation, and QoS-driven multipath routing with congestion awareness simultaneously. This hybrid synergy is an effective solution to counter major shortcomings of existing WBAN routing schemes, including energy imbalance, congestion and trust instability. Compared to the previous LEOC-MP framework²⁰, benchmark schemes, including T-MAC, MS-MAC, and DT-MAC, the proposed QEEC-Routing shows remarkable improvements in providing energy-efficient, delay-minimized, and trust-reinforced data communication for superior IoT healthcare monitoring systems in terms of QoS and network sustainability. The main achievements of the proposed work are as follows:

• To develop a QEEC-Routing framework for IoT-enabled WBANs that ensures reliable, low-latency, and energy-efficient communication under dynamic network conditions.

• To design an adaptive cluster formation mechanism using the MRO algorithm, enabling balanced energy consumption among nodes and extending overall network lifetime.

• To establish a trust-driven CH selection model based on the TQV-RNN that dynamically evaluates node reliability through multidimensional behavioral parameters, enhancing data integrity and security.

• To optimize data transmission paths using the IHNA algorithm, ensuring congestion-free, energy-efficient, and QoS-compliant routing between sensor nodes and mobile gateways.

Related works

Awan et al.²¹ proposed a new multi-hop routing protocol for decreasing congestion in IoT-enabled WBAN environments. Their approach is based on developing a routing strategy that enables sensor nodes to send data more efficiently, reducing delays and facilitating the delivery of high priority information, while having the energy consumption low enough to extend the life of the network. For routine, non-urgent traffic, the protocol chooses the next hop taking into account three major factors: how much energy the potential forwarder still has, how much it is suffering from congestion and what the signal-to-noise ratio is between the source node and the forwarder²². While this approach mitigates the congestion and delays for the emergency data, it is not adaptive clustering or cross-layer energy optimization. The proposed QEEC-Routing solves these problems by merging energy-conscious clustering and fuzzy-based routing, resulting in the balance of energy consumption and reduced latency.

They also include data collection and filtering methods to help reduce network congestion and reduce energy consumption. When the data is important to a patient’s safety, the system switches to a priority-driven routing approach, so that delay and confusion is minimized and throughput is maximized in an emergency situation, thus ensuring that the information about life-saving devices is able to reach its destination in a timely manner. Bulaghi et al.²³ proposed a framework called SENET that is aimed at integration with IoT-based WBANs for reliable data exchange in a closed-loop system. In order to enhance the determination of routing paths, the authors used several optimization strategies, including world competitive contests (WCC), particle swarm optimization (PSO), ant colony optimization (ACO), and genetic algorithms (GA). Their architecture not only achieves noticeable energy reductions in WBANs, but also points out the power of the WCC method for choosing appropriate sensor locations in the system. Although the SENET framework is effective in conserving energy using metaheuristics, it is limited in terms of notable computational demands and lack of dynamic optimization. In contrast, QEEC-Routing makes use of lightweight energy measures coupled with adaptive decision-making to guarantee faster convergence and scalability in real-time.

The WCC approach is unique in terms of its reduced energy consumption²⁴, its capacity to reduce the number of essential sensors and its general reliability. In various test scenarios, it scored an average of 38.44 in its performance, which was better than that of the competing techniques. Dhanvijay et al.²⁵ proposed secure aware mobility management protocol named CoSMP for IoT based WBAN systems. Its primary aim is to secure sensing data that mobile nodes generate and secure communication between each node and the web client. CoSMP does this by establishing pairwise key between networks and AES-CFMAC algorithm for encryption and authentication. Their evaluation found that the protocol kept very low delay values − 303.55 ms, 1.6 ms, 700 ms and 80 ms under different conditions. While CoSMP provides the security of mobility, it does not provide for energy balancing and routing flexibility in densely populated WBANs. QEEC-Routing addresses these problems by using dynamic cluster head selection and energy-efficient multi-hop routing, in order to extend the network’s life.

Khan et al.²⁶ presented an energy harvested, cooperation-based routing protocol (EHCRP) for IoT driven WBAN systems²⁷. Their design considers several important factors affecting the routing performance such as the remaining battery power in each sensor node, the number of hops between each node and the sink, the node mobility, the signal-to-noise ratio, and the bandwidth available in the network. Using these inputs the protocol estimates the cost of each possible path and selects the best route forwarder node accordingly. This decision making process is useful for ensuring the reliability of data delivery and for making overall communication within the network more efficient. EHCRP improves reliability of forwarding across cooperation, but requires complex path cost computation. The proposed method helps streamline this procedure by introducing a fuzzy treatment of the quality of the link and available energy to reduce the overhead and maintain the precision.

Misra and colleagues²⁸ proposed a MAC protocol with an aim to reduce energy use in WBANs. One of the problems they solved was that the body-worn sensors sometimes began transmitting when they didn’t want to, thereby raising the possibility of their signals colliding with on-going transmissions from the hub, or with frames sent by other sensors. To overcome this, the authors proposed an adjusted superframe layout which can help prevent these unplanned overlaps. An arranged access instrument is used to manage the criticality of each node in an effective way, minimize the probability of a collision, and also boost the performance of the network. While the optimization of the MAC layer can help to minimize collisions, it does not provide adaptability at higher layers and awareness of routing. QEEC-Routing QEEC-Routing enhances the efficiency by adopting a combined approach of clustering and routing optimization at the network layer.

Ullah et al.²⁹ proposed an efficient cross-domain admission control system tailored for IoT-based WBAN environment. Their method is designed to ensure the system remains both strong and verifiably secure while also keeping the system lightweight and practical. To achieve this, the application provider side is based on certificate-less signcryption, on the other hand, the WBAN side is based on identity-based signcryption, which allows not only secure data handling but also simplified access control. Performance tests have shown that the approach costs only 1.92 ms of computation and incurs a communication overhead of just 1296 bits, proving this approach to be practical and overall efficient³⁰. While cross-domain admission control increases security, it does not address problems of energy limitation and routing efficiency. QEEC-Routing on the other hand promises secure and energy efficient data transmission, by means of coordination across the layers and route selection according to energy considerations.

Izza et al.³¹ presented an improved verification mechanism using IoT-enabled WBAN systems, which is implemented by using the RFI technology. Their ECDSMR protocol hardestens data protection by encrypting the data transmitted from the RFID tag to the reader by using symmetric encryption algorithm after establishing a shared session key³². It also leverages the use of digital signatures that have message recovery capabilities and memorandum links to securely share important patient information with healthcare personnel. Together, these measures make the entire authentication process more secure and reliable. Although ECDSMR improves the authentication, it does increase the computational and communication requirements. In contrast, QEEC-Routing guarantees good security with low computational overheads by using light weight encryption and efficient route management.

Pandey et al.³³ proposed an improved version of the EEAWuR-MAC protocol, which was aimed at dealing with routine as well as time critical data in IoT-based WBANs. Their work investigated how the protocol behaves: they looked at energy consumption, delays and throughput both during normal data (ND) transfers and situations when emergency data transfers instead. In order to cope with these various traffic types, the system is based on a priority-driven CSMA/CA mechanism. In case of emergency when a body sensor gives a report and transmits a back-up request, the WBAN coordinator immediately schedules another transmission opportunity to ensure the urgent data can be delivered without any unnecessary delay³⁴. EEAWuR-MAC is in fact well suited to prioritize essential data while lacking the ability to sustain an adaptive route with variable load conditions. The proposed QEEC-Routing scheme dynamically adjusts routing paths using both residual energy and link reliability.

Anbarasan et al.³⁵ has developed B-DEAH system specifically for IoT-WBANs. This revolutionary system contains a number of important components, starting with patient key registration. They proposed an extended version of PRESENT to effectively handle and secure the registration of patients. An optimizer is used to create clusters and select cluster heads which helps in better organization and structure of the network. The MOORA algorithm is used in order to improve the routing in these clusters to improve the efficiency of data transmission³⁶. B-DEAH improves clustering with the help of the MOORA algorithm, but it is not efficient in the adaptability of mobile environments and varying traffic conditions. In comparison, QEEC-Routing enhances the dynamic adaptability using fuzzy clustering and hybrid optimization based routing approach.

Samriya et al.³⁷ proposed a clustering approach to better utilise energy in IoMT networks. Their solution is based on the chicken swarm optimization technique³⁸ to optimize the way in which clusters are created. A customized fitness function is used for selecting the cluster heads based on a number of parameters such as remaining energy, queue delays, communication overhead, link reliability and the overall importance of each node. By choosing the most appropriate cluster heads, the algorithm contributes to building more efficient clusters, resulting in a consequent reduction of energy consumption and an extended operational lifetime of the network³⁹. Remarkably, it achieves a notable reduction in energy consumption of 3–7%, highlighting its promise for creating energy-efficient and durable IoMT networks. While the chicken swarm-based clustering enhances energy efficiency, it overlooks aspects of routing and delay optimization. QEEC-Routing combines clustering and routing into a single framework to reduce delay and boost throughput. Table 1 gives the summary of research gaps which gathered from the existing state-of-art routing schemes.

Table 1 Summary of research gaps.

Problem methodology

Table 2 lists the most important symbols and notations used in the mathematical formulations of the proposed QEEC-Routing scheme. It specifies all the variables, parameters, and coefficients in the MRO, TQV-RNN, and IHNA algorithms to promote a clear and consistent interpretation. The following table can be used as a fast guide to the equations associated with clustering, trust computation, and optimal route path choice.

Table 2 Symbols and notations used in mathematical formulations.

Research gaps

IoT-WBANs have a significant role in real-time health monitoring, particularly for patients suffering from chronic ailments or are in a critical care situation. The requirement of continuous and rapid data transmission in these situations cannot be highlighted enough. QoS aware routing strategies play a pivotal role in the reliable and timely delivery of vital medical information to healthcare providers. This in turn makes it easier for swift and potentially life-saving intervention. Moreover, the nature of WBAN devices is such that they are resource constrained, frequently running on limited battery power and processing power. In order to maximize the operational life of such campaigns with minimum energy consumption, the use of efficient routing techniques becomes imperative. QoS-aware routing is helpful to optimize the usage of resources, thus contributes to the prolonged life expectancy of the network as well as seamless monitoring of the network. Furthermore, IoT-WBANs struggle with the dilemma of managing heterogeneous data traffic, which can be anything from nonemergency to emergency medical data. QoS-aware routing mechanisms are applied to deal with this diversity. These mechanisms are designed to prioritize and effectively process the range of data types into the system to ensure that emergency data take precedence, even when non emergency data are present. A previous work⁴⁰ proposed the link quality and energy-efficient optimal clustering-multipath (LEOC-MP) arrangement, which was intended to improve the link quality, network’s life and reliable multi-path data transfer in WBANs. LEOC-MP used some innovative features such as improved pelican optimization (ICO) for clustering, auto-regressive probabilistic neural network (AR-PNN) for CH selection, and moderated puffer-fish optimization (MPO) algorithm for opting the optimum and straight path for data transfer. However, the LEOC-MP routing scheme had some issues related to congestion which made it not suitable for the IoT environment, and it needed to be further improved. In order to overcome the congestion problem, Zeb et al.⁴¹ suggested an improved version of the MT-MAC procedure, called DT-MAC, that is designed to guarantee the successful delivery of messages in IoT-WBANs. The protocol includes an instrument of abdication of a node among the simulated clusters in order to maintain the honesty of the system and exploits the concept of smallest connected controlling set for creating the network, optimizing energy operation. Comparative evaluations against reference algorithms, like MT-MAC, showed good improvements. The DT-MAC protocol showed an increase in latency of about 3% which equaled a 13–17% increase in delivery and a response-time reduction of about 15%.

The following research issues are provided in the overview of the literature review^{21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40}. In general, IoT-WBAN devices have limited computer resources. Achieving this required QoS while within these constraints, particularly for intensive applications like QoS monitoring, can represent a major challenge²¹.To keep a QoS can be challenging when multiple devices send at the same time or the network is congested²³. A new routing protocol should address the methods to deal with congestion situations and preserve the QoS of critical data. Determining the best suited QoS metrics and thresholds for IoT-WBANs is difficult^23,26,28. The routing protocol should specify what an acceptable QoS is and how to effectively measure and monitor it. Ensuring the QoS requirements are not violated during the entire end-to-end data transmission process from the sensor nodes to the central monitoring system can be complex^23,29. This involves co-ordination between WBAN devices, gateways and wider networks. Patients may move or change positions^{21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40}, which could affect the quality of the communication links. Creating a routing protocol that can continuously access network conditions in real time and be able to change the routing decisions according to the QoS metrics adds complexity^35,40. It has to be able to evaluate and modify the QoS during the data transmission.

System architecture of proposed QEEC-Routing

A complete description of the proposed QEEC-Routing approach for IoT-based WBANs can be found in the system layout in Fig. 1, which presents the entire workflow of the proposed method. The process starts with deployment of sensor nodes throughout the body, where they monitor important physiological signals and other health-related information. This gathered information is used as the basis for the routing decisions that are made throughout the network. After, clusters are formed based on the MRO algorithm and cluster heads (CHs) are selected carefully to balance the energy usage and extend the life of the network. These CHs take up the responsibility of collecting data from their cluster members and sending these information effectively so as to ensure smooth flow of information. Since this system is used in a healthcare context, it is particularly important to ensure that each node is reliable.

Fig. 1

Complete system setup used in the QEEC-Routing method for IoT-driven WBAN applications.

To overcome this challenge, the system applies a TQV-RNN model to estimate the level of trust of each node. These trust scores depend on factors such as the frequency of the movement of a node, the strength of the received signals, or the congestion of the links. Nodes that gain a higher trust value are selected to act as either cluster heads or forward important data, this helps to strengthen the overall reliability of the network. For routing, the IHNA is used to determine the most efficient route through the network. It considers a number of operating conditions and constraints to determine a route that best facilitates smooth data delivery. Once this optimal route has been established, the network moves on to the actual data transmission, which is especially critical in the case of medical data such as vital signs or other health data. During the entire communication process, QoS parameters are monitored to ensure they do not go out of acceptable limits. The essential function of this routing strategy is to enable real-time health monitoring by minimizing the delays and ensuring performance standards required for reliable healthcare applications.

Proposed QEEC-routing

In this segment, the detailed working process of QEEC- Routing system for IoT- WBANs is presented which consists of cluster formation using MRO, trust degree computation using TQV-RNN, and path finding using IHNA algorithm. Together, these elements constitute a comprehensive routing solution that aims at improving network performance and reliability, and this sets the stage for a detailed discussion of the components of the methodology.

Optimal cluster formation

Cluster formation in IoT-WBANs consists of nodes of sensor systems in the form of body-worn devices, in order to route and manage the data in a more efficient way. This step plays an important role in enhancing the overall performance of the network as well as better utilization of the available resources. In order to reinforce this process, a modified raccoon optimization (MRO) technique⁴² is used to effectively form clusters. Because MRO adapts well to constantly changing conditions of WBAN environments, it helps to maintain stable and efficient clusters as the network evolves. By optimizing the creation of clusters, MRO minimizes the amount of energy wasted—an important factor to the lifetime of WBAN devices. It also supports network expansion by adding new sensor nodes with ease. At the beginning of the algorithm, an investigation agent (which is modeled as a raccoon) is placed inside the search space and given an initial exploratory function. During this stage, the possible search region for each agent at the iteration “I” is represented as ${r}_{cl(h)}$ (0 ≤ h ≤ MaxIter). The loading phase starts with this value set to zero iterations with the location of the exploration manager’s starting location defined as “${r}_{cl(0)}$”. The parameter “$op{t}_{J}$” is then introduced representing the random initial placement of the exploration guide

$$opt_{J} = r_{cl(sa)}$$

(1)

In this stage, the algorithm operates with a pool of promising candidate solutions prz, representing the possible locations which can be identified by the search agent (SA) using its sensing ability and the knowledge stored in it. Instead of testing all possibilities, the method only tests a carefully selected set of candidates selected at random, in order to efficiently guide the search.

$$prz_{(sa)} = \{ R_{sc(0)} ,\,R_{sc(1)} ,....R_{sc(nrz)} \}$$

(2)

Now nrz is the number of candidate nodes which are included in the search agent’s range. This value is based on the number of orientation nodes that are necessary to precisely locate the body sensor nodes within the system. In addition, $r_{(sc)(h)}$, 0 ≤ h ≤ NRZ, denotes the randomly chosen functions that represent the current estimated positions of the NLOS nodes before they are fully localized.

$$r_{(sc)(h)} = (p_{0} ,\,p_{1} ,....p_{nrw} ),\quad where\quad 0 \le h \le nrz$$

(3)

In this context, the distance between the search agent’s current location and the potential solution indicates that the the $\alpha (d_{loc} ,\,r_{sc} )$ candidates within prz must meet the required condition.

$$0 < \alpha (d_{loc} ,\,r_{sc} ) \le prz,\quad where\quad 0 \le h \le nrz$$

(4)

Each phase of prz is treated as an important phase that helps to examine the full range of possible solutions with greater precision. For this reason, “$pr{z}_{0}$” is first introduced as a key component of the process and is, generally defined as follows.

$$prz_{0} = R_{sc(h)} |h \in \{ 0,\,1,\,....nrz\} \quad and\quad 0 < \alpha \left( {d_{loc} ,\,R_{sc} } \right) \le prz$$

(5)

The selection of $r_{best(0)}$ is guided by the expansion problem, with the fitness function serving as the criterion for optimization.

$$r_{best(0)} = R_{(g)} ,\quad where\quad R_{(g)} \in prz_{0} \quad and$$

(6)

$$F(R_{(g)} ) = \max \left\{ {F(R_{(h)} )|R_{(h)} \in prz_{0} } \right\},\quad where\quad 0 \le h \le nrz$$

(7)

After generating the reachable population in the previous step, a new population is formed that corresponds to the search agents. Similar to prz, the MRO algorithm introduces another group known as the visible zone population (pvz). This set contains the solutions that can be directly observed and are considered feasible within the ranges defined by rrz and rvz.

$$pvz = \left\{ {U_{sc(0)} ,\,U_{sc(1)} , \ldots ,U_{sc(rvz)} } \right\}$$

(8)

The value pvz represents the count of contenders situated beyond rrz and below rvz but still within the search agent’s scope. The notation $U_{sc(h)}$ (0 ≤ h ≤ nrz) corresponds to arbitrarily selected candidate positions that characterize the present estimated positions of previously localized NLOS protrusions. The set $U_{sc(h)}$ is further modeled as a p-dimensional optimization problem.

$$U_{sc(h)} = (q_{0} ,\,q_{1} , \ldots ,p_{nvz} ),\quad where\quad 0 \le h \le nvz$$

(9)

In the second population assessment and exploitation phase, rvz consistently provides better performance than rrz. If an optimal solution appears between the exploration agent’s current position and the candidate point in the designated pvz region $\alpha \left( {d_{loc} (pvz} \right),R_{sc(h)} )$, it is taken as the improved solution

$$prz < \alpha \left( {d_{loc} (pvz} \right),R_{sc(h)} ) \le pvz,\quad where\quad 0 \le h \le nvz$$

(10)

Similar to prz, pvz is considered an important constraint because it helps identify a precise set of candidate solutions. For this reason, the initial population within the visible region is first mapped out and visually $pvz_{(0)}$ represented as follows.

$$pvz_{(0)} = \{ U_{sc(h)} |h \in \{ 0,\,1,\,....,nvz\} \;and\;prz < \alpha (d_{loc} (pvz),\,R_{sc(h)} ) \le pvz\}$$

(11)

Once the pvz region is formed, the top-performing member within that zone is identified, much like in the prz stage. The best candidate solution in the pvz area is denoted as $T_{best(0)}$. This value is obtained by solving the given optimization problem and using the corresponding training function.

$$T_{best(0)} = U_{(g)} ,\quad where\quad U_{(g)} \in prz_{0} \quad and$$

(12)

$$F(U_{(g)} ) = \max \left\{ {F(U_{(h)} )|R_{h} \in prz_{0} } \right\},\quad where\quad 0 \le h \le nvz$$

(13)

During this loading phase, the safety worth is determined founded on the “Prn” solution.

$${XR}_{b} = 0$$

(14)

In each iteration, h(0 ≤ h ≤ MaxIter), prz ($\,r_{best(h - 1)}$), the highest value in the visible region ($T_{best(h - 1)}$), and the present position of the exploration agent $d_{loc(h - 1)}$ relative to the previous repetition are selected.

$$\begin{aligned} & d_{loc(h)} = d_{1} ,\quad where\quad d_{1} \, \in \{ d_{loc(h - 1)} ,\,r_{best(h - 1)} ,\,T_{best(h - 1)} \} \quad and \\ & F(d_{1} ) = \max ,\,\{ F(j)|j \in \{ d_{loc(h - 1)} ,\,r_{best(h - 1)} ,\,T_{best(h - 1)} \} \} \\ \end{aligned}$$

(15)

On the extra pointer, if the position of the exploration manager is moved by admiration to the body sensor nodes during each main loop iteration $\left( {d_{loc(h)} \ne d_{loc(h - 1)} } \right)$, the ‘Prn’ worth is reset to zero (Prn = 0).

$$XR_{n} \left( {d_{loc(h)} = d_{loc(h - 1)} } \right) \to (XR_{n} + 1) \wedge d_{loc(h)} \ne d_{loc(h - 1)} \to 0$$

(16)

At this stage, the trust level is evaluated during every iteration by comparing the displacement factor with the safety threshold. This comparison helps keep the process from becoming trapped in a local optimum.

$$\alpha \left( {d_{loc(h)} ,\,\,d_{loc(h - 1)} } \right) > rvz$$

(17)

Once the migration phase finishes, the security parameter prn is returned to its initial value of zero.

$$\begin{gathered} \left( {d_{loc(h)} = d_{loc(h - 1)} } \right) \Leftrightarrow (prz_{(h)} \cap prz_{(h - 1)} ) = \varphi \hfill \\ \,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\, \wedge \hfill \\ \left( {d_{loc(h)} = d_{loc(h - 1)} } \right) \Leftrightarrow (pvz_{(h)} \cap pvz_{(h - 1)} ) = \varphi \hfill \\ \end{gathered}$$

(18)

After the main loop runs up to the MaxIter limit, the algorithm identifies the strongest fitness values for both “OPTG” and $d_{loc(h)}$ the comparison model by scanning through the reference nodes to find the most suitable solution. To form clusters more effectively, the fitness of every raccoon search agent is handled with extra care, using multiple reference points to ensure that each agent’s fitness is evaluated in a reliable and meaningful way.

$$fitness_{fn} = z_{1} \times fn_{(1)} + z_{2} \times fn_{(2)}$$

(19)

Both parameters where ‘$z_{1}$’ and ‘$z_{2}$’ are assigned a value of 0.5 so that all features are weighted evenly during the localization process. Within the ROA, In ROA, ‘$fn_{(1)}$’ and ‘$fn_{(2)}$’ framework, these two parameters represent, respectively, the estimated optimal position between the orientation node and the unknown body sensor nodes, and the separation of those unidentified nodes from the issues caused by body-sensor conditions. A concise outline of the cluster-formation procedure using MRO is presented in Algorithm 1.

Figure 2 shows the workflow of the modified Raccoon Optimization (MRO) algorithm for forming clusters in IoT enabled WBANs. The procedure starts with the initialization of the population and random placement of the search agents. This is followed by an iterative calculation of the potential region zone (prz), and evaluation of fitness. Subsequently, optimal clusters are formed using multi-reference fitness assessment in order to optimize energy efficiency and stability of the network.

Fig. 2

Flowchart of MRO-Based cluster formation.

Algorithm 1

Forming clusters with the MRO method

Trust degree computation and CH selection

The process of determining a node’s level of trust involves bringing together a number of factors—how often the node moves, signal strength received by the node and how often the node is in a state of congestion. These inputs are combined to give a trust score which indicates how reliable the node is to take on important network tasks. To do that, a two-layer Quaternion-Valued Recurrent Neural Network (TQV-RNN)⁴³ is applied. RNNs are made to process sequential information, making them suitable for analyzing data that are time-dependent, such as mobility trends and changes in signal strength. Their quaternion-valued counterpart adds this capability to the model by enabling it to work with richer and multidimensional data. In this setting, the TQV-RNN is especially powerful because it is able to find subtle relationships among the different parameters involved in trust assessment. As a result, it produces more precise trust scores, improving the selection of reliable cluster heads within the network. The TQV-RNN models the evolution of quaternion-based states Q(0), Q(1), Q(2), and so on, as it progresses through the evaluation sequence.

$$E_{r} (B + 1) = \left\{ {\begin{array}{*{20}l} {\sigma (s_{r} (B)),} \hfill & {0 < |M_{r} (D)| < + \infty } \hfill \\ {e_{r} (B),} \hfill & {otherwise} \hfill \\ \end{array} } \right.$$

(20)

$$M_{H} (B) = \sum\limits_{r = 1}^{u} {Y_{HG} e_{r} (B)}$$

(21)

where $M_{H} (B)$ is the performing competence of the h-th hidden layer at importance’s.

$$Y_{HG} = \overline{{Y_{HG} }} \quad and\quad Y_{HH} \ge 0,\quad \forall_{H} ,\,G \in \{ 1, \ldots ,V\}$$

(22)

The quaternionic association yields a low storing limit the optimal memory.

$$S_{HG}^{s} = \frac{1}{U}\sum\limits_{\eta = 1}^{S} {\sum\limits_{\xi = 1}^{S} {I_{h}^{\eta } } R_{\eta \xi }^{1} \overline{V}_{G}^{\xi } ,}$$

(23)

Here $\overline{D}_{\eta \xi }^{1}$ signifies the (η, ξ)-entry of the quarter n valued opposite of the medium assumed by

$$R_{\eta \xi } = \frac{1}{u}\sum\limits_{G = 1}^{u} {\overline{I}_{G}^{\eta } } I_{G}^{\xi } = \frac{1}{U}\,\,\,\,\left\langle {X^{\xi } ,\,X^{\eta } } \right\rangle ,\quad \forall \mu ,\,u \in \{ 1,\,2, \ldots ,S\}$$

(24)

Because the network matrix c is invertible, each essential memory component functions as a key element of the TQV-RNN, supporting the model’s predictive training process.

$$M_{H} (Z) = \sum\limits_{R = 1}^{u} {Y_{HG}^{R} e_{G} (B) = } \sum\limits_{G = 1}^{u} {\left[ {\frac{1}{u}\sum\limits_{\xi = 1}^{S} {X_{H}^{\xi } \overline{X}_{G}^{\xi } } } \right]} e_{G} (B)$$

(25)

We explain the basic structure of the TQV-RNN in terms of trajectory validation. Consider a system whose components correspond to the core memory units, where the decoding process of U is carried out and the functions R, σ, and fe {•} are applied piece by piece. The overall workflow of the proposed QEEC-Routing approach is illustrated in Fig. 3. The method for calculating trust levels and selecting cluster heads is outlined in Algorithm 2.

Fig. 3

Detailed operational workflow of the proposed QEEC-Routing process illustrating clustering, CH selection, and optimal path finding.

Algorithm 2

Trust computation and CH selection through the TQV-RNN technique

Given an fundamental value P (0), a TQV-RNN utilizing synchronized apprise is represent as shadows.

$$Y(M) = d(fe\{ X^{*} e(B)\} /u)$$

(26)

$$e(B + 1) = \sigma (IY(B))$$

(27)

In the TQV-RNN model, the time step is derived from the internal relationship between p(z) and the weight assigned under the given conditions. More precisely, we define this as the (η, ξ) component of the inverse of the optimal structure, expressed as follows.

$$R_{\eta \xi } = d\left( {\frac{1}{u}fe\{ \left\langle {I^{\xi } ,\,I^{\eta } } \right\rangle \} } \right),\quad \forall \eta ,\,\xi \in \{ 1, \ldots ,S\}$$

(28)

Besides, to abridge the calculation,

$$u_{H}^{\xi } = \sum\limits_{\eta = 1}^{S} {I_{h}^{\eta } R_{\eta \xi }^{ - 1} }$$

(29)

for all G = 1, …, A then η = 1, …, Y. Therefore, the beginning probable of a TQV-RNN is given by

$$M_{H} (B) = \sum\limits_{\xi = 1}^{S} {Y_{\xi } (M)v_{H}^{\xi } ,} \quad \forall H = 1, \ldots ,u$$

(30)

The grid-vector representation is applied to organize and align the hidden layers within the TQV-RNN. Let the quaternion-based structure correspond to the primary memory elements of the system. We then define the unchanged reference frame along with the quaternion-valued grid that is formed under these conditions.

$$R = d(fe\{ I^{*} I\} /u)\quad and\quad u = IR^{ - 1}$$

(31)

Here innervations competence R and fe {•} are measured in a segment astute way. Assumed the fundamental value Q(0), a QRPNN represents recursively.

$$Y(B) = d(fe\{ I^{*} o(B)\} /u)$$

(32)

$$e(B + 1) = \sigma (CY(B))$$

(33)

where R: [− 1, 1] → f then σ : → t are measured in a fragment wise method.

The trust degree thresholds of CH selection are dynamic in the proposed QEEC-Routing scheme, as opposed to being fixed. The trustworthiness of every node is continually modified depending on several behavioral and communication parameters, such as the packet forwarding ratio, residual energy, link reliability, and interaction frequency. The system also determines the normalized trust score of every node and compares it with an adaptive threshold during each round of clustering, and the adaptive threshold is also recalculated based on the current network situation (average trust value and node density inside the cluster). This dynamic adjustment ensures that the CHs are solely selected as regular reliable and energy-efficient nodes, regardless of the mobility and traffic conditions. The threshold is automatically adjusted to maintain the security, stability, and energy efficiency balance at an optimal level using a self-tuning fuzzy logic controller to avoid the possibility of malicious or weak nodes becoming a CH.

Optimal path finder

After the selection of the cluster head (CH), the optimal Path finding stage becomes crucial for finding the optimum path from the source node to the mobile agent for data transmission. This step ensures reliable delivery, considering the network load, signal quality and overall stability. As deciding the optimal way for IoT-WBAN environments is a complex task, the Improved Hypercube Natural Aggregation (IHNA) algorithm ⁴⁴ is adopted to deal with this operation. IHNA is designed specially to make the choice of path optimal by considering factors like the stability of the links, energy consumption and Quality of Service (QoS). Since IoT-WBANs operate in environments where the conditions are changing regularly, IHNA is designed to respond to these changes, ensuring that the selected path remains efficient and dependable as the network fluctuates. To begin the IHNA procedure, random transmission values $R_{\dim }$ and $p_{d}$ are assigned to the hypercube’s size and center parameters. These values form the basis of the first hypercube in the initial group. Next, search points $B_{pop}$ are produced through a uniform distribution (Un), and these candidates form an initial pool, denoted as P. Each candidate is assessed, and its evaluation is stored in a corresponding vector $F_{vect}$. Once this setup is complete, the size of the hypercube is recalculated using the minimum and maximum levels of the population X, along with the centroid value $p_{d}$. The uniform distribution then generates a new set of points in P, and each point’s $F_{vect}$ quality is computed and recorded. At this step, the best fitness value $p_{best}$ at iteration h is represented by f. The parameter values listed in Table 1 are obtained through the calculations that follow. The sequence vector (Ln) is determined using the following expression.

$$Ln = Min(P\,\,bounds)$$

(34)

$$Un = Max(P\,\,bounds)$$

(35)

Dimensions of m-dimensional HCs:

$$c = Un - Ln$$

(36)

Central standards are compute as follows.

$$p_{d} = \frac{(Ln + Un)}{2}$$

(37)

The new position $p_{\,new}$ is obtained through a local exploration step defined by $p_{\,new} = p_{new}^{best} + \rho \Delta F$ where F denotes the detached function and ρ lies within [0, 1]. Once the best score is identified, it yields a revised value, $p_{d}$ the phase concludes with the generation of an updated opinion, explained below

$$p_{d\,new} = \frac{{(p_{d} + p_{new}^{best} )}}{2}$$

(38)

$$p_{best} = p_{new}^{best}$$

(39)

$$p_{d} = p_{d\,new}$$

(40)

To construct the hypercube, the algorithm uses values extracted from P, namely $R_{\dim }$, c, and $p_{d}$. The next center point is computed by averaging the current and best positions: $p_{d} = (p_{d} + p_{best} )/2$.

This standard averaging technique helps prevent rapid convergence to a nearby local minimum $F_{best}.$ Meanwhile, the mean fitness is determined using $F_{Mean} = F\left( {\left( {p_{last - center} + p_{best} } \right)} \right)$ which reduces unwanted fluctuations during the search

$$p_{b} = \frac{{(p + p_{d} )}}{c}$$

(41)

Regularized $p_{Min}$ (current x for minimum):

$$p_{Min\,b} = \frac{{(p_{Min} - p_{d} )}}{c}$$

(42)

Regularized coldness (should be restricted by 0 and sqrt(m)):

$$c_{b} = \frac{{Sum\left( {\left( {p_{b} - p_{Min\,b} } \right)^{2} } \right)^{0.5} }}{c}$$

(43)

Regularized coldness (should be bounded by 0–0.1):

$$c_{bb} = \frac{{c_{b} }}{\sqrt a }$$

(44)

This represents the natural variation among the $c_{bb}$ best candidate solutions. Once the conditions are evaluated, and if they are not met, the algorithm proceeds to the search-space exploration stage. The optimal design vector is then calculated using the following expression.

$$R_{\dim } = R_{Dim} * TH$$

(45)

$$p_{d} = p_{best}$$

(46)

The center of the hypercube is updated to the new $p_{best}$, and its size is adjusted accordingly. With this update, the hypercube maintains its scale for small movements but gradually contracts when needed. The detailed steps of the optimal path-finding process using the IHNA method are outlined in Algorithm 3.

Figure 4 illustrates the workflow of the IHNA algorithm, which is employed to select the optimal path in the proposed QEEC-Routing scheme. The initialization of a random population and the creation of sequence vectors according to the hypercubes starts the procedure. It goes as follows: the normalized parameters and adaptive coefficients are calculated, and the candidate solutions are iteratively optimized. Ultimately, the algorithm evaluates the fitness of every path and selects the path that is most energy-efficient and reliable for the transmission of data.

Fig. 4

Flowchart of IHNA-based optimal routing path selection.

Algorithm 3

Optimal path identification with IHNA

Results and analysis

This section presents the results of our experimentation and provides a comparative analysis between the proposed routing approach with some existing IoT-WBAN routing protocols under different simulation environments, i.e., the impact of mobile nodes and varying node movement change. The proposed method was developed and tested in NS3 network simulator. Its performance was then compared with several established protocols for this purpose with T-MAC⁴⁵, MS-MAC⁴⁶, MT-MAC⁴⁷ and DT-MAC⁴¹. To evaluate the overall efficiency and reliability, we have investigated several performance measures, e.g., energy consumption, packet delivery ratio, network lifetime, throughput, delay and routing overhead. These metrics form a solid basis for understanding the impact of the proposed technique on the network behavior and pave the way to the detailed discussion that follows.

In the proposed system, all parameter tuning for the MRO, TQV-RNN and IHNA algorithms was performed using a systematic sensitivity analysis to ensure reproducibility and best convergence. For MRO, the most important parameters, population size (N = 30), maximum iterations (Itr = 100), exploration coefficient (α = 0.6) and exploitation coefficient (β = 0.4) were adjusted, which were experimentally tuned to mediate the global and local search performance. In the TQV-RNN, the learning rate (0.001), batch size (64), recurrent depth (three layers), and quaternion activation function (tanh-based) were optimized using cross-validation to get a high accuracy with a low computational load. For IHNA, the aggregation constant [γ = 0.8), inertia weight [ꞷ = 0.7)] and attraction coefficient [λ = 1.2)] were optimized by means of grid search method in order to stabilize the convergence and reduce oscillations. These parameters were chosen after several simulation trials to maximize the performance parameters including energy efficiency, accuracy of trust, and the reduction of latency to ensure that the model is reproducible and adaptable to future implementations.

Simulation setup

Table 3 presents the simulation structures used in this study, which will give a full idea of the network setup and settings. In this simulation, 50 nodes were deployed in a network area of 1000 m * 1000 m. Data packets between these nodes were fixed in length, 512 bytes in size, used a data rate of 250 kbps and the message payload was 64 bytes. To introduce mobility into the network, the simulation included a changing amount of mobile nodes, from 10 to 50, which exhibited speeds of 5–25 m/s. These mobile nodes moved according to a random mobility model that brought an element of unpredictability to the network dynamics. The simulation took into account different modes of operation of in terms of energy consumption. While in sleep mode, the nodes consumed energy at a 1.4 mW rate, which is significantly lower than that of 62 mW consumed for both the transmit and receive modes. The transmission power output was determined to be 55.18 mW, which determined the power of the signals transmitted in the network. The communication between nodes was controlled by the MAC protocol (IEEE 802.11) which ensured standardization and compatibility. The transmission range, which is the maximum distance on which the nodes could communicate effectively, was fixed at 250 m. This parameter has a significant impact on the network coverage and connectivity. The simulation execution time was 1000 s, which allowed the observation of the network behavior and performance over an extended period.

Table 3 Simulation parameters.

In this study, the IEEE 802.11 MAC was chosen because it is available and is actively implemented in the NS-3 simulator, enabling consistent assessment and replication. Although IEEE 802.15.6 was specifically designed for WBANs, it lacks comprehensive support in simulations and is more complex to implement. In addition, the similarities between the two MAC standards are that they both have an identical contention-based access (CSMA/CA), and hence similar tendencies in their performance with respect to energy efficiency and delivery reliability. Therefore, IEEE 802.11 offers a fair estimate of the evaluation of the proposed routing scheme under general WBAN communication conditions.

Comparative analysis with impact of mobile nodes

Table 4 describes the results comparison of the proposed and existing routing schemes with respect to varying quantities of mobile nodes from 10 to 50 with a mobile node speed of 5 m/s. Figure 5 depicts the energy consumption results for the proposed and existing routing schemes with different numbers of mobile nodes. The QEEC-Routing outperforms existing schemes in terms of energy efficiency. Specifically, compared with T-MAC, MS-MAC, MT-MAC, and DT-MAC, the proposed QEEC-Routing scheme demonstrated significant improvements. The QEEC-Routing scheme reduces energy consumption by 55.715%, 48.549%, 38.615%, and 23.927%, respectively, compared to these existing schemes. The energy efficiency of the QEEC-Routing scheme is of paramount importance for the prolonged operation of IoT-WBANs. It plays an important role in the conservation of limited energy resources of the wearable devices. Furthermore, this improved energy efficiency guarantees that the transmission of medical data is reliable and continuous, and contributes to the high quality of service demanded for healthcare applications.

Table 4 Results comparison of proposed and existing routing schemes with varying number of mobile nodes.

Fig. 5

Energy consumption with number of mobile nodes.

Figure 6 shows the results for the packet delivery ratio with different numbers of mobile nodes for various routing schemes. The results show clearly improved presentation regarding packet delivery ratio of the QEEC-Routing arrangement in comparison with existing schemes. Specifically, when compared with T-MAC, MS-MAC, MT-MAC and DT-MAC the proposed QEEC-Routing scheme showed significant improvements. It is 10.365%, 7.774%, 5.183% and 2.591% better than these existing schemes respectively. This improvement in the efficiency of packet delivery is of tremendous importance for IoT-WBANs, since it directly affects successful transmission of important medical data. By improving the delivery ratio, the QEEC-Routing scheme helps in guaranteeing a high QoS in healthcare applications. Reliable and timely delivery of medical data is important for patient monitoring and ensuring that healthcare providers are getting the material they need in a timely manner for decision-making and interventions. Figure 7 shows the outcomes of the network lifetime for various routing schemes with varying number of mobile nodes. The QEEC-Routing approach showed better network longevity compared to other methods being used. Specifically, when compared to T-MAC, MS-MAC, MT-MAC and DT-MAC, the proposed QEEC-Routing scheme shows considerable improvements. The QEEC-Routing scheme increases the network lifetime by 17.641%, 13.231%, 8.82% and 4.41% with respect to these existing schemes respectively. Extending the network lifetime is important for IoT-WBANs because it will contribute to seamless monitoring of patients and efficient operation of healthcare application.

Fig. 6

Delivery ratio with number of mobile nodes.

Fig. 7

Network lifetime with number of mobile nodes.

Throughput performances of the multiple routing schemes with different mobile nodes are shown in Fig. 8. The QEEC-Routing approach has a higher throughput compared to other existing approaches. Compared with T-MAC, MS-MAC, MT-MAC and DT-MAC our QEEC-Routing scheme shows large improvements in throughput. Compared to these existing schemes the QEEC-Routing scheme achieves 22.664%, 16.689%, 11.404% and 4.622% better throughput, respectively. Improved throughput is critical for IoT-WBANs because it allows efficient data transmission and is a part of the QoS demanded in healthcare applications. Figure 9 shows the analysis of the latency of the various routing techniques, while the QEEC-Routing technique shows a much lower delay and hence the efficiency of the data transmission. Compared with T-MAC, MS-MAC, MT-MAC, and DT-MAC, the proposed QEEC-Routing scheme shows great improvements in terms of latency. The QEEC-Routing scheme decreases the latency by 58.703%, 51.599%, 41.545% and 26.219% respectively as compared to these existing schemes. Lower latency is critical to IoT-WBANs, since it guarantees quicker data transmission and contributes to the QoS needed in healthcare applications. Figure 10 describes the routing overhead performance with different routing schemes with different node densities, which reveal the effectiveness of the proposed model to minimize the communication overhead. The proposed QEEC-Routing scheme shows better routing overhead in comparison to the existing schemes. Compared to T-MAC, MS-MAC, MT-MAC and DT-MAC, our proposed QEEC-Routing scheme showed significant reductions in routing overhead. The routing overhead is reduced by 74.26%, 68.392%, 59.059% and 41.903% respectively with respect to these existing schemes by the QEEC-Routing scheme. A decreased routing overhead is desirable in IoT-WBANs, which optimizes network resources and helps to achieve efficient, data transmission, hence improving the QoS for healthcare applications.

Fig. 8

Throughput with number of mobile nodes.

Fig. 9

Latency with number of mobile nodes.

Fig. 10

Routing overhead with number of mobile nodes.

Comparative analysis with impact of node speed

Table 5 shows a comparative study of the proposed routing scheme with the existing routing schemes with different mobility speed varying from 5 to 25 m/s and with total number of mobile nodes fixed at 50 nodes. In Fig. 11, the energy consumption results are shown for various routing schemes for different mobile node speeds. The energy efficiency of the QEEC-Routing scheme is remarkable compared to schemes that exist. Compared to T-MAC, MS-MAC, MT-MAC and DT-MAC, the proposed QEEC-Routing scheme had lower energy consumption. Our QEEC-Routing scheme beats these current schemes with energy consumption improvements of 74.667%, 68.852%, 59.574% and 42.424%. The increased energy efficiency realized by the QEEC-Routing scheme is an important factor in extending the operational life of IoT-WBANs, since it helps to ensure the energy resources of wearable devices are preserved.

Table 5 Results comparison of proposed and existing routing schemes with varying mobile node speed (m/s).

Fig. 11

Energy consumption with node speed.

Figure 12 shows the results of the packet delivery ratio for different speeds of the mobile node when routing protocols are used. The proposed routing algorithm showed better performance than the existing methods in terms of improved packet delivery ratios. Specifically, compared to T-MAC, MS-MAC, MT-MAC, and DT-MAC, the QEEC-Routing scheme has remarkable improvements in packet delivery ratio of 10.556%, 7.917%, 5.278% and 2.639%, respectively. This improvement in the QEEC-Routing scheme is a result of its QoS-aware design, optimal clustering and pathfinding strategies. By properly utilizing Trust-degree computation algorithm and path optimization algorithm, QEEC-Routing scheme ensures reliable and low-interrupted transmission of data packets.

Fig. 12

Delivery ratio with node speed.

In Fig. 13, the results of the network lifetime are presented for various different routing schemes at different mobile node speeds. The proposed approach was observed to be consistently better than the existing algorithms with significantly improved network lifetimes. Specifically, compared with T-MAC, MS-MAC, MT-MAC and DT-MAC, QEEC-Routing scheme has improvement in network lifetime, with efficiency improvements of 27.202%, 21.575%, 15.948%, and 10.321%, respectively. The outstanding growth in network lifetime provided by the QEEC-Routing scheme can be explained by the energy-efficient design and the optimized routing techniques. The QEEC-Routing scheme has a strong performance in terms of resource utilization and energy conserving, which is essential for extending the operational life of IoT-WBANs. This efficiency is mainly accomplished through its QoS-aware approach which ensures judicious resource usage. The application of the MRO algorithm for cluster formation is an energy efficient data transfer and network longevity.

Fig. 13

Network lifetime with node speed.

Figure 14 shows the through put performance comparison of routing schemes at various node speeds, from which QEEC-Routing has a better data transmission efficiency. Specifically, as compared with T-MAC, MS-MAC, MT-MAC and DT-MAC, the proposed QEEC-Routing scheme shows considerable improvements in throughput mainly by improving efficiency by 24.262%, 17.854%, 11.161%, and 6.603%, respectively. The great boost in throughput provided by the QEEC-Routing scheme can be explained by the efficient routing schemes and optimized cluster-based approach. The proposed scheme ensures that the data can be transmitted in a timely and reliable manner, hence contributing to the requirement of the QoS for the healthcare applications. This reliability in data transmission causes a higher throughput.

Fig. 14

Throughput with node speed.

Figure 15 shows that QEEC-Routing has much lower latency than other schemes in different mobile node speeds. Compared with T-MAC, MS-MAC, MT-MAC, and DT-MAC, QEEC-Routing scheme has large improvements with efficiency enhancements of 65.919%, 59.195%, 49.164% and 32.594% respectively. TQV-RNN technique finds the best suited paths for data transmission which prevents congested or inefficient paths. Streamlined data paths lead to less data transfer latency. Figure 16 gives the routing overhead of different schemes at different mobile node speeds. The QEEC-Routing scheme shows an amazing efficiency of reducing routing overhead compared to the existing T-MAC, MS-MAC, MT-MAC, and DT-MAC schemes. The improvements in the QEEC-Routing scheme are shown with efficiencies of 49.365%, 42.237%, 32.772% and 19.597%, respectively.

Fig. 15

Latency with node speed.

Fig. 16

Routing overhead with node speed.

Results comparison of high-density nodes

The performance comparison of the proposed QEEC-Routing scheme with the existing routing schemes T-MAC, MS-MAC, MT-MAC, and DT-MAC shows the significant improvements in both the packet delivery ratio and network lifetime, which can be summarized in Table 6. In terms of the Packet Delivery ratio, QEEC-Routing was always better than all other schemes. At a node density of 1000, QEEC-Routing achieves a packet delivery ratio of 93.235%, which is notably higher than DT-MAC (75.525%), MT-MAC (83.152%), MS-MAC (81.058%), and T-MAC (75.856%). The percentage improvements were 22.47% over DT-MAC, 12.74% over MT-MAC, 12.87% over MS-MAC, and 17.37% over T-MAC. As the node density is increased, the benefit of QEEC-Routing is still a significant one. At a density of 5000 nodes, the scheme achieve a delivery ratio of 88.512% which is 16.20% higher than DT-MAC, 18.50% higher than MT-MAC, 19.96% higher than MS-MAC and 23.17% higher than T-MAC, respectively. Similarly, QEEC-Routing has the best network lifetime. At 1000 nodes, it extends the network lifetime to 90.125%, which is higher than DT-MAC (85.625%), MT-MAC (83.124%), MS-MAC (79.290%), and T-MAC (75.790%). The increases were 4.54% for DT-MAC, 7.00% for MT-MAC, 10.67% for MS-MAC, and 14.34% for T-M. At 5000 nodes, QEEC-Routing achieves a network lifetime of 84.325%, outperforming DT-MAC by 9.08%, MT-MAC by 12.04%, MS-MAC by 19.08%, and T-MAC by 25.10%. These results, which are presented in more detail in Table 6, illustrate the success of the QEEC-Routing scheme in improving the packet delivery and network lifetime. The improvements were observed to be between 16.20 and 23.17% in packet delivery ratio and between 4.54 and 25.10% in the network lifetime depending upon the node density. This performance highlights the capability of the proposed scheme to maintain efficiency and extend network lifetime in high-density scenarios.

Table 6 Performance comparison of proposed and existing routing scheme for high density nodes.

Statistical analysis

The statistical validation, cross-validation, and descriptive analysis of the proposed method using the extant routing protocols T-MAC, MS-MAC, MT-MAC, and DT-MAC are presented in Tables 7, 8 and 9, respectively. Based on Table 7, the statistical validation of the Wilcoxon signed-rank test shows that the proposed scheme is significantly better in all benchmark protocols, with p-values smaller than 0.05 for all performance metrics. The fivefold cross-validation outcomes, as illustrated in Table 8, also confirm the consistency of the models, with QEEC-Routing having the lowest cross-validation mean of 1.196 and standard deviation of 0.1673, which indicates its ability to be generalized, as well as a smaller variance as opposed to all currently existing protocols. Table 9 presents the descriptive statistics of energy consumption, and the 25th, 50th (median), and 75th percentile of the values of QEEC-Routing are 1.235, 1.563, and 1.987 mW, respectively, representing a narrower and more compact range of energy distribution than competing schemes. These results confirm the hypothesis that QEEC-Routing can provide statistically significant energy efficiency, reliability of delivery, and overall performance of the network compared with state-of-the-art techniques.

Table 7 Statistical validation of QEEC-Routing compared with existing routing schemes.

Table 8 Cross-validation analysis of proposed and existing routing schemes.

Table 9 Descriptive statistics of energy consumption (mW) for all routing schemes.

State-of-the-art comparison

Table 10 shows the suggested QEEC-Routing scheme with the current state-of-the-art WBAN routing schemes in 2022–2025. The data indicate that QEEC-Routing consumes the least amount of energy (1.235 mW) and delivers the highest number of packets (96.356), which is considered efficient and reliable in data transmission. It also achieves the highest throughput (5426 Mbps), which is better than those of EEART and M-EEMH. These enhancements underscore the performance of the proposed fuzzy-based clustering and energy-conscious routing scheme. Overall, it can be concluded that QEEC-Routing is a better performance router that proves to be robust and effective in comparison with modern WBAN routing methods.

Table 10 state of art comparison.

Computational complexity analysis

The time complexity of the modified Raccoon optimization (MRO) algorithm is O(N.C), where N is the total number of sensor nodes and C is the number of candidate cluster nodes. Each iteration entails the calculation of the fitness functions of all the raccoon agents and the renewal of their locations according to the visible and potential areas. The complexity of space is O (N + C), because every agent only maintains the current and best fitness values. The adaptive search and limited number of iterations guarantee rapid convergence, which is appropriate for low-power WBAN. The TQV-RNN model has a time complexity of O(N·L·H²), where N is the number of nodes, L is the number of RNN layers, and H is the number of hidden neurons. Matrix multiplications and quaternion updates are performed by each neuron and are quadratically scaled by the dimension of the hidden representation. The storage of weights, activations, and recurrent states has a space complexity of O(L·(H + H²)). Although the algorithm is computationally intensive, the quaternion representation is found to converge faster, and fewer redundant operations are performed, making it efficient in constrained WBANs. The IHNA algorithm has a time complexity of O(N·C·R), where N is the number of nodes, C the number of possible paths, and R the number of iterations required for convergence. The stages in each iteration include distance normalization, coefficient updating, and path fitness evaluation. The space complexity is O (N + C + R), which considers the temporary storage of path vectors and aggregation coefficients. It has high routing precision with its multi-parameter adaptive updates and a computational load that is acceptable in resource-limited WBAN devices.

Table 11 depicts the computational efficiency analysis of the proposed and existing routing scheme. The proposed scheme of QEEC-Routing has a computational complexity of O (N.C.R), with each node being involved in the formation of clusters, selection of CH and finding the optimal route by performing an iterative evaluation of energy levels and optimizing the route. Although this creates a little more overhead in terms of computation than the more traditional MAC-based techniques, the adaptive design provides a balance in load distribution and quicker convergence in the optimization of paths. The storage required to maintain neighbor lists, cluster parameters, and routing tables is captured by the space complexity O (N + C + R). In general, QEEC-Routing demonstrates efficiency in computing and memory consumption, besides providing substantial performance increment in both energy efficiency and network reliability.

Table 11 Computational efficiency analysis of the proposed and existing routing schemes.

Conclusion

The proposed QEEC-Routing is designed for IoT-WBANs to achieve efficient data transmission and reliable communications. This approach combines three high-tech computational approaches to improve the routing performance. MRO algorithm has been used to get optimal cluster formation, with an equitable use of energy among the sensor nodes. The TQV-RNN is used for trust evaluation of nodes accurately, leading to the enhancement of reliability of communication links. For path selection, the IHNA decides the most efficient route (and stable) within the network. Simulation experiments performed with the NS-3 simulator proved that QEEC-Routing was significantly better than conventional routing techniques. The proposed system has a reduction in energy consumption of 41.71% for different node densities and 61.37% for different node speeds. Under similar conditions, it also improves the packet delivery ratio by 6.47% and 6.56%, respectively. Furthermore, QEEC-Routing increases the network lifetime by 11.03% for different number of nodes and 18.76% for different mobility speeds. The throughput is improved by 13.84% and 14.94% while the latency is reduced by 44.52% and 51.72% compared with the existing methods. Additionally, the routing overhead reduced by 60.09% and 35.99%, which shows the significance of QEEC-Routing in solving the major challenges, i.e., energy efficiency, reliability, QoS assurance, and network sustainability in IoT-WBAN environment, especially in healthcare monitoring applications.

In future work, the proposed QEEC-Routing scheme will be validated via real-world experiments based on an IoT-enabled WBAN testbed. The setup will comprise of wearable sensor nodes integrated with medical monitoring devices and mobile sink nodes for data aggregation and an IoT gateway connected to a cloud-based healthcare platform. Performance metrics such as energy utilization, packet delivery ratio, delay and network lifetime will be observed in a realistic context including node mobility, interference and body shadowing effects. This physical deployment will help to check the practicality, reliability and adaptability of the proposed routing scheme in addition to simulation, for its effectiveness with real-time health monitoring applications.