Systems Challenges in Tiered Sensor Networks

Опубликовано 6 сентября 2016, 6:19

Someday soon, a ninth-grade biology class, the EPA, and a police department may each seek information from the sensors in a single instrumented city park. How can the police department collect acoustics and other high-rate signals from the distributed sensor network, given the low-bandwidth radios these networks use? How will the biology class and police department be able to request different information from the network at the same time? How might the network combine separate user tasks to optimize the use of network resources? In summary, how can we turn this sensor network into reusable infrastructure? My work builds toward this infrastructure by providing the flexibility it requires – adapting to environmental conditions and supporting concurrent users – while operating under tight constraints.  The VanGo framework provides software filters to reduce high-rate data at its source, lessening communication and power overhead. PC-class devices, with their greater resources, calibrate and configure data reduction software running on constrained embedded sensing nodes. Tenet provides a way for these PC-class devices to construct tasks for the sensing tier, and a mote runtime that can support several tasks concurrently. The lower tier thus becomes flexible and generic, supporting a wide range of application services. Finally, planning for efficient lower-tier operation requires evaluating the constraints relevant to sensor network applications. Snack (Sensor Network Application Construction Kit) provides a configuration language and compiler so top-tier devices can express and solve this constraint-satisfaction problem, merging individually specified tasks into efficient combinations to make the best use of sensor resources.