The survival of critical sensor data that is collected during manmade or natural disasters is arguably more important than traditional sensor-network design objectives, such as prolonging the sensors' lifetime. Under this project, a sensor data ghosting framework is being developed. This framework creates minimal data redundancy, known as sensor data ghosts, which roam around the sensor network toward a sink. Under sensor failures, the data ghosts are expedited toward the sink for their timely recovery. Recovery of the sensor data at the sink becomes feasible due to the availability of just enough data (generated due to data ghosting). Coding and networking solutions are being developed and integrated in a synergetic manner to achieve sensor data survivability and persistence. In particular, network coding approaches that map efficient channel codes over network graphs are being developed. On the networking side, a novel topology evolution solution that is capable of reacting rapidly to random failures and providing expedited delivery of critical data to the sink is being researched. This topology evolution approach rearranges a traditional sensor-network geometric topology toward a small-world network topology, which allows a small number of long-haul "shortcuts" toward the sink. Related networking research, such as directed diffusion and prioritized forwarding, is also being investigated under topology evolution. In addition to enabling new levels of sensor data survivability, this project has significantly broader theoretical and practical impact than the target sensor application. This includes the development of new types of "codes on network graphs" and approaches in adaptive network topologies.
