Information Technology Reference
The applied solution is to combine the shortest path metric with a “remaining
battery life” metric, i.e., to introduce some sort of reluctance in sensor nodes which
is inversely proportional to remaining battery life. In that way, efficiency of routing
is partly sacrificed, but the burden of routing is spread more evenly across the net-
work, prolonging its lifetime.
Another important aspect of routing in sensor networks is the issue of addressing.
Traditional networks use data-centric addressing - the basis for the addressing scheme
is some kind of datum, which in IP networks is the 4-byte number known as an IP
address. Most of the sensor network architectures use a similar approach; for example,
in TinyOS each node is assigned a 1-byte number address, with the gateway node
being assigned the address 0. However, in some use-cases, due to unstable nature of
sensor networks (and ad hoc networks in general) it is useful to have geographical
addressing: A message directed to a certain area would always reach a node
positioned in that area, even if node previously responsible for that area had already
used up its batteries and another node had taken its place [ 5 ]. This would not work
with the data-centric approach, as the data-centric address of the new node would
not be the same; the message addressed to the old node would be lost.
Finally, in some sensor networks addressing is not important at all; the only
important thing is to diffuse collected data to the sink (gateway node) so they could
be processed remotely. For information on a diffusion-based protocol, an interested
reader is directed to the Rumor Routing protocol [ 6 ].
Positioning and Localization
In many applications of sensor networks, it is important to record the actual
location where the measurement was made. For example, if there is a sensor net-
work monitoring the gradient of the temperature, each temperature measurement
would have to be accompanied by the information on where it was recorded. For
this to happen, each sensor node would have to have knowledge of its location. In
accepted terminology, positioning the sensor node means providing it with abso-
lute coordinates (such as longitude and latitude); localizing the sensor node means
providing it with information on its position relative to other sensor nodes.
An obvious choice is to provide all sensor nodes with a GPS device. While very
straightforward, this approach has significant drawbacks: GPS devices cost money
(a $50 GPS module deployed on 100 sensor nodes would increase the network cost
by $5,000), plus they significantly increase the rate of energy consumption, shortening
the battery life considerably.
Fortunately, it is not necessary to provide all nodes with a GPS if we want to have
full positioning information. In a 2-D case, if at least three devices are provided
with GPS information (either through a GPS device or if GPS positions are input
manually) all other devices can gradually establish positioning information through
a simple algorithm: