The capability of a mobility model to detect certain patterns of user behavior (e.g., favorite walks or walking habits) enables solutions for a number of challenging networking problems, including efficient opportunistic communications and handoff / cellular planning. We argue that the limited viewpoint of a single mobile node and its scarce resources (e.g., energy, memory or processing) are major obstacles for accurate estimations. Targeting at hybrid network environments, we offload prediction capabilities to the fixed nodes that may be available in the area, offering a global view and the capability of resource-demanding calculations. Here, we introduce a solution running on top of the infrastructure nodes that: (i) implements a mobility model which provides a number of mobility forecasts to the mobile users in the area, (ii) supports proactively the routing decisions of opportunistic mobile devices being taken at times there is not connectivity. We introduce the corresponding semi-Markov model and demonstrate its efficiency using scenarios deployed in a pre-selected city center, where a number of mobile nodes seek for Internet access.

Final publication is available at link.springer.com (Lecture Notes in Computer Science).