Digital Signal Processing Reference
In-Depth Information
IEEE 802.15.4 specifications, can be used to detect interference on the channel. As
a result, the energy consumption can be modeled as only varying with the number
of channels that are scanned (or listened to) in parallel:
E
tot
=
n
s
t
a
P
Rx
,
(5.1)
where
n
s
is the number of channels scanned and
t
a
is the active period per super-
frame.
In sensor networks, switching energy also presents a large portion of the con-
sumed energy. As we explain in Sect.
5.4
, we redefine the problem statement. The
goal of the current chapter is to have all terminals on the same channel, so switching
isn't necessary anymore. This way, a terminal that performs well in the redefined
problem, also minimizes switching energy.
5.3 Basic Solution: Random Frequency Selection
The simplest DFS algorithm is a scheme where nodes randomly (uniform distri-
bution) pick a channel every period. It can of course be expected that the average
delay in this Random Frequency Selection (RFS) scheme will be large. However,
since it does not rely on any coordination between the nodes and does not rely on
an environment model, it is very robust.
5.4 The Problem from a Different Angle
In the current chapter, our goal is to minimize the average packet delay to the sink.
This target function is unfortunately quite difficult to model, so we will approximate
it by inspecting the optimal solution. In any learning system, defining a target func-
tion, which can be measured at RT, is a vital step in the solution design (see [67,
Chap. I]).
As mentioned in Sect.
5.2
, we assumed the sensor network to have a low duty
cycle. Under this assumption a packet has the least delay if it can travel the fur-
thest distance each hop. Obviously, the optimal solution is then the one where all
terminals share a common channel. In that case, we can ensure that each packet,
independently of where it was generated, can travel the furthest distance and have
an efficient last hop.
We can now come up with a simple approximation of our target function: we
no longer look at the delay, but try to get all the terminals on the same channel.
Clearly, any solution presented in this chapter can thus also be used to dynamically
and distributed find a common control channel (CCC), which is of high importance
for Opportunistic Radio networks [68, 69].
