Compact Forbidden Set AODV Routing Information Protocol's Radial-Based function Set on a Neural Network

INTRODUCTION

One of the first distance-vector routing systems, uses the hop count as a routing metric (RIP). RIP prevents routing loops by restricting the number of hops permitted on a path from source to destination. The maximum number of hops permitted by RIP is 15, which places a cap on the size of networks that the protocol can handle. The routing tables of RIPv1 routers are updated every 30 seconds. Due to the limited routing tables in the early installations, the traffic was not heavy. However, when networks expanded in size, it became obvious that there could be a sizable traffic spike every 30 seconds even if routers were initialised at random times. (RIP) is a dynamic routing system that uses hop count as a routing metric to determine the best route between the source and destination networks. Operating at the OSI Network layer, it is a distance- vector routing protocol with an AD value of 120. Port 520 is used for the RIP protocol[1]. The hop count is the number of routers that connect the source and destination networks. The optimum route to a network is determined to have the fewest hops, and as such, it is added to the routing table. RIP minimises routing loops by capping the number of hops that may be included in a path.

The AODV routing protocol is the Compact Forbidden Set Routing Information Protocol which initializes the broadcast timer while doing route discovery, and it receives route replies that arrive only within the duration. If a response is received after the timer has expired, it will be rejected and discarded. Many factors, such as network density and traffic conditions, determine the time it takes for route responses to arrive.  As the network becomes denser, the time it takes for a route reply to reach the source node increases. of detected routes.

• A missing route raises the chances of missing the shortest route and also increases the number of hidden paths.

• In congested situations, the increasing number of rebroadcasts causes higher network traffic, which increases the number of missed cases.

• The overhead introduced by the broadcast technique has a significant impact on the network's throughput and performance.

In universal function approximation, Radial Basis Functions (RBF) are an alternative to Multi Layer Perceptrons (MLP). RBF was developed to solve multivariate interpolation and numerical analysis difficulties. In neural network applications, where the training and query targets are both continuous, their prospects are identical. RBF network performs a local mapping, whereas MLP performs a global mapping, in which all inputs result in an output. Only inputs near specified receptive fields will activate them. RBF neural networks have a major advantage over MLP in terms of performance.[2] In a multidimensional space, RBF neural networks optimally fit the training data. The global and local mappings in neural networks are shown in Figure1. RBF Network Structure A typical RBF network is a three-layered network. It is made up of three layers: an input layer, an output layer, and a hidden layer. The input for the pattern comes from the environment, which will be multi- dimensional. To train the data, it should be reduced. The Hidden Layer, which is the same size as the input layer, is made up of some locally tuned neurons describes each unit. where N is the size of the training sample and x — xj denotes the vector's Euclidean norm (x-xj). The RBF's centre is defined by the jth input data point xj, and the pattern vector x is applied to the input layer. The links connecting the source nodes to the nodes in the hidden units are thus direct connections with no weights, unlike a multilayer perceptron. The output layer has a smaller number of computational elements or a single linear computational unit.[4,5]provides the network's response. The Gaussian function is employed as the radial basis function for each computational unit in the network's hidden layer. All of the Gaussian hidden units are usually, but not always, given the same width o. In instances of this nature, the parameter that separates one hidden unit from another is the centre xj. The choice of the Gaussian function as the radial-basis function in RBF networks is based on the fact that it possesses a number of desired qualities.

Route Discovery, RBF-NN Approximation, and Traffic Aware Route Selection are three steps of the proposed AODV advancement technique for route selection and packet forwarding. This section delves into the above-mentioned stages in further depth. It depicts at any point in time, a snapshot of the network What happens while discovering route in this congested network is there will be delayed delivery of route replies which will be ignored and generates missing routes. This section addresses the issue by identifying hidden pathways and selecting a route based on the traffic conditions on the chosen road.

• Neighbor Discovery

Neighbor Discovery is the process of determining a set of neighbours located around any node in the MANET. The source node sends out a Hello message to the entire network and starts the hello timer. The hello reply message is received by the source node till the timeout expires. The hello message is received by nodes within the coverage of the source node, and when it is received, it generates a hello reply. Upon receiving any hello reply message, the source node adds the node ID from where the reply has been received to its neighbour table.[6,7]

• Route Discovery

The route finding is achieved by sending message into the network. The source node generates Route request message and broadcast on the network. The route request message is received by the neighbours node in the route request to its own ID and neighbour table. The node's neighbour table lists the nodes within its transmission range, and the node checks the presence of destination node ID in the neighbour table.[8.] If one of the neighbour IDs matches the destination ID, an route reply message is generated and sent back to the source node. If there is no such neighbour exist for a neighbour that receives the route request, it then rebroadcasts the same to its neighbour and the method is done recursively till all the node receives the message. When a node detects the presence of a destination ID in its neighbour database, it constructs a route reply message, adds its own ID at the end, and sends the reply packet back to the sender. In the first portion of Figure 3, the route request propagation is shown, with dotted lines indicating duplicate packets. As seen in part 2 of Figure 4, the route reply is unicast to the source node. The format of traffic message created in the suggested approach is shown in Table 3.1. The first column displays the packet format; the second column displays the node ID to which the packet must be delivered; and the third column displays the node's details, as seen below. 4:140:18:120:120:5 - The value 4 indicates the node ID, 140 indicates the remaining energy in Joules, 18 indicates the traffic in bytes, 120:120 indicates the x and y-axis coordinate coordinates, and 5 indicates the node's displacement speed. Similarly, the information in the next 4,5,6,7 columns is formatted in the same way. The seventh column indicates that node ID 6 has a link to the destination, has 190 Joules of leftover energy, and has an 8-byte-per- second traffic rate.[9,10] The value 1 represents the Figure 3.10 depicts the best route selection when there are several paths between the source and the destination. The best path is used for 60 seconds in a simulated scenario. Based on traffic conditions, path 1 is picked as the optimum path for the first 30 seconds, while path 2 is taken as the optimum path for the next 30 seconds.

SIMULATED RESULTS

The network simulator NS 2 is used to simulate the RBFNN-based AODV protocol. The initial network setup in the network simulator with 30 nodes is shown in Figure 5 Figure 6 depicts a snapshot of the source and destination during runtime, whereas Figure 7 depicts node mobility. And also depicts the many parameters that determine the network's QoS, including as throughput, packet delivery ratio, routing overhead, and average end-to-end delay.[13] The calculated results of one set of simulations in network simulator are shown in Figure 10.

To increase the performance of AODV routing and the performance of mobile ad hoc networks, traffic aware radial basis neural network based AODV routing has been presented. Unlike pure AODV routing, the proposed routing protocol takes into account a variety of aspects and employs a number of attributes like as speed, location, residual energy, and traffic conditions to select a specific path for forwarding data packets. By utilising radial basis neural networks, the suggested technique further includes dimensionality reduction and missing routes caused by the least constant timer values of route timer. [11.12]The suggested method and increasing network throughput by 1.5 percent. [15]When compared to AODV, the overhead is reduced by 17.5 percent, and the PDR is improved by 3%. The results and discussion section of the thesis contains a detailed comparison of the results.

1. Haiqing Liu, Licai Yang & Yao Zhang 2015, ‗Improved AODV Routing Protocol Based on Restricted Broadcasting by Communication Zones in Large-Scale VANET‘, Arabian Journal for Science and Engineering, Springer, vol. 40, no. 3, pp. 857-872. 2. Maamar, S, Bilami, A & Benmohamed, M 2011, ‗M-AODV: AODVvariant to improve the quality of service in MANETs‘, IJCSI International Journal of Computer Science Issues, vol. 8, no. 1, pp.429-436. 3. 3,.Mahesh Kumar Yadav, Ram Kishan Khola & Deepak Dembla 2012,‗Modeling, Analysis & Implementation of Improved AODV Routing Protocol in MANETs‘, International Journal of Computer Applications, vol. 41, no. 21, pp. 22-26. 4. Maleki, A & Sadeghzadeh, M 2014, ‗Improved AODV route recoveryin mobile ad-hoc networks using a genetic algorithm‘, ManagementScienceLetters, vol. 4, no. 9, pp. 2171-2176. 5. Natalia Castro Fernandes, Marcelo Duffles Donato Moreira & OttoCarlos Muniz Bandeira Duarte 2013, ‗An Efficient and RobustAddressing Protocol for Node Auto configuration in Ad HocNetworks‘, IEEE/ACM transactions on networking, vol. 21, no. 3, pp.845-856. 6. Mangrulkar, RS, Chavan, PV & Dagadkar, SN 2010, ‗ImprovingRoute Selection Mechanism using Trust Factor in AODV RoutingProtocol for MaNeT‘, International Journal of Computer Applications,vol. 7, no.10, pp. 36-39.7. Marco Conti & Silvia Giordano 2014, ‗Mobile Ad Hoc Networking:Milestones, Challenges, and New Research Directions‘, IEEECommunications Magazine, pp. 85-96. 7. Marek Natkaniec, Katarzyna Kosek-Szott, Szymon Szott & GiuseppeBianchi 2013, ‗A Survey of Medium Access Mechanisms for Providing QoS in Ad-Hoc Networks‘, IEEE communications surveys& tutorials, vol. 15, no. 2, pp. 592-620. ComputerScience, pp. 1-4. 9. Mawloud Omar, Sabrine Hedjaz, Souhila Rebouh, Katia Aouchar,Bournane Abbache & Abdelkamel Tari 2015, ‗On-demand sourcerouting with reduced packets protocol in mobile ad-hoc networks‘,Elsevier, International Journal of Electronics and Communications (AEÜ), vol. 69, pp. 1429–1436. 10. Meena Rao & Neeta Singh 2014, ‗An Improved Routing Protocol(AODV nthBR) for Efficient Routing in MANETs‘, in the proceedingsof the second International Conference on Advanced Computing,Networking and Informatics, Springer, vol.2, pp. 215-223. 11. Meghanathan, N 2012, ‗A MANET Multicast routing protocol forstable trees based on the inverse of Link Expiration Times‘, IEEEconsumer communication and networking conference, pp. 947-951. 12. Mitko Bogdanoski, Dimitar Bogatinov & Saso Gelev 2014,‗Performance Analysis of Statistical QoS Routing Schemes forMultihop Wireless Ad Hoc Networks‘, International Journal of FutureGeneration Communication and Networking, vol.7, no.5, pp. 119-128. 13. Mohammed Awad, Héctor Pomares, Ignacio Rojas, Osama Salameh &Mai Hamdon 2009, ‗Prediction of Time Series Using RBF NeuralNetworks: A New Approach of clustering‘, The International ArabJournal of Information Technology, vol. 6, no. 2, pp.138-144.

Research Scholar, Department of Computer Science, Capital University, Jharkhand