Abstract
Pairwise contacts in Delay-Tolerant Networks (DTNs) for applications such as bus or smartphone based social networking commonly show some regular repeating patterns. Most existing routing protocols only implicitly exploit these patterns to predict future contacts. To enhance message delivery rate, most of the schemes allow messages to be replicated and forwarded to encountered nodes. However, there is no efficient mechanism for dynamically controlling message replication to achieve high message delivery rate with very low message overhead. In this paper, we present FGDR, a routing protocol designed for DTNs by leveraging fine-grained contact characterisation and dynamic message replication. In FGDR, the history contact is characterised in a fine-grained manner using a sliding window mechanism, and an up-to-date future contact prediction can be made based on the most recent history data. We design an efficient message replication scheme, in which replication is controlled in a fully decentralised manner by taking into account the expected message delivery rate, the replication history, and the quality of the encountered node. A replica can be generated only when it is necessary to fulfill the expected message delivery rate. We evaluate our scheme through trace-driven simulations, and results show FGDR can achieve much higher message delivery rate with lower message overhead in comparison with existing schemes.