Digital Signal Processing Reference
In-Depth Information
3.4.2.6 Determine DT Procedure
⑦
Finally, a DT procedure is developed. As mentioned before, the DT procedure
should be generic enough (parameterizable) to be able to efficiently handle as many
RT situations as possible. The RT procedure can be expressed as a fixed procedure
that should be carried out. Alternative, the procedure can be defined in a way that
enables run time learning or calibration. The RT learning engine then adapts the
DT procedure according to the encountered RT situation, which is learned through
feedback. A DT procedure is established for each possible scenario. The DT proce-
dure outputs an action distribution, which has to be tuned based on feedback, as a
function of the environment.
For instance, in Chap. 7 we introduce the hidden terminal starvation problem.
For this scenario, we know that the actions: increase rate, increase carrier sense
threshold and decrease power are dominated by their antagonists. The DT procedure
for this scenario would then look like:
=
p(
decrease transmission rate
)
0
,
(3.1)
p(
increase transmission rate
)
=
α
1
,
(3.2)
p(
decrease carrier sense threshold
)
=
α
2
,
(3.3)
p(
increase carrier sense threshold
)
=
0
,
(3.4)
=
p(
increase transmit power
)
α
3
,
(3.5)
p(
decrease transmit power
)
=
0
,
(3.6)
where
α
1
...
3
are the parameters the RT learning engine needs to learn for the ob-
served RT situation.
Alternatively, in Chap. 6 we define a RT procedure for the energy-efficient trans-
mission of multiple terminals in a single wireless network. The RT procedure in that
case will be defined completely at DT assuming that the environment, hardware and
multi-user interaction can be sufficiently isolated at DT.
Due to switching cost, it can be suboptimal for the terminal to change its behav-
ior after it has detected a scenario change. Hence, possible gains should always be
compared with the cost needed to switch from one DT procedure to another.
3.4.3 Run-Time Operation
The challenge at RT is to determine, for each user, the optimal configuration. This
requires to monitor the environment, to identify the observed scenario, to execute
the RT procedure and to calibrate it. As mentioned in the introduction of this chapter,
the different steps in the RT flow are the most important in the smart and cognitive
radio operation. As a result, in each of the following chapters, different case studies
will be introduced that focus each more on a specific step of the RT flow.
