An improved PRoPHET - Random forest based optimized multi-copy routing for opportunistic IoT networks

Abstract

Opportunistic networks are one of the important categories of ad hoc networks in Internet of Things (IoT), which considers human social activities like daily routines, activities and many more to provide efficient communication. In opportunistic networks, mobile nodes are used to establish communication between nodes despite of non-availability of a dedicated route between them. Furthermore, nodes don’t acquire any knowledge in advance about the characteristics of the network such as the network topology and the location of the other nodes. Hence, designing a routing algorithm becomes a challenging task since traditional routing protocols used in the Internet are not feasible for the characteristics inherent type of network. The proposed work propounds a multi-copy routing algorithm based on machine learning named iPRoPHET or improved PRoPHET (Probability routing protocol using history of encounters and transitivity). iPRoPHET, uses dynamically changing contextual information of nodes and the delivery probability of PRoPHET to carry out message transfer. The iPRoPHET uses machine learning classifier known as random forest to classify the node as a reliable forwarder or a non-reliable forwarder based on the supplied contextual information during each routing decision. The classifier trained with large amount of data extracted using simulation leads to precise classification of the nodes as reliable or unreliable nodes for carrying out the routing task. Obtained results from the simulation proves that the proposed algorithm outperforms with respect to delivery probability, hop count, overhead ratio, latency but over costs with respect to average buffer time in par with similar multi-copy routing algorithms. The uniqueness of this paper lies in data extraction, categorization and training the model to obtain reliable and unreliable nodes to facilitate efficient multi-copy routing in IoT communication.