A Sensor Infrastructure
An information technology revolution for the life sciences, as well as
civil engineering, mechanical engineering, and agriculture is the
emergence of tiny micro-sensors that can be deployed in vast networks to
obtain environmental data at unprecedented spatial and temporal density.
The instrumentation is placed within the experiment, rather than the
experiment being performed within an instrument. Starting from
visionary work in SmartDust,
we have developed a series of small wireless sensor nodes, a TinyOS operating system, and
complete low-power RF network stack and have made all aspects openly
available. There are now nearly a hundred groups internationally using
our technology for research spanning from networking to in depth habitat
monitoring and emergency response. Although individual data rates are
low, the sheer number of devices connected continuously to the physical
world generate a tremendous amount of data. Moreover, given the power
and resource constraints of these devices, processing must be performed
within the network itself so that only the most valuable information is
transmitted and every bit of information that is conveyed is precious
enough to be stored. There is a wide range of open problems associated
with the potential merging of database and networking technology, and we
have developed an initial TinyDB which transforms the vast network of
tiny devices into a streaming, interactive database. The number of
potential applications of this technology is staggering and we need to
build a deep storage infrastructure to capture the volumes of streaming
sensor data in a manner that makes it available to processing and web
access.
We have completed substantial deployments in areas of monitoring
structural response to seismic shock, indoor environmental monitoring,
and long term monitoring of the nesting habits of certain migrating
seabirds. Each of these efforts presents the opportunity to obtain a
vast amount of data that was previously unobtainable. The existence of
such information is leading scientist-users to develop new analysis
techniques, progress on which is furthered by the retention of all
possible data. Eventually, the data processing algorithms will be
distributed over the sensor network itself.
|