Traffic related accidents and route congestions remain to dwell significant issues in the globe. To overcome this, VANET was proposed to enhance the traffic management. However, there are several drawbacks in VANET such as collision of vehicles, data transmission in high probability of network fragmentation and data congestion. To overcome these issues, the Enhanced Pigeon Inspired Optimization (EPIO) and the Adaptive Neuro Fuzzy Inference System (ANFIS) based methods have been proposed. The Cluster Head (CH) has been selected optimally using the EPIO approach, and then the ANFIS has been used for updating and validating the CH and also for enhancing the data transmission procedures. The dijkstra’s algorithm has been used for identifying the shortest path for data transmission. The results showcases that the proposed technique has attained the maximum Packet Delivery Ratios (PDRs) as 73.23% at a sensor radius of 130 m and 70.42% at a velocity of 10 km/h. Moreover, the proposed method has outperformed the existing technique in terms of the CH formation delay, the end to end delay and the PDR.
According to the 2016 United Nations Census Report, the population of the cities is greater than that of the pastoral areas for the first time in human history. Currently 54.5% of the world’s population is living the in urban regions and by 2030 it is expected that 60% of the world’s population will subsist in the metropolitan regions [
Moreover, by means of influencing the 5G enabled Vehicular
Initially, the concerned vehicles would be clustered in the network. Therefore, the cluster head would be selected in the beginning using the Enhanced Pigeon Inspired Optimization (EPIO) method. In the proposed EPIO, for enhancing the performance of the PIO, the Opposition based PIO with Cauchy distribution has been adopted and used here. In order to avoid vehicle collisions, the probability of an accident has been presented based on the expected location of a node and thus offers the required premature caveat and the follow-up measures if the probability exceeds a predefined limit value. Subsequent to the structure of the CH, for updating and validating the CH and for improving the data transmission procedures, the ANFIS based prediction model has been proposed here. Also, for improving the data transmission process, the dijkstra’s algorithm has been presented for finding the shortest path. The performance of the proposed approach has been appraised and contrasted with the existing techniques based on the hop by hop aspect in terms of the CH formation delay, the end to end delay and the packet delivery ratio.
The rest of the article has been organized as follows. Section 2 discusses the various related works. The proposed method has been discussed in Section 3. The experimental results and discussions have been described in Section 4. Finally, the paper has been concluded in Section 5.
This section discusses several new traffic management researches on congestion control on VANET. In [
In order to improve the competence of road traffic on VANET the Density-Based Dynamic Cluster (DBDC) was proposed [
As the road congestion increased, in [
In our proposal, an Enhanced Pigeon Inspired Optimization (EPIO) approach and an Adaptive Neuro-Fuzzy Inference System (ANFIS) based methodologies have been presented for enhancing the traffic management and the data transmission procedures in VANET. The architecture of the proposed scheme for traffic management using the EPIO and the ANFIS techniques through VANET has been illustrated in
In our proposal, a dynamic zone clustering method has been employed in VANET. For enhancing the traffic management in VANET, vehicles in a particular area have been divided into eight zones. Each zone has been considered as a cluster here. Clusters are virtual groups in VANET that can be organized by the CH and the clustering algorithm. Every cluster in the VANET zone would possess a CH and a list of Cluster Members (CMs). Here, CMs refer to the vehicles of each cluster.
The cluster head has been selected in each zone based on the dynamic zone based clustering methodology. The PCH (Percentage of Cluster Head) estimation process has been used here for avoiding data congestion. In each zone the vehicles have been divided initially for estimating the maximum buffer size. This is then followed by the computation of the distance between the source and the sink nodes in each of the individual zones. For example: the node nearest to the sink node possessing a higher percentage would be opted as a cluster head for avoiding data congestion. Therefore, the CH has been selected based on the various parameters like: node location (vehicle location), velocity and buffer size based on the PCH as the CH should possess a higher stability among its neighboring vehicles. To choose the CH optimally, the EPIO technique has been deployed.
The fitness function
where, v represents the velocity of the vehicles due to the increasing number of vehicles, it is inversely proportional, p indicates the location distance, w1, w2 and w3 represent the co-efficient, b represents the buffer size of each of the vehicles and w represents the weight of the final value based on the clustering process.
In our proposed EPIO algorithm, the opposition based PIO technique with Cauchy distribution has been implemented. Here, the Opposition Based Learning (OBL) has been used for enhancing the performance of the conventional PIOA and the dynamic Cauchy probability distribution has been used as a mutation operator. The process of selecting the optimal cluster head has been discussed as follows:
The PIO algorithm has been offered with reference to the magnetic field and sun, map and compass based operator models. The Landmark operator model has been offered based on its identities. In order to idealize the various incoming characteristics of the pigeons, two operators have been constructed based on a certain defined set of procedures:
The Map and compass operator: Pigeons could perceive the earth’s field by utilizing the magnets for creating a map in its brain. They tend to assume the height of the sun for adjusting the direction of the compass. As they fly towards their target, they would become less dependent on the sun and the magnetic particles. In the PIO model, virtual pigeons have been employed naturally. In this map and compass operator, rules have been described with reference to its location Xi and velocity Vi of a pigeon i. Locations and velocities in the D-dimension search space have been modernized in every iteration. Based on the
where, Xg is the current global best location, rand is the random number and R is the map and compass factor gained by contrasting the individual locations between the pigeons.
where
Opposition Based Learning (OBL): In OBL, the initial population of the individual methodologies would be approximate and gradual in its subsequent iterations until an optimal solution is reached. The accumulation time of this technique is related to the distance between the initial assumption and the optimal solution. If the choice of the original solution appears to be optimal, then it can be integrated quickly, or else, it would consume longer time durations for accomplishing the integration procedures. One of the best ideas for improving the initial solution by evaluating the existing candidate solution and its counter solution at the same time is to learn from the OBL and select is the one that appears to be more suitable for the initial solution. This is because, according to the probability theory, any predicted solution is 0.5 times larger than its actual solution. This technique would be constructive not only to start the population, but also to develop the ultimate key for the individual iterations. OBL’s is an optimization problem, also at the same time it is capable of estimating the present aspirant solution together with its counter solution.
where Pi,j is the jth location vector of the ith pigeon in the population, OPi,j is the opposite position of Pi,j, apj and bpj are the least and the greatest values of the jth dimension in the current population respectively.
Different mutation operators have been proposed in the evolutionary optimization literature for enhancing the performance levels by avoiding the pre-integration procedures. Among them, the spread of the Gauss and the Kuchi has become popular. Compared to the Gaussian probability distribution, the Cauchy probability distribution tends to escape the local optimum due to its long-tail probability distribution function. This prompts us to utilize the Cauchy probability distribution as a mutation operator for improving the execution of a regular PIO. In this algorithm, the dynamic Cauchy mutation has been applied on the pigeons for enhancing the performance of the PIO. The one-dimensional Cauchy density function has been denoted by
where t > 0 is a scale parameter. The Cauchy distributed function can be computed using
where Ft (x) represents the Cauchy distributed function and x represents the solution of the fitness function. The cause for exhausting such a mutation operator is to raise the probability of evading from a local optimum [
where
After selecting the optimal cluster head using the EPIOA, the vehicle collision is to be avoided. Collision probability has been evaluated due to the node’s predictable state. The generation of the warning message provides guidelines to the vehicles based on this probability. Based on the relative distance and speed with a pair of front and rear nodes the expected state can be represented. The probability of avoiding collision [
where ρs is the probability of the number of collisions among the nodes, η is the relative distance among the nodes and vf and vr represents the velocity of the front and the rear node respectively.
where, ρc denotes the collision probability and ρm denotes the maximum collision probability. The proposed method further computes the probability depending on the direction of the nodes. Since highways are bi-directional, at a given time the nodes can move in the opposite directions. The probability of a collision between two or more nodes traveling on the opposite sides of the highway may be higher depending on their relative distance and relative speed (as indicated in
where, k indicates the hamming distance. This determines the problem of imprecision in the probability evaluation for the nearby nodes and also for the nodes moving in the opposite directions. Consider the collision probability derived using
In the previous step the CH has been selected using the opposition based learning algorithm. After a certain time interval, the CH of the individual zones that plays an important role may be updated and validated correspondingly. Congestion may occur while broadcasting the messages over the VANET channels (i.e., the message channel turns into surpass via communication, and event-driven communications). It has been observed that an increased number of vehicles in the cluster area are endeavoring to transmit concurrently in the impenetrable circumstance. This indeed would reduce the packet delivery ratio and thus congestion may encounter. Therefore, the ANFIS based model has been introduced for updating and validating the CH and for enhancing the data transmission process.
The ANFIS is a multi-layer feed-forward network that includes both the terminals and the directional links. The ANFIS model functions with respect to the ambiguous Sugeno model with an adaptive system structure that supports both the learning and the adaptation based procedures. For example, two inputs provided through ‘x’ and ‘y’ with the output ‘z’ have been used for the ambiguous logical inference. Let the rule base have two fuzzy “if-then” rules of the Takagi and the Sugeno’s type [
Under the Rule 1 and Rule 2, Zi has been observed as the output around the fuzzy area stated by the fuzzy rules, the fuzzy sets have been denoted as Ai and Bi where as pi, ri, and qi represent the acquired design parameters for the training procedures. The ANFIS architecture employs these rules as represented in
Here, the ANFIS model has been trained using the grid partition technique. The proposed ANFIS based prediction model predicts the CH based on the inputs such as location distance, velocity and buffer size of the vehicles. For efficient creation of the CH, the ANFIS prediction model has been learned with the samples acquired from the initial EPIOA based clustering algorithm. Therefore, there is no need of any additional training of the ANFIS and is thus trained with respect to the initial node conditions. Based on the training data, the ANFIS can be trained and updated. This updated ANFIS engine has been named as the self learning CH predictor. After certain time interval, the ANFIS based prediction model would predict the particular node as the CH or not. It’s appears to be a rapid process and hence the CH formation delay appears to be less.
After the CH formation and validation, the data transmission from the source to the destination would be accomplished. Therefore, in order to enhance the data transmission between the nodes, the dijkstra’s algorithm can be adopted and utilized for computing the shortest path for the encountered data transmission process. The Dijkstra’s algorithm was initially proposed in the year 1956 by Edsger Dijkstra and published in the year 1959 [
Results of the proposed methodology have been analyzed with different performance metrics like: Packet Delivery Ratio (PDR), end to end delay and Cluster Head (CH) formation delay. Also, the proposed method has been analyzed and contrasted with the existing hop-by-hop technique.
The average ratio of successfully received packets at the sink node to the total number of packets generated in the source node.
It is the time variation among the communication data source from the destination. Data transfer between the sources to the destination may be lost due to the node coverage area or their location.
Initially, the CH selection would be done based on the EPIOA based clustering algorithm, after the t-time slot the CH updating or validating procedures would be accomplished based on the ANFIS prediction. This time duration of the CH update is called as the CH formation delay.
Parameters | Value |
---|---|
No. of Lane | 2 |
Length | 4 km |
Zones | 8 |
Packet size | 500 |
No. of vehicle | [50,100,150] |
Sensor radius | [80,100,130,160] m |
Change in velocity | [5,10,20,30] km/h |
In the first test case, the performance of the proposed method has been analyzed and derived from various velocity levels of the vehicles such as 5, 10, 20, and 30.
Proposed scheme | Existing scheme | |||||||
---|---|---|---|---|---|---|---|---|
Velocity (km/h) | ||||||||
PDR | 0.3661 | 0.7042 | 0.5070 | 0.3098 | 0.3483 | 0.6699 | 0.4823 | 0.2947 |
CH formation | 0.2876 | 0.2865 | 0.2865 | 0.2865 | 0.5753 | 0.5731 | 0.5731 | 0.5731 |
End to end delay | 1.0428 | 2.7285 | 1.5857 | 1.1428 | 5.7857 | 4.8857 | 4.6 | 4.7428 |
In the second test case, the performance of the proposed method has been analyzed with different sensor radius.
As shown in
The performance of the proposed and the existing method for the end to end delay aspect has been illustrated in
Proposed scheme | Existing scheme | |||||||
---|---|---|---|---|---|---|---|---|
Sensor radius (m) | ||||||||
PDR | 0.1549 | 0.2957 | 0.7323 | 0.4929 | 0.1473 | 0.2813 | 0.6967 | 0.4689 |
CH formation | 0.2875 | 0.2865 | 0.2865 | 0.2865 | 0.5750 | 0.5731 | 0.5731 | 0.5731 |
End to end delay | 0.4714 | 0.9571 | 2.4857 | 1.9285 | 3.3285 | 4.1428 | 5.1857 | 5.3 |
Therefore, the proposed scheme has outperformed in terms of the end to end delay aspect. The performance of the proposed and the existing method for the packet delivery ratio has been illustrated in
To enhance the traffic management in VANET, the EPIOA based clustering methodology has been proposed initially for selecting the optimal CH. The CH has been used for reducing the collision of the vehicles and for enhancing the data transmission procedures among the nodes efficiently. In VANET, the nodes appear to be dynamic in nature. Therefore, for updating and validating the CH and for enhancing the data transmission processes the ANFIS based prediction model has been proposed for updating and validating the CH. Additionally, for enhancing the data transmission process, the dijkstra’s technique has been proposed for identifying the shortest path for accomplishing the data transmission procedure. Also, the proposed method has been analyzed under two conditions such as different velocities of the vehicles and the sensor radius. The results show that the proposed methodology outperforms the existing technique in terms of the CH formation delay, the end to end delay and the packet delivery ratio.