Structure Discovery and Information Brokerage in Sensor Networks

Опубликовано 6 сентября 2016, 6:16

Sensor networks are distributed systems embedded in a physical space. This embedding, as well as the nature of the physical laws governing the phenomena observed or monitored by the network, can greatly influence the design of system architectures and information processing paradigms for such systems. In this talk we describe a number of lightweight tools for discovering the structure of the network at a high level, or of the signal landscapes it observes, and then distributing a small amount of this global knowledge to each node. We use this local data for globally-aware, efficient naming and routing, information dissemination and aggregation, and information discovery. We argue that this helper information blends well with current architecture proposals such as SP-SNA or Tenet, and is robust to low-level link volatility and other transient network effects. A number of the proposed techniques incorporate ideas from algebraic topology, computational geometry, and partial differential equations.