An Efficient Prediction-Based Routing in Disruption-Tolerant Networks

Routing is one of the most challenging, open problems in disruption-tolerant networks (DTNs) because of the short-lived wireless connectivity environment. To deal with this issue, researchers have investigated routing based on the prediction of future contacts, taking advantage of nodes' mobility history. However, most of the previous work focused on the prediction of whether two nodes would have a contact, without considering the time of the contact. This paper proposes predict and relay (PER), an efficient routing algorithm for DTNs, where nodes determine the probability distribution of future contact times and choose a proper next-hop in order to improve the end-to-end delivery probability. The algorithm is based on two observations: one is that nodes usually move around a set of well-visited landmark points instead of moving randomly; the other is that node mobility behavior is semi-deterministic and could be predicted once there is sufficient mobility history information. Specifically, our approach employs a time-homogeneous semi-Markov process model that describes node mobility as transitions between landmarks. Then, we extend it to handle the scenario where we consider the transition time between two landmarks. A simulation study shows that this approach improves the delivery ratio and also reduces the delivery latency compared to traditional DTN routing schemes.