Software Issues in Wireless Sensor Networks - Application and Multidisciplinary Aspects of Wireless Sensor Networks - page 30

Chapter 3
Software Issues in Wireless Sensor Networks
Aleksandar Crnjin
As is the case with all other computing devices, sensor nodes too have to be programmed
in order to be able to do something useful. Programming sensor nodes falls some-
where between programming desktop computers and small embedded devices:
There is only conceptual similarity between programming a home computer with
2 GHz CPU and gigabytes of RAM and storage, and programming a tiny device
with just a dozen kilobytes of RAM and an 8-bit processor running on just a few
kilohertz. Still, programming sensor networks is not all that similar to embedded
devices either: Programming embedded systems typically amounts to writing an
assembly program and loading it into programmable ROM of the embedded device,
while a sensor node usually has an operating system (such as TinyOS or ConTiki)
which provides support for basic I/O operations, operation of the radio transceiver,
and so on (Fig. 3.1 ).
Further difference is that sensor nodes are not conceived as being able to do
anything useful autonomously; they always operate as a part of a sensor network.
Having this in mind, programming paradigms for sensor networks have appeared
which allow a programmer to program the sensor network as a whole.
We will now introduce a taxonomy using which we can identify three different
approaches to programming sensor networks.
• Full Custom approach, in which we program a sensor node as if it were an
embedded device with no OS
• Scaled Traditional approach, in which a sensor node is programmed the way an
ordinary computer system is: using an operating system to operate the peripherals,
putting the hardware aspects below the abstraction boundary, and thus enabling
the programmer to put more effort into algorithms
• Spatial Programming approach, where we use the spatial programming
paradigm so as to program the whole network as a unique entity, putting the
internode communication aspects below the abstraction boundary as well