Latency‐aware reinforced routing for opportunistic networks

Abstract

In opportunistic networks, the path connecting two nodes is not continuous at any time instant. In such an environment, routing is an extremely taxing word owing to the ever-changing nature of the network and random connections between nodes. Routing in such networks is done by a store carry forward mechanism, in which local information is used to make opportunistic routing decisions. In this study, the authors present a novel dynamic and intelligent self-learning routing protocol that is an improvement of the history-based routing protocol for opportunistic (HiBOp) networks. The proposed method presents a novel solution for the estimation of average latency between any two nodes, which is used along with reinforcement learning to dynamically learn the nodes' interactions. Simulation results on a real mobility trace (INFOCOM 2006) show that latency-aware reinforced routing for opportunistic network applied to HiBOp outperforms the original HiBOp protocol by 14.4% in terms of delivery probability, 15% in terms of average latency and 34.7% in terms of overhead ratio.