Efficient hybrid enhanced genetic algorithm and ant colony system model for rerouting multimedia message in multiple node–link failures within wireless network

Failure‐resilient wireless networks have attracted the interest of the research community and have been an open area of concern in the studies of wireless network in recent years. Accordingly, different research on restoration techniques have been carried out to proffer solution to the network component failures. In the same vein, this article proposes a hybridized enhanced genetic algorithm and ant colony system (EGAACS) survivability model, which can instantly resolve node–link failure problems, thus improving the quality of service of the wireless network. The EGAACS is a hybrid model that combines the principles of the enhanced genetic algorithm (EGA) and ant colony system (ACS) models to form a capacity efficiency solution that outperforms the ACS and EGA models in terms of path cost and transmission delay. The resilience of this proposed EGAACS model is verified for different wireless network sizes (20, 30, 40, and 50 node networks). Simulation results show that the proposed EGAACS model generates the best close to optimal paths in terms of the path cost and transmission delay in comparison to the EGA and ACS models. In fact, the performance of the proposed EGAACS model is more conspicuous as the size of the network increases. More importantly, the proposed EGAACS model is suitable for real‐time wireless network as it exhibits moderate computational time complexity.


| INTRODUCTION
The ability of a wireless network to continue providing services in line with requirements, despite node-link failures and other undesirable events, is referred to as network survivability. 1 Wireless networks are expected to offer high dependability by achieving quick restoration in the event of node-link failures. This requires putting share capacity in 2 | RELATED WORK This section briefly discusses some relevant work used to solve node-link failure problems and the background of the EGA and ACS models. Two approaches to path planning were proposed in Tewolde and Sheng, 11 one based on GA and the other on ACO. On a real-world deployment of numerous robotic manipulators equipped with certain spraying instruments in an industrial setting, a comparison of the two procedures was done. The study asserted that while both systems yield results that are nearly identical for small problem sizes, as problem size and complexity increase, the ACO-based algorithm outperforms the GA-based algorithm despite requiring a longer implementation time.
Sariff and Buniyamin 12 carried out a comparative performance study on the GA and ACO algorithm. The performance of algorithms were verified in three different workspaces with varying complexities, and the ACO approach was found to be more reliable, whereby its performance surpasses the GA approach in terms of the number of iterations and computational time complexity. Despite that the ACO and GA methods have shown to be effective in resolving path planning problems, the authors conclude that these two techniques have some limitations. They showed that ACO has a quicker convergence speed and a powerful local search capability, but if the problem is big enough, the algorithm can quickly enter a local optimum. However, because GA is a random optimized process, it does not have a problem with local convergence; however, the rate of its convergence slows down. Therefore, it is thought that a combination of the ACO and GA approach, which is the subject of this article, can be a promising substitute. Buniyamin et al 13 proposed an expanded ACO algorithm to address the global path planning problem and assessed its efficiency in comparison to the GA. They proved that the solutions can look for a path that is close to optimal. In spite of this, it was found that the ACO algorithm was more reliable and efficient.
Furthermore, some works in the literature that provided answers to the path planning issues merged the GA and ACO algorithms. Using a hierarchical hybrid algorithm, Ma and Wen-Jing 14 put forward a path planning solution that aims to find a path in a network of roads. The ACO algorithm is applied to each of the numerous subnetworks that the authors created, and the optimal paths that the ACO algorithm produces are utilized to populate the GA. In their simulation work, the authors showed that their approach can find the optimum path more rapidly and efficiently than the work in Xu et al 15 with less iterations. Additionally, Gao et al 16 integrated the GA and ACO algorithms to proffer solution to the the robot navigation problem. To enhance the quality of service (QoS) of the paths outputted by the ACO at the start of the algorithm, special functions were proposed: Crossover and mutation operations. This operation was used to improve the ACO algorithm solution's QoS and prevent it from reaching a local optimum. The augmented method generates a better solution in terms of direction, implementation time, length, and iterations when compared to the traditional ACO algorithm.
Geetha et al 17 proposed the ACOG, a hybrid ACO algorithm and GA. They improved the efficiency of the ACO algorithm by using three genetic operations (structure-preserving crossover, Darwinian reproduction, and structure-preserving mutation), which prevented it from reaching a local optimum. More so, the authors noted in their paper that their method is a multiobjective algorithm because it takes into account three different criteria: Smoothness, path length, and security. They demonstrated that their approach provides a path that is close to optimal by comparing it to the work in Zhou et al. 18 As mentioned, these techniques discussed used the traditional genetic algorithms and/or ACO algorithms to solve wireless network failures. However, these authors do not incorporate a system that will generate the required bandwidth to reroute the messages in real-time, which automatically increase the transmission delay of their model in finding a new route in event of node-link failures, whereas the proposed model in this article adopted the EGA and ACS, which incorporate heuristics that calculate the bandwidth required in real-time to reroute multimedia messages as soon as (real-time generation of required bandwidth to reroute messages) the rerouting path is generated. In addition, the generation of the alternative paths are not implemented immediately; there is a node-link failure; it is either the alternative paths are generated before there are failures or the network is completely down. This proposed EGAACS model finds the alternative path in real-time; that is, immediately, there is a node-link failure with little or no transmission delay. Besides, the proposed EGAACS model solves the ACS problem of falling into a local minimum and the premature convergence problem associated with the GA model, which is a major setback with these discussed techniques. Accordingly, this proposed EGAACS model will generates a path from the source node to the destination node with reasonable path cost and transmission delay. Table 1 summarizes these related models in the literature with the proposed EGAACS model.

| Swarm intelligence based on ACS
An ACO is a heuristic algorithm on agents that simulates the natural behavior of ants by developing mechanisms of cooperation and learning, allowing the exploration of positive feedback between agents as a search mechanism. 8 Ant colonies' foraging behavior is significant and fascinating, especially how ants can choose the quickest routes between food sources and their nest. 9 Ants leave a chemical substance called pheromone on the ground as they go from food sources to their nest and the other way around, creating a pheromone trail. Ants are able to detect pheromones, and they are more likely to choose a path that has a high pheromone concentration. The pheromone trail enables the nestmates of the ants to direct them toward the food. Figure 1 depicts how real ants determine the shortest path. Ants arrive at a decision at A. Some ants proceed up the top path at B, while others proceed down the lower path. The choice is made at random. Since ants move at a roughly constant speed, those who choose the lower, shorter path at C get at the opposing decision point before those who choose the top, longer way. The shorter path at D allows pheromone to build more quickly. The quantity of pheromone left behind by the ants is indicated by the number of dashed lines. In general, all ACS algorithms adhere to the algorithmic scheme outlined in Algorithm 1. The initial pheromone value is set, along with a number of other parameters, during the initialization procedure. Up until an end condition is satisfied, the main loop is then iterated. At each iteration, the ants are entrusted with coming up with workable solutions to the optimization problem. The use of a local search technique could then potentially improve the solutions that were provided. As ants deposit pheromone on the components used during search, the amount of pheromone is updated; the pheromone value can either rise or decrease.

| EGA
All GA, in general, follow the algorithmic scheme described in Algorithm 2. A set of initial population or paths called chromosomes are generated at the beginning of the algorithm. To ensure that a genetic algorithm converges on the best solution, only the fittest solutions should survive. Each chromosome is assessed to determine its fitness using the fitness function. The fittest solutions are then subjected to genetic transformations such as the crossover and mutation operations, which results in new best solutions. This procedure is carried out until the end condition is satisfied.
The EGA is a computational technique for simulating evolution processes and natural selection in organisms. 10,19 It involves steps such as generation of the initial population, evaluation, selection, crossover, mutation, and regeneration. The initial population is significant because it symbolizes the solution and is typically produced at random. A problemspecific function is then used to assess the population. GA will then choose some of them depending on the likelihood that they will mate in the next process. Subsequently, a new and improved path will be generated using crossover and mutation operations. The concept behind the GA is that each generation of solution should be superior than the one before it. This procedure is iterated until some sort of termination point is met. GA is widely applicable in many optimization problems because it does not require a thorough understanding of the problem at hand. Furthermore, this algorithm can search for the global optimum, and it is well suited to the problem. 20 Yet, the GA is prone to premature F I G U R E 1 How real ants find the shortest path.
convergence, making it difficult to achieve the best solution, which is one of its drawbacks. Also, GA requires longer processing time when dealing with large amounts of data. 21 Generally, all GAs adhere the algorithmic structured described in Algorithm 2. At the beginning of the algorithm, a set of paths is produced. Also, in GA, only the fittest solutions should survive such that the algorithm can converge on the best solution. The fitness function is used to assess each chromosome's fitness. The fittest solutions are then subjected to genetic transformations such as crossover and mutation, which result in new solutions. This procedure is repeated until the stopping point is reached.

| EGAACS MODEL DEVELOPMENT
As discussed earlier, the EGAACS model incorporates the best features of both the EGA and ACS models to provide an effective path to the destination node from the source node in case of node-link failures. As such, the EGA and ACS models' parameters and mathematical analysis are first discussed.

| Mathematical analysis of ACS algorithm
The ACS parameters are used for generating an alternative path in case of node-link failures. Accordingly, the ACS model used to describe the survivability of telecommunication network is explained as follows.
• Initial pheromone concentration: where the number of nodes in the network and the length of the route found by a nearest neighbor heuristic are given as n and L nn , respectively. 22 • Local pheromone update: where the parameter governing pheromone decay is given as ρ, ρ ¼ 0:1 was adopted from literature such that 0 < ρ < 1 and τ 0 , the amount of pheromone on the edge ði,jÞ at time t is given as τ i,j ðtÞ, and the initial value of pheromone on all edges is given as τ 0 . 22 • Edge attractiveness: where the distance between i and j is given d. 22 • Computation of edge probability: where α and β are the two parameters that decide the relative influence of the pheromone trail and τ i,j is the amount of pheromone on the link between nodes i and j. 22 Table 2 summarizes the ACS model parameters.

| Mathematical analysis of EGA algorithm
The step-by-step approach used by the EGA model to generate an alternative path in case of node-link failures is described as follows.
• Chromosome value encoding: The chromosome value encoding a communication path is the interconnection of links, where the distance between two nodes is the path cost. A typical example of communication paths are shown in Table 3. • Fitness evaluation of network paths: The genetic algorithm searches for the optimal path with the highest fitness, where the fitness function is used to assess the quality of paths within the network. The fitness function that involves computational efficiency and accuracy is defined as where f i is the fitness value of the i th chromosome (path), l i is the length of the i th chromosome (path), g i ðjÞ is the gene (node) in the network path, C is the path cost between nodes in the chromosome, and B is the path bandwidth.
T A B L E 2 ACS model parameters.

Chromosomes
Links forming the chromosomes (paths) • Selection from the collection of network paths (ranking): The selection mechanism chosen is rank-based fitness assignment. The fitness assigned to each network path depends only on the position of the path cost, required bandwidth, and number of nodes in the network path. Ranking introduces a uniform scaling across the collection of network paths as defined as where p i is the probability that individual path, i will be selected, f i is the fitness of individual path, i, and P f j represents the sum of the fitness of all individual paths from the source node to the destination node. Table 4 summarizes the EGA model parameters.

| Bandwidth needs for optimal paths
The amount of data that can be sent over a network is referred to as the bandwidth. 23 Congestion must be controlled in order to maintain good QoS for the flow in the produced path. The congestion can be as a result of the following: • The network path has been allocated a restricted amount of bandwidth.
• There is a strong demand for the transmission path and a high volume of work at the nodes, which lowers QoS.
The throughput can be used to estimate the minimum bandwidth needed to convey an interactive media message as defined by Equation (7) 24 : where W denotes the size of a multimedia packet to be transmitted and D denotes the packet delay. Network parameters including network latency, transmission delay, and per-packet router processing time determine the delay value. Therefore, the needed bandwidth whose path includes two or more nodes is defined by Equations (8) and (9).
In Equations (8) and (9), λ and ω are both constant values, a is an unknown function that depends on n and I, while n is the number of nodes in the routing path, and I ¼ P n I n is the total amount of a single route discovery path length known as the path cost.

| Proposed EGAACS model
The proposed EGAACS model hybridized the close to optimal paths obtained from the ACS and EGA models. This means the proposed EGAACS model accepts two inputs at the start of its path finding process. The outputted path of T A B L E 4 EGA model parameters. The link bandwidth the ACS algorithm serves as one of its input while the outputted path from the EGA algorithm is taken as its other input. Afterwards, the EGAACS model performs two genetic operations sequentially: Point crossover and omitting mutation, to find a more reliable path in comparison to the paths generated by the ACS and EGA model alone. The crossover operator combines subsets of both parent chromosomes to create kids with some genetic information from both parents. Crossover can be classified into two types: One (single) point and multipoint crossover. There is just one crossover site in a single point crossover, and there are multiple crossover sites in multipoint crossover. 25 Crossover looks at present solutions in order to come up with better ones. In the case of routing issues, crossover involves physically exchanging each route of the two chosen chromosomes in such a way that the child produced by the crossover will only be one route. This article adopts the single crossover operation. Furthermore, the mutation operator modifies genes at random to partially shift the solution pace to new regions. If the values of successive iterations are the same, mutation is performed. 25 It is possible that the crossover operation create populations that are degenerate. A mutation operation is used to undo this. There are different types of mutation operation; examples include inversion, insertion, reciprocal, omitting, and exchange. Inversion involves picking two random places and reversing the string between them. A node is placed at a random point in the string in the event of insertion. Nodes at two random places are exchanged in reciprocal exchange. The omitting mutation operator is used in this article, which is accomplished through mutating one chromosome/node at random within future group. The chosen gene is then removed.

EGA parameters Definition
Having fed the paths from the ACS and EGA models, the proposed EGAACS model first evaluates the fitness of these paths using Equation (5). Subsequently, the model performs the first genetic operation (point crossover) and checks for duplication of chromosomes (D C ). The model then outputs the path with the highest fitness if there exist any duplication. Otherwise, it formulates Child 1 (C1) and Child 2 (C2) from the Parent paths. Thereafter, a fitness comparison is performed between the Parent and Child. If the Child fitness is greater than that of the Parent, the Child serves as the new path. On the other hand, the second genetic operation (omitting mutation) is performed. Therefore, new sets of paths are generated. At this point, the algorithm checks again if D C exist and the process continues until its output is close to the optimal path. The process is hereby summarized in Algorithm 3  Figure 2 represents a mobile ad hoc network (MANET) for processing multimedia packets in media houses with virtual links connecting the workstations (nodes) and the processing offices where MANETs are installed, while the connections are linked to the Internet through a gateway (access points). The transmission/rerouting are serial in such a way that packet processing must be complete in a MANET before moving to the next MANET to produce good quality multimedia product before marketing it to the public. The main objective of this arrangement is that each MANET is a dedicated network to handle a kind of processing that is different from the processing in other MANETs. For details of such wireless network packet streaming, refer to Taha and Ali. 26 During the processing activities, a workstation was discovered to be down and at the same time causing the virtual link that will carry the packet to other workstations for the next processing phase to be down. To resolve the failure, an alternative path must be generated to transmit the multimedia packet for processing to continue. This arrangement depicts a wireless network and can be represented as a complete weighted graph, G ¼ ðN, EÞ, where N is the set of n nodes and E is the set of edges (paths) fully connecting all networks. 8 Each edge ði,jÞ E is assigned a cost d ij , which is the distance between nodes i and j. The cost, d ij can be defined in the Euclidean space as Thus, the optimal solution is to choose the route with the lowest total path cost among all possible combinations of N nodes. For even n nodes, the number of permutations can be very large because there are n! different ways to permute n numbers. Figure 3 depicts the transition diagram of Figure 2. The circles represent different network nodes, while the line connecting them represents the virtual link.

| Alternative paths generation: ACS
The ACS model is used here to analytical provide an alternative path to transmit the multimedia message in the event of node-link failure as depicted in Figure 3. The parameter specifications for this ACS-based capacity efficiency model is shown in Table 5. Note that most of the parameter values in Table 5 have been carefully chosen for best tuning. Otherwise, the parameter values have been adapted from best practices in literature for the purposes of analysis. 8 F I G U R E 2 Node-link failure of multimedia transmission. Accordingly, the analysis in finding the alternate route in the multimedia message network of Figure 3 from the first to the last iteration is summarized in Figure 4.
From Figure 4, the generated path from the source node to the destination node is "A -C -H -J -N" with a path cost of 7; that is, total distance ¼ 1 þ 4 þ 1 þ 1 ¼ 7 mm. Therefore, Equations (8) and (9) can be used to calculate the required bandwidth for routing the multimedia message over this optimal alternative path. For example, if the size of the multimedia message is 600 bytes and the IP header is 20 bytes, the actual message size is ð600 À 20Þ ¼ 580 bytes. The delay is calculated by multiplying the length of the links by the propagation speed. Refer to Kamal and Taha 27 for more details about delay calculation in wireless networks. If the starting time is 5 ms (0.005 s) for example, the time it will take the message to travel from the source to the destination will be ð7 * 5Þ ¼ 35 ms. This implies that the outputted path has a delay of 35 ms. Thus, applying Equations (8) and (9): Delay, D ¼ 25ms ¼ 35 1000 ¼ 0:035s a ¼ 7 þ 5 ¼ 12mm ¼ 0:012m λ and ω are taken to be 1. Bandwidth, B ¼ W DÀa ¼ 5800:035 À 0:012 ¼ 25,217:39bytes=s.

| Alternative paths generation: EGA
The EGA model is employed to analytically provide an alternative route to transmit the multimedia message in the wireless network of Figure 3 from the source node to the destination node. Table 6 showcases the parameters assumed in the EGA model. The EGA model finds an alternate path to the destination in the event of node-link failure using these four steps.
• Step 1-Chromosomes generation (first generation): All available paths to the destination nodes are first generated. The paths represent the chromosomes and the nodes represent the genes. These paths generated are candidate solutions. In this case, there are eight possible paths the packet can be routed through to the destination. These paths are as follows: Step 2-Fitness: The fitness function and rank have to be calculated to determine the fitted chromosomes for the new generation. Thus, fitness function is computed using Equation (5). As indicated, Equation (5) comprises three key parameters: The path cost and the number of nodes in the chromosome and the bandwidth required to transmit the packet. Figures 5 and 6 depict the computation of these parameters.
Having computed the path cost, number of nodes in the chromosome, and required bandwidth, the fitness function for all the paths is compiled using Equation (5) as shown in Figure 7 • Step 3-Ranking: After the fitness function for all the available paths to the destination nodes have been determined, the ranking system for the paths is established. The ranking provides a uniform scaling across the collection of all network paths. Figure 8 depicts the ranking calculation process using Equation (6).  To perform the crossover operation, assume that Parent 1 = Path 3 (fitness=0.0043) and Parent 2 = Path 7 (fitness = 0.0044). If the crossover point is taken to be "J," thus the Child of the Parents are given to be Child 1: Subsequently, the Child's fitness are computed to be 0.0041 (Child 1) and 0.0045 (Child 2) using Equation (5).
F I G U R E 7 Chromosome fitness computation.
F I G U R E 6 Required bandwidth computation.
F I G U R E 8 Chromosomes ranking for optimal path selection.
Therefore, Child 2 moves to the next generation since its fitness value is higher than that of the Parent while Child 1 will be mutated because the Parent's fitness is greater than that of the Child. Nevertheless, observing Child 1 (A -D -H -J -N), there is only one node after the crossover node "J," which is the destination node. This implies that Child 1 cannot be mutated; as such, its Parent (with the higher fitness value) moves to the next generation. So, the second generation paths will be Notice that any further crossover operation with this new generation paths will amount to duplication of chromosomes, D C . Therefore, the chromosome with the highest fitness is selected. That is, the chosen path is "A -B -E -G -K -J -M -N" with a path cost of 11.
As shown, the generated path from the transmitting node to the final receiving node is "A -B -E -G -K -J -N" with a path cost of 11; that is, total distance ¼ 2 þ 2 þ 2 þ 1 þ 2 þ 1 þ 1 ¼ 11 mm. Similarly, Equations (8) and (9) can be used to calculate the needed bandwidth for routing the multimedia packet over this alternative path.

| Alternative paths generation: Proposed EGAACS
The proposed EGAACS model takes as input the outputted paths from the ACS and EGA models as mentioned earlier.
Thus, the parameter assumed in Tables 5 and 6 for the ACS and EGA models are adopted for the EGAACS model for this analytical scenario. The outputted paths from the ACS and EGA model are thus labeled as Path 1 and Path 2, respectively. Path 1: A -C -H -J -N and Path 2: First, the crossover operation is performed, where node "J" is taken as the crossover node. Thus, the Child of the Parent 1 (Path 1) and Parent 2 (Path 2) are given to be Child 1: A -C -H -J -M -N and Child 2: Next, the fitness of the Parent and Child is compared. Using Equation (5), the fitness are computed as Parent 1 = 0.0045, Child 1 = 0.0044 and Parent 2 = 0.0045, Child 2 = 0.0044. As computed, the Child of Parents 1 and 2 have lower fitness values; as such, they will be mutated. Nonetheless, observing Child 2 (A -B -E -G -K -J -N), there is only one node after the crossover node "J," which is the destination node. This implies that Child 2 cannot be mutated. Hence, its Parent with the higher fitness value moves to the next generation. Also, mutating Child 1 by removing the only node ("M") between the crossover node "J" and the destination node "N" will make Child 1 have the same path as its Parent. Therefore, the Parent moves to the next generation. This means they can be any more crossover operations to avoid duplication of chromosomes, D C . Accordingly, the chromosome with the highest fitness is selected. That is, the chosen path is "A -C -H -J -N" with a path cost of 7.
Comparing the path cost of the path generated by the ACS model and that of the EGA model, the path generated by the ACS model achieves a low path in sending the packet from node "A" to node "D." On the other hand, the ACS model and EGAACS model selected the same path in this analytical scenario. Yet, it is emphasized that in more detailed wireless network, the proposed EGAACS model selects a path with low path cost in comparison to the ACS and EGA models as documented in Section 5.

| EXPERIMENTAL SET-UP, EVALUATION, AND DISCUSSION
The performance of the proposed EGAACS model is verified in comparison with the ACS and EGA model in this section. Their performances are documented for varying parameters such as the failed link, nodes, and multimedia packet size. More so, their performances was verified for different sizes (20,30,40, and 50 nodes) of the multimedia message wireless network. A fixed value of 20 bytes and 5 ms is assumed for the packet header and transmission time between nodes, respectively, in all experimental set-up. Hence, subsequent sections present the results of simulation runs using MATLAB. 28 Tables 5 and 6, The link, failed nodes, and generated alternative paths for the EGA model, ACS model, and the EGAACS model are shown in Tables 7-9.

| Routing in 20 nodes network
As shown from the tables, an increase in the number of node-link failures results in an increase in the path cost, whereas an increase in the path cost slows down (delay) the transmission speed or the time at which the sent message will get to the destination node. Thus, observe closely from the tables that the delay increases as the path cost increases. This implies that a reasonable value of the path cost is desirable at all time for quick message transmission. Therefore, the path cost plays an important role in the design of any multimedia message transmission model. Having explained this, observe from the tables that the EGA model offers the worst path cost in comparison to the ACS model and the pEGAAC model. The ACS model obtained the second best path cost from the source node to the destination node. The proposed EGAAC model offers the best path cost from the transmitting node to the destination node, even as the number of node-link failures increase. Although the computational complexity of the EGAACS model surpasses the ACS and EGA models, the trade-off obtained from path cost and delay is key; hence, the proposed EGAACS model can be encouraged for real-time message transmission. To buttress the results presented in the tables, Figures 10-12 depict the generated paths using the EGA model, the ACS model, and the proposed EGAACS model, respectively, for experiment 5.   wireless networks. Notice, as the network nodes increase from 20 to 30 in Figure 13, the proposed EGAACS model offers the best performance both in terms of the path cost and delay in comparison to the ACS and EGA models. For example, at experiment 10 in Tables 10-12, the proposed EGAACS model offer a path gain of 2.2608 and 1.4281 over the EGA and ACS models, respectively. This in return reduces the delay of the EGAACS model in forwarding the packet to the destination node. For the EGAACS model, the message gets to the destination node in 0.05 s, whereas the EGA and ACS models conveyed the message to the destination node in 0.06 and 0.055 s, respectively. Furthermore, as the nodes in the network increase to 40, the proposed EGAACS model offers a reasonable delay and path gain over the ACS and EGA models. For instance, observe experiment 15 in Tables 13-15, the EGAACS model achieves a path gain of 3.6359 and 1.6911 over the EGA and ACS models, respectively. In terms of the delay, the EGAACS model transmitted the message to the end-user node in 0.07 s, whereas the EGA and ACS models transmitted the message to the end-user node in 0.09 and 0.075 s, respectively. This indicates a time gain of 0.02 and 0.005 s over the EGA and ACS model, respectively. Similarly, as the nodes in the network increase to 50, observe from experiment 20 in Tables 16-18  in Figure 17. This implies that for large wireless networks, the proposed EGAACS model will provide higher path cost gain in comparison to the ACS and EGA models in conveying the message from the source node to the destination node. Thus, the EGAACS model can be deployed for real-time wireless network as it serves as a performance capacity efficient alternative to the ACS and EGA models.

| CONCLUSION
This article proposed an EGAACS-based capacity efficiency framework for wireless network failure survivability. This proposed model generates a close to optimal alternate route to forward the multimedia packet at every level of failure. The performance of this EGAACS model is compared with the ACS and EGA models for different network scenarios. First, the article demonstrated that the proposed EGAACS model achieves a reasonable path cost in conveying the message from the transmitting node to the end-user node in comparison to the ACS and EGA models. Second, it was shown in the article that the EGAACS is suitable as the nodes in the network increase in comparison to the ACS and EGA models. In addition, the EGAACS achieves this low path cost in sending the message to the destination node at a time delay that is better than that of the ACS and EGA models. Although the proposed EGAACS model exhibits high computational time complexity in comparison to the ACS and EGA model, it can still be deployed for real-time system as its offer suitable trade-off in terms of its delay and path cost gain.

DATA AVAILABILITY STATEMENT
Data sharing not applicable to this article as no datasets were generated or analysed during the current study.