Improving the Performance of Epidemic Routing Protocol using Game Theory Approach for Non-cooperative Node in Delay Tolerant Network

Abstract

Routing is a challenging issue in Delay Tolerant Network (DTN), due to the problem of intermittent connection. Epidemic Routing Protocol (ERP) is the one DTN routing protocol, having the highest delivery ratio with minimum latency features. But these happen only when nodes cooperate. In reality, not all nodes may cooperate, because of limited resources and fewer social ties. Those non-cooperative nodes (NN) drop the message, which contributes to poor routing performance. The existing ERP not considered this behavior of nodes. In our study, a Game Theory Based ERP (GT-ERP) for the NN is developed. In this model, nodes select id 0 and 1 to act as cooperative and NN respectively. We measured the NN effect from [0-100] degree number of the NNs. The proposed GTERP motivates this NN to form cooperation among nodes. A theoretical framework for a rational node that maximizes its payoffs has been presented. These were due to nodes have a cooperative mechanism that is reaching Nash Equilibrium (NE) and the nodes never leave that state. Using GT, it is verified that the proposed GT-ERP achieved cooperation among nodes. The proposed GT-ERP is simulated on ONE Simulator. From the simulation result during the first phase the delivered packet increased from1020 to 1166, dropped packet decreased from 438 to 292, delivery ratio increased from 0.7 to 0.8 which is 10% improved by the proposed GT- ERP, the overhead ratio decreased from 14.05 to 13.56. This was due to the proposed GT-ERP has a cooperative mechanism. But the average latency increased from 2045.61 to 2356.21, because of the proposed GT-ERP used time in selecting a strategy. Lastly, as the number of NN reaches 100 the delivered packet increased from 254 to 955, the dropped packet decreased from 1204 to 503, the delivery ratio increased from 0.174 to 0.655, the overhead ratio decreased from 24.58 to 15.38 and the average latency decreased from 3827.98 to 3163.35. We have concluded that as the number of NN increased the proposed GT-ERP achieved better performance in all metrics including average latency in comparison to the original ERP.