Image Processing Reference
In-Depth Information
By taking a look at the characteristics of the sensor nodes, we notice that the energy is spent for
three main functions: environment sensing, wireless communication, and local processing. Each of
these three components has to be optimized to obtain minimum energy consumption. For the sensing
of the environment component, the most energy efficient available sensors have to be used. From this
point of view, we can regard this component as a function of a specific application and a given sensor
technology.
The energy needed for transmitting data over the wireless channel dominates by far the energy
consumption inside a sensor node. More than that, it was previously shown that it is more efficient to
use a short-range multihop transmission scheme than sending data over large distances []. A new
strategy characteristic to the sensor networks was developed based on a trade-off between the last two
components and it is, in fact, one of the main characteristics of the sensor networks (see for example
techniques developed in Refs. [,]). Instead of blindly routing packets through the network, the
sensor nodes will act based on the content of the packet [].
Let us suppose that a certain event took place. All nodes that sensed it characterized the event with
some piece of data that needs to be sent to the interested nodes. Many similar data packets are created,
or at least, some redundancy exists in the packets to be forwarded. To reduce the traffic, each node on
the communication path should examine the contents of the packet it has to forward. hen it should
aggregate all the data related to a particular event into one single packet, eliminating the redundant
information. he reduction of traffic by using this mechanism is substantial. Another consequence
of this mechanism is that the user will not receive any raw data, but only high-level characterizations
of the events. his makes us think of the sensor network as a self-contained tool, a distributed network
that collects and processes information.
From an algorithmic point of view, the local strategies employed by sensor nodes have as a global
goal to extend the overall lifetime of the network. he notion of lifetime of the network usually hides
one of the following interpretations: one can refer to it as the time passed since power on and a
particular event such as the energy depletion of the first node or of % of the nodes, or even the
moment when the network is split in several subnetworks. No matter which of these concepts will be
used, the nodes will choose to participate in the collaborative protocols following a strategy that will
maximize the overall network lifetime.
To be able to meet the goal of prolonged lifetime, each sensor node should
•
Spend all the idle time in a deep power down mode, thus using an insignificant amount
of energy.
•
When active, employ scheduling schemes that take into consideration voltage and fre-
quency scaling.
It is interesting to note, at the same time, the contradictory wireless industry trends and the
requirements for the wireless sensor nodes. he industry focuses at the moment in acquiring more
bits/s/Hz while the sensor nodes need more bits/euro/nJ. From the transmission range point of view,
the sensor nodes need only limited transmission range to be able to use an optimal calculated energy
consumption, while the industry is interested in delivering higher transmission ranges for the radios.
Nevertheless,theradiosdesignednowadaystendtobeasreliableaspossible,whileawirelesssensor
network is based on the assumption that failures are regarded as a regular event.
Energy is not the only resource the sensor nodes have to worry about. he processing power and
memory are also limited. Large local data storage cannot be employed, so strategies need to be devel-
oped to store the most important data in a distributed fashion and to report the important events
to the outside world. A feature that helps dealing with these issues is the heterogeneity of the net-
work. here might be several types of devices deployed. Resource-poor nodes should be able to ask
more powerful nodes to perform complicated computations. At the same time, several nodes could
associatethemselvestoperformthecomputationsinadistributedfashion.
