Context-adaptive forwarding in mobile opportunistic networks

When two opportunistic network peers encounter, utility functions are generally employed to select the messages that have to be exchanged, with the purpose of maximizing message delivery probability and reduce congestion. These functions compute weighted sums of various parameters, like centrality, similarity, and trust. Most of the existing solutions statically compute the weights based on offline observations and apply the same values regardless of a node's context. However, mobile networks are not necessarily constant in terms of behavior and characteristics, so the classic approach might not be suitable. The network might be split into sub-networks, which behave differently from each other. Thus, in this paper, we show that, by dynamically adjusting the behavior of a node based on its context, through the adjustment of the utility function on the fly, the opportunistic forwarding process can be improved. We show that nodes behave differently from each other and have different views of the network. Through real-life trace-based simulations, we prove that our solution is feasible and is able to improve an opportunistic network's performance from the standpoint of hit rate, latency, and delivery cost.