Professional Documents
Culture Documents
INTEGRATED
NETWORK
SENSORS
For pervasive computing performance, exploit the
physical limits of these densely distributed networks of
embedded sensors, controls, and processors.
W
ireless integrated network sen- [2–5]. Scales range from local to global for applica-
sors (WINS) provide distrib- tions in medicine, security, factory automation, envi-
uted network and Internet ronmental monitoring, and condition-based
access to sensors, controls, and maintenance. Compact geometry and low cost allow
processors deeply embedded in WINS to be embedded and distributed at a fraction
equipment, facilities, and the of the cost of conventional wireline sensor and actu-
environment. The WINS network represents a new ator systems.
monitoring and control capability for applications WINS opportunities depend on development of a
in such industries as transportation, manufacturing, scalable, low-cost, sensor-network architecture. Such
health care, environmental oversight, and safety and applications require delivery of sensor information to
security. WINS combine microsensor technology the user at a low bit rate through low-power trans-
and low-power signal processing, computation, and ceivers. Continuous sensor signal processing enables
low-cost wireless networking in a compact system. the constant monitoring of events in an environment
Recent advances in integrated circuit technology in which short message packets would suffice. Future
have enabled construction of far more capable yet applications of distributed embedded processors and
inexpensive sensors, radios, and sensors will require vast numbers of devices. Con-
processors, allowing mass ventional methods of sensor networking represent an
production of sophisti- impractical demand on cable installation and net-
cated systems linking work bandwidth. Processing at the source would
the physical world to drastically reduce the financial, computational, and
digital data networks management burden on communication system
(b)
(a) (c)
components, networks, and human resources. undo the dispersion [12]. Finally, many obstructions
Here, we limit ourselves to a security application can render electromagnetic sensors useless. Regardless
designed to detect and identify threats within some of the size of the sensor array, objects behind walls or
geographic region and report the decisions concerning under dense foliage cannot be detected.
the presence and nature of such threats to a remote As a simple example, consider the number of pix-
observer via the Internet. In the context of this appli- els needed to cover a particular area at a specified res-
cation, we describe the physical principles leading to olution. The geometry of similar triangles reveals that
consideration of dense sensor networks, outline how the same number of pixels is needed whether the pix-
energy and bandwidth constraints compel a distrib- els are concentrated in one large array or distributed
uted and layered signal processing architecture, outline among many devices. For free space with no obstruc-
why network self-organization and reconfiguration are tions, we would typically favor the large array, since
essential, discuss how to embed WINS nodes in the there are no communications costs for moving infor-
Internet, and describe a prototype platform enabling mation from the pixels to the processor. However,
these functions, including remote Internet control and coverage of a large area implies the need to track mul-
analysis of sensor-network operation. tiple targets (a very difficult problem), and almost
every security scenario of interest involves heavily
Physical Principles cluttered environments complicated by obstructed
When are distributed sensors better than a single lines of sight. Thus, if the system is to detect objects
large device, given the high cost of design implicit in reliably, it has to be distributed, whatever the net-
having to create a self-organizing cooperative net- working cost.
work? What are the fundamental limits in sensing, There are also example situations (such as radar) in
detection theory, communications, and signal pro- which it is better to concentrate the elements, typi-
cessing driving the design of a network of distrib- cally where it is not possible to get sensors close to tar-
uted sensors? gets. There are also many situations in which it is
Propagation laws for sensing. All signals decay with possible to place sensors in proximity to targets,
distance as a wavefront expands. For example, in free bringing many advantages.
space, electromagnetic waves decay in intensity as the Detection and estimation theory fundamentals. A
square of the distance; in other media, they are sub- detector is given a set of observables {Xj} to determine
ject to absorption and scattering effects that can which of several hypotheses {hi} is true. These observ-
induce even steeper declines in intensity with dis- ables may, for example, be the sampled output of a
tance. Many media are also dispersive (such as via seismic sensor. The signal includes not only the
multipath or low-pass filtering effects), so a distant response to the desired target (such as a nearby pedes-
sensor requires such costly operations as deconvolu- trian) but background noise and interference from
tion (channel estimation and inversion) to partially other seismic sources. A hypothesis might include the
greater the likelihood of having a link with suffi- works should be designed so the radio is off as much
ciently small distance and shadowing losses. The sig- of the time as possible and otherwise transmits only at
nals then effectively hop around obstacles. the minimum required level.
Exploitation of these forms of diversity can lead to Processing also gets cheaper with time but is not
orders of magnitude reduction in the energy required yet free. Because application-specific integrated cir-
to transmit data from one location in a WINS net- cuits (ASICs) can clock at much lower speeds and use
work to another; the energy cost is then dominated by less numerical precision, they consume several orders
the reception and retransmission energy costs of the of magnitude less energy than digital signal processors
radio transceivers for dense peer-to-peer networks. (DSPs). While the line between dedicated processors
Radio systems involve a close connection between and general-purpose (more easily programmed)
networking strategy and physical layer. The connec- machines is constantly shifting, generally speaking, a
tion is even stronger in light of the multiple-access mixed architecture is needed for computational sys-
nature of the channel, since interference among users tems dealing with connections to the physical world.
is often the limiting impairment; the management of The ratio in die area between the two approaches—
multiple-access interference is explored in [6]. ASIC and DSP—scales with technological change, so
Energy consumption in integrated circuits. Unfor- ASICs maintain a cost advantage over many chip gen-
tunately, there are limits to the energy efficiency of erations. Convenient programmability across several
complimentary metal oxide semiconductor (CMOS) orders of magnitude of energy consumption and data
communications and signal-processing circuits. Over- processing requirements is a worthy research goal for
all system costs cannot be low if the energy system is pervasive computing. In the meantime, while
large. A CMOS transistor pair draws power each time researchers continue to pursue that goal, multiproces-
it is flipped. The power used is roughly proportional sor systems are needed in WINS.
to the product of the switching frequency, the area of
the transistor (related to device capacitance), and the Signal-Processing Architectures
square of the voltage swing. Thus, power consump- Security applications require constant vigilance by at
tion for any given operation drops roughly as the least a subset of the sensors. We want a low false-
fourth power of feature size. The components that alarm rate and a high detection probability. So as
consumption for reception is nearly the same as that reduced operating power, improved bit rate,
of transmission, the protocol should be designed so improved bit error rate, improved communication
radios are off as much of the time as possible. That privacy (by way of reduced transmit power), simpli-
is, a device’s medium access control (MAC) address fied protocols, and reduced cost. These benefits are
in a network should include some variant of time- not obtained simultaneously but need to be extracted
division multiple access. individually, depending on design emphasis.
A time-division protocol requires that the radios In network design today, architects also have to
exchange short messages periodically to maintain address: How can Internet protocols, including TCP
local synchronism. It is not necessary for all nodes to and IPv6, be employed within sensor networks? While
have the same global clock, but the local variations it is desirable to not have to develop new protocols or
from link to link should be small to minimize the perform protocol conversion at gateways, several fac-
guard times between slots and enable cooperative sig- tors demand custom solutions. First, IPv6 is not truly
nal processing functions, including fusion and beam- self-assembling; while addresses can be obtained from
forming. The messages can combine network a server, this particular protocol presupposes attach-
performance information, maintenance of synchro- ment at lower levels already. Second, present-day
nization, and reservation requests for bandwidth for Internet protocols take little account of the unreliabil-
longer packets. The abundant bandwidth resulting ity of physical channels or the need to conserve energy,
from the spatial reuse of frequencies and local pro- focusing instead on support for a wide range of traffic.
cessing ensures relatively few conflicts in these Embedded systems can achieve far higher efficiencies
requests, so simple mechanisms can be used. At least by exploiting the traffic’s limited nature.
interface
for event detection wireless
network
University of California, Los Angeles; the first gen- event classification interface
actuator control
eration of field-ready WINS devices and software and identification
The
Com puter
Alan Watt, Fabio Policarpo
Image
To Order 1998
Call Toll Free: 751 pages Hardcover
1.800.342.6626 (USA & Canada) ISBN 0-201-42298-0
1.212.626.0500 (outside US) ACM Order Number : 702980
Fax: 1.212.944.1318 Member Pr ice: $53.95
Nonmember Pr ice: $59.95
To order online: h t t p : / / w w w. a c m . o r g / c a t a l o g /