A CONVERSATION WITH THE EDITOR
The Future of Sensors: A
Conversation with David Rejeski
Adena Schutzberg
David Rejeski has been thinking about sensors for
some time. He’s the Director of the Foresight and
Governance Project at the Woodrow Wilson International
Center for Scholars. He specializes in technology policy
and assessment, environmental policy, and strategic
planning.
Like the federal highway program and the
currently-in-development NSDI, we are also on the way to
an enhanced sensor infrastructure. The one we have now,
says Rejeski, is geared to stationary sensing.
Environmental work in the past has focused on monitoring
pollutants at addresses, that is, at power plants and
hazardous waste sites. While we’ve not solved all of the
challenges these pose, we have a pretty good handle on
what is going on at the point location sources, he offers.
Changing How and Where We Sense
Today’s environmental challenges, Rejeski
explains, are becoming more low-level, chronic, mobile and
distributed; in many ways, a monitoring nightmare. Most of
our existing environmental monitoring systems, built on a
few static sensors, can’t track thousands, or tens of
thousands, of mobile sources of pollution, such as autos
or farmers fertilizing their fields. What we need, he
feels, is a new infrastructure. The good news is that much
of that infrastructure
is under development. “The environmental
community needs to look at every expansion of our
networked infrastructure as a potential platform for
sensing,” Rejeski notes. “Automobiles, cell phones,
appliances, radio frequency ID tags; everything with a
network connection is up for grabs. Remember that
today’s car has more computing power than the Apollo 13
spacecraft, so it’s no great leap to add on some sensor
packages and a communications system. Many cars already
have data loggers and a GPS transducer.”
Other possible sensing nodes are “cell phones, or
whatever portable devices we carry in the coming years.
These are gaining computing power and may be able to
collect data, do some low-level processing and then send
it off to a receiver every hour or so or serve as
communications nodes for other sensors scattered
throughout our environment. Continuous data streaming at
this point is impractical since it drains batteries
quickly. The idea would be to have these tasks be handled
‘in the background’ of the devices as owners use them
as phones or personal computing devices.”
Rejeski doesn’t deny that privacy issues rear
their heads as sensor platforms become more ubiquitous and
invisible. Issues such as how the data is collected, used,
and shared, as well as the anonymity (or not) of the owner
will need to be seriously examined before such systems are
deployed. Rejeski’s sense is that a new generation of
environmental volunteers who see the potential benefits of
such data may agree to carry a sensor. “Those who are
particularly sensitive to some pollutants may be
interested in understanding their exposure or being part
of a team of data gathers who help society develop a
better picture of pollution” he suggests. Today,
scientists often have access to data interpolated from a
few sensors, typically many miles apart. As the public
appreciates the value of more data points, they may be
more willing to help out. Also, Rejeski offers, experience
with location-based services (and even location-based
games) may help the younger generation to be active
proponents of this vision.
The Role of Homeland Security
Funding for homeland security may accelerate the
deployment of sensor networks, just as it’s pushed other
infrastructure projects, Rejeski envisions. Currently
there are limited numbers of sensors per pollutant and per
geographic region. However, to detect various biological,
chemical, or radiological threats, authorities will need
more. And, other agencies may be able to “piggy back”
on at least the communications network that’s created.
He muses that while the Department of Homeland Security
may not want to share its real-time data on airborne
pathogens, it may be amenable to hosting EPA’s sensors
on its network.
I asked Rejeski about the costs involved in
developing a sensor infrastructure. There are essentially
two cost points, he explained: the network infrastructure
and the sensor packages themselves. The networks are
currently under development for other purposes, so the
costs may be relatively low to the environmental
community. The sensor package choices and pricing are
dropping. The fact that it’s possible to buy
self-contained sensor packages for a variety of
measurement tasks and deploy them, effectively, “right
out of the box” is a huge step forward. It’s something
that’s occurred in just the last year or so.
Rejeski explains that early users of these networks
are already exploring their potential, tracking the
microclimate in wineries (Figure 1), the temperature of
forest fires, or the nesting habits of birds.
Environmental applications will typically fall into two
categories: the realm of “exploration,” that is,
answering the question, “what’s it like out there?”
(Figure 2) and those used for regulation. Sensors
distributed in bird (Figure 3) or animal habitats fall
into the first category. They keep track of a variety of
environmental factors, and equally important, keep people
out of the fragile habitats. Sensors used for intelligence
gathering for the military do the same things. Regulatory
applications for sensor nets are much further away because
the massive investments in the existing monitoring
infrastructure make switching to new technologies
difficult. The burden of proof for the new systems will be
very high and a variety of “proof of concept”
applications may be needed before regulators feel
comfortable switching technologies. Early use of GPS and
other sensors to charge fees for driving in London, or to
determine insurance costs in Oregon, are still generating
quite a lot of policy debate.
Steps Forward
The biggest step forward that the sensor network
provides, says Rejeski, is the ability to match the
spatial and temporal density of your measurements to the
problem. Consider that in a city there may be but one or
two air quality monitors. But on the ground, where we
live, there are likely to be significant differences in
air quality, differences that occur next to major traffic
arteries, or between parks and built-up areas and affect
human exposure to pollutants. The same holds true for
dozens of important environmental parameters within a
forest or watershed. The ability to deploy, at reasonable
cost, hundreds or thousands of sensors, instead of a few,
may change our deep understanding of environments at these
locations.
Rejeski is particularly interested in combining new
sources of sensed information. As sensors platforms mature
and prices drop, we may be able to integrate genetic
information with environmental information, he says. The
advent of lab-on-a-chip technologies will make it possible
to examine DNA outside of a laboratory setting. In time,
it will be possible to look at the world around us through
a genetic lens, for instance, searching for a particular
pathogen based on its DNA sequence and feeding that data
(along with other information) into a sensor net.
Challenges Ahead
Rejeski is quick to point out that despite a long
history using sensors, there are still many things
scientists don’t yet know. One open question concerns
determining sensors’ most effective use. For example,
it’s not yet known if it’s more advantageous to have a
single or a few larger, more sensitive sensors, than a
large number of less sensitive ones distributed in more
locations. Some work is underway on that topic, but we
don’t yet know the answer, he says, and the answer will
vary from application to application.
He ticks off a few other challenges: We need to
develop standard ways for sensors to communicate with one
another, and with the Internet. There are issues of sensor
calibration. If a sensor is hanging on the side of a
building or in a tree for a year or two, it will need to
perform its own self-calibration so that it doesn’t
report faulty data.
The technical challenges do remain, but perhaps the
most important one, according to Rejeski, is education of
the policymaking community. In a 2001 paper he reported
that of the 8 billion computer chips manufactured that
year, only 2 percent ended up inside computers. Where did
the rest go? Toasters, microwaves, cars, watches, etc. The
idea that the “stranded intelligence” in these
millions of devices might one day be linked together by a
network is still new to many people, including most
decision makers who remain glued to their PCs. Rejeski
points to the next generation of digital citizens, who
have grown up in a world of video games, web blogs, and
instant messaging, as the ones who may make a new
environmental sensor infrasturcture possible.
Back
