A
Remote Ecological Micro-Sensor Network
A
project funded by DARPA's Sensor IT program
Edoardo S.
Biagionia
K. W.
Bridgesb
Brian J.
S. Cheea
aDepartment
of Information and Computer Sciences, University of Hawaii at Manoa
bDepartment
of Botany, University of Hawaii at Manoa
This project is developing an
innovative `micro-sensor' network that will be applied to a high-priority
ecological problem. The field problem
being studied involves a combination of the remote visual surveillance of
federally listed rare, threatened or endangered plant species and the intensive
monitoring of the weather in the area in which the plants occur.
This ecological project
shares many of the DOD surveillance needs.
Our project provides the near-real time observing of important events
such as visitation by pollinators, consumption by herbivores and even human
visits, along with monitoring a number of weather conditions and events. This directly parallels the military's
requirements to `rapidly and accurately detect, track and report threats such
as winged and wheeled vehicles, persons, and bio-chemical agents' (BAA00-25).
There will be times when the
type of remote reconnaissance being developed by this project fulfills military
needs. For example, the military shares
a commitment to the protection of federally listed species that occur on its
lands. One of the species proposed in
the monitoring test here, Silene hawaiiensis1, has one of its
few known populations in the Pohakuloa Training Area on the Island of Hawaii.
The current state of our
surveillance of these vital resources is limited or non-existent. The sites with the rare plants are
necessarily remote. This means that
they are not visited very often so we have little idea of what is happening to
them or their environment. The sites
lack access to the conventional power and communications infrastructure. This limits our collection of environmental
data to an occasional `stored-data' recorder.
Weather data are first examined weeks or months after their
collection. The data collection is so
sparse that it can't be considered a network.
Because Internet access was not designed into the system, the data are
rarely shared with other people interested in the area.
We have a radical
remedy. We will have a dense network of
weather instruments. We will measure
not only what is happening at the site of the rare plants, but also in the
adjacent areas since we are equally interested in why the species are not
growing there, too. We will `visit' the
plants often so we can see what they are doing. When do they grow? When
do they flower? What is eating
them? By recording pictures of the
plants, we can spot these events and even review earlier pictures to gain more
insight into these life-event processes.
We will finally be able to apply management systems based on a
substantial understanding of how these plants function in their environment.
The Challenges of
Building a New Reconnaissance System
The first challenge is to
make good observations of our rare plants.
We have found that the current-generation of high-resolution digital
cameras (two megapixel) produces excellent results. We will point high-resolution cameras at the rare plants and the
environment (often you need to see the weather, not just make measurements). We will trigger the cameras periodically and
sometimes in response to specific events.
Our second challenge is to
spread out a network of digital weather instruments to measure the
environment. We need to get the
instruments that measure rainfall, wind, temperature, humidity and solar
radiation, into appropriate places.
Rare plants are often in microhabitats.
We need to measure what is inside these habitats, as well as what is
happening in the surrounding areas.
The third challenge involves
the physical communications environment.
We need to gather information from a variety of locations and move it
back to a place where it can be interpreted.
The communication environment concerns are complex because of
considerations for an actual deployment.
Since we will test the system's ability to monitor rare plants, we will
need to place our equipment in a variety of habitats, ranging from scattered
low shrubs to dense tropical forests.
Topographic extremes will vary, too.
Some sites are on gentle-sloping volcanic shields, while others are in
narrow, steep-sided valleys.
Line-of-sight locations are rare.
We need to get our data through the forests, around mountains and down
valleys. Environmental conditions can
be severe, too. Some locations frequently
freeze. Others have rainfall that is
highly acidic (often dropping to as low as pH 2.0).
Finally, we are challenged
to build a system that is reliable. We
need to have network nodes that automatically configure themselves. They must also adapt to their location and
role in the overall system. They must
conserve their resources, such as power.
They must also be cryptic so that don't interfere with what they are measuring
or observing and minimize the possibility of their being harmed or stolen.
Our Innovative Solution:
A Network of Pods.
We are building the
communication network using a series of `pods.' These pods have two functions: each pod can collect instrument
information and the pod can pass these data on to other pods. A pod is built around a power source,
embedded computer, GPS receiver and wireless communications gear. A plug permits the connection of a network
of weather instruments.
The system functions so that
data are sent from pod to pod until they reach a base station pod. This base station is located at a place with
convenient access to the Internet. In
the aggregate, we deploy one set of pods around an area being monitored. Another set of pods is arranged as a
communications chain, linking the monitored area back to the base station. All the pods, whether they are in the
monitored area or in the communications chain, have the potential to collect
data. This provides an
information-gathering network that is both intensive and extensive. These are the two scales that are needed for
the evaluation of many problems.
Each of the links in our
communication system is relatively short.
The pods are arranged so that their communications ranges overlap. This adds redundancy so that if a node in
the communication chain fails, for example, it is simply bypassed.
Our system is designed to
use pods that are relatively inexpensive, flexible, and easily deployed. The pods also need to be very robust. We expect them to endure a wide variety of
extreme environmental conditions. They
must provide an environmentally tolerant communications environment. For example, the pods need to respond to
difficulties, such as the removal of a network node or low-power conditions,
and adaptively change functions like the rate of data transmission so that they
perform adequately with the modified resources.
There are Three Types of
Pods
The pods are not all the
same. The most general type of pod is
the one that has just been discussed, the communications pod. This pod handles all the network communications
functions. In addition, it accepts a
network of plug-in weather sensors. The
specialized pod has all the functions of a communications pod, but
differs by physically housing instruments like a high-resolution digital
camera. Such equipment has much higher
data streaming requirement and greater power consumption. The third type is the base station pod. It functions to communicate, like the other
pods. In addition, it provides for the temporary
local storage and further processing of data and connects to the Internet. The base station is not used to collect
instrument data as it is expected to be located outside the area being studied.
Studying a `real world'
problem, such as the ecology of rare species, suggests some features needed by
the micro-sensor surveillance network that might otherwise be overlooked. For example, we have discovered that all of
the equipment that we deploy in the field must be cryptic and blend into the
environment. Anything that appears out
of place draws attention. All too
often, this results in some harm to the instrument. Equally important, the obvious presence of monitoring equipment
might bias the observations or even negate its purpose.
The need for cryptic pods
implies that the pods must use power very efficiently. Otherwise, they would require large
batteries and, perhaps, extensive photo voltaic arrays for recharging. We place power-conservation as one of the
primary design features, in part to keep the pods small and more easily hidden.
We have found that
high-resolution digital cameras provide a significant opportunity to make
critical observations that have not been feasible in the past. A simple example shows how this works. During the examination of a scene, the
person doing the interpretation sees what looks like a change. By having an on-line library of past images,
it is possible to view the scene, as it was earlier, and determine if there
actually has been a change.
We anticipate that we will
be able to schedule field visits so that they happen during times when critical
observations are important. Currently,
we go to the field when it is convenient or at regularly scheduled times. And, all too often, we remark, `We should
have been here a month ago.' Our
current way to schedule our visits is inefficient and misses important events. It is easy for us to anticipate that we will
better use our human resources by using the proposed system to help schedule
our field time so that we make critical observations at the optimal times.
We are working on a data
interpretation system (in a separate project that parallels the one being
proposed here) that combines four main sources of information. The current conditions in the area being
monitored come from data collected as proposed in this project. This is combined with the near-real time
weather data from METAR stations2 and a variety of satellite
images. These provide a regional context. Historical data are available to establish
standards and comparisons, both locally and regionally. Finally, weather predictions, such as NOAA's
rainfall prediction models3, are integrated into the system.
1 Images of Silene hawaiiensis, one of the rare
plants proposed for monitoring, can be found at http://www.botany.hawaii.edu/b308old/s97/scans/sil_info.htm
2 METAR data are hourly temperature, air pressure, wind
speed and direction and general weather observations made at thousands of airports
around the world. The data are
accumulated at a central site and are available soon after they are collected.
3 These
NOAA models provide 12 to 48 hour forecast plots of wind and rainfall. They are available at http://www.nws.noaa.gov/pr/hnl/pages/rsm_page.html.
Last Updated: 3/10/00