Wireless sensor networks - where parallel and distributed processing meets the real world

Summary form only given. Vast networks of intelligent sensors that are deeply embedded in physical world will revolutionize practices in the life sciences, civil engineering, manufacturing, security, agriculture, ubiquitous computing, and many other areas. They also present a wonderful new venue for parallel and distributed processing. Bandwidth, storage, and energy limitations make in-network processing essential - within the node and among collections of nodes. The algorithms should be resource efficient, but also deal with noise, uncertainly and dynamically changing connectivity. Ideally, application programming is done at the level of unstructured ensembles, rather than individual nodes. To explore these issues, we have built a series of inch-scale wireless sensor platforms, along with an operating system and networking substrate for vast collections of tiny, power-constrained devices - TinyOS. This open experimental platform is being used by hundreds of research projects internationally in a wide range of disciplines. This talk describes the challenges in making networks of such devices robust and programmable, including platform architecture, operating system design, network discovery and routing, and explores novel distributed algorithms developed for such networks.