A dynamic data routing solution for opportunistic networks

Abstract

When two nodes in an opportunistic network meet, a utility function is generally employed to select the data that have to be exchanged between them, in order to maximize the chance of message delivery and to minimize congestion. The utility function computes weighted sums of various parameters, such as node centrality, similarity, trust, etc. Most of the existing solutions pre-compute the weights based on offline observations, and apply the same values regardless of a node's context. However, since mobile networks are extremely varied in terms of node type and behavior, this approach might prove not to be optimal. The network might be split into sub-networks that behave differently from each other (for instance, a group of nodes from the network might have many contacts, whereas some nodes might spend hours without encountering other peers). Thus, in this paper we wish to lay the foundation for a dynamic data routing solution for opportunistic networks. We show that nodes do indeed behave differently and have different views of the network, but that familiar nodes (i.e., that meet each other often for long periods of time) are alike in terms of behavior. Furthermore, we adapt an existing dissemination solution to dynamically adjust the weights of the utility function based on a node's context, and show through simulations that it behaves better than the static version. This would allow us to pre-compute the weights of the utility function and dynamically change them as a node's view of the network is modified, leading to a more efficient dissemination.