摘要：Opportunistic networks have recently seen increasing interest in the networking community. They can serve a range of application scenarios, most of them being destination-less, i.e., without a-priori knowledge of who is the final destination of a message. In this paper, we explore the usage of data popularity for improving the efficiency of data forwarding in opportunistic networks. Whether a message will become popular or not is not known before disseminating it to users. Thus, popularity needs to be estimated in a distributed manner considering a local context. We propose Keetchi, a data forwarding protocol based on Q-Learning to give more preference to popular data rather than less popular data. Our extensive simulation comparison between Keetchi and the well known Epidemic protocol shows that the network overhead of data forwarding can be significantly reduced while keeping the delivery rate the same.
Loading...
