Island Hopping: Efficient Mobility-Assisted Forwarding in Partitioned Networks Natasa Sarafijanovic-Djukic Michal Piorkowski Matthias Grossglauser EPFL, LCA Mobility with Concentration Points Our main assumptions: heterogeneous spatial distribution concentration points (CPs): node density much higher than average stable over time Our main goal: exploit CPs to help ad-hoc networking routing in mobile partitioned networks Justification of our assumptions: real-life mobility traces Michal Piorkowski – LCA, EPFL 2 Exploiting Stable Concentration Points Mobile partitioned networks routing 2 If it is possible for nodes: to assign unique and stable labels to CPs 1 to know this labeled CP graph to know at which vertex the destination is Then: 4 D 3 S 5 routing could be done more efficient 3 Components of Our Routing Scheme the set of neighbors 2 1.Collaborative Graph Discovery CP graph 1 2. Last Encounter Locating destination location 3. Island Hopping 4 D 3 S 5 routing decisions 4 1. Collaborative Graph Discovery (COGRAD) Each node knows only the set of neighbors: 2 has no external information (such as GPS, or fixed beacons) 1 4 3 does not know if it moves or not Labeling: assign unique labels to CPs 5 maintain the labels stable as long as possible Edge Discovery: discover the labeled edges of the CP graph 5 Our Contributions Mobility model with CPs validation: taxi mobility traces Routing scheme for mobile partitioned networks COGRAD: discovering the CP graph by observing only the set of the neighbors Last Encounter Routing: locating nodes Island Hopping: mobility-assisted forwarding that exploits knowledge of CP graph destination location 6
© Copyright 2026 Paperzz