KNNR:K-nearest neighbour classification based routing protocol for opportunistic networks

Abstract

Opportunistic Networks (OppNets) are an extension of Mobile Adhoc Networks (MANETs), where no assumption is made regarding the preexistent path between the source and the destination node. Hence, the nodes in OppNets are required to rely upon intermediate nodes for successful message delivery. Therefore, the biggest challenge in OppNets for a carrier node is to make a decision whether the neighbour node will be a good carrier for the message in the future or not. Hence, in this paper a new routing protocol called K-Nearest Neighbour based Routing protocol (KNNR) is proposed which judicially forwards the message through intermediate nodes towards the destination. The proposed protocol initially stores the past behaviour of nodes in a dataset. Whenever a decision has to be made related to an intermediate node, the protocol studies this dataset and finds instances that closely resemble the intermediate node based on their network parameters using K-Nearest Neighhbour (KNN) algorithm. To evaluate the efficiency of the proposed protocol simulation results are compared with the existing routing protocols i.e. Epidemic, HBPR and ProPHET. It was observed that the KNNR protocol efficiently reduces average latency, overhead ratio and average hop count while at the same time increases the message delivery probability.