Digital Signal Processing Reference
In-Depth Information
becoming very complex. To make an appropriate DT model, we need to predict at
DT, what kind of environments the wireless terminal will face and model these en-
vironments. This task is very cumbersome, which is the main reason why validation
of research results is evolving from simple channel models over network simula-
tors [42] to elaborate real-life testbeds, e.g., the ORBIT grid of WinLab [43], the
IBBT iLab.t [44] or the BOWL project of TU Berlin [45]. The task of validating re-
search results remains very important! However full validation cannot realistically
be achieved at DT anymore. As a result, the framework should be flexible enough
to adapt to new and unforeseen RT situations. This is heavily supported by the CR
paradigm.
3.4.2.2 Identify Control Dimensions
②
One of the assumptions of the framework is the availability of control dimensions
or knobs to control or steer the system's performance. The focus of this work is on
the radio knobs, so the lower communication layers, so no transport and application
layer knobs are considered. The number of knobs available in a system, and the
range to set them can be implementation dependent. As a control layer to handle
a wide variety of control dimensions, the CR is pushing systems to extend their
control dimensions.
Control Dimensions (Knobs):
(K
i,j
)
,1
k
. For a given wire-
less network architecture consisting of
n
terminals,
k
control knobs or dimensions
exist for each terminal, such as modulation, code rate, output power. The control di-
mension settings are discrete in real implementations, inter-dependent and together
have a complex non-linear influence. We define a setting of all knobs
j
for terminal
i
to be configuration point
K
i
.
In this topic, we also consider actions. Actions represent a change over time,
from one knob setting to the other. As the system can only control the knobs, it is
the only way of acting and interacting with the environment.
Actions:
(A
i
)
,1
≤
i
≤
n
,1
≤
j
≤
n
. For all
n
wireless terminals, an action at time
t
k
is
defined as the relative change in its configuration point from
t
k
to
t
k
+
1
:
A
i
|
t
k
≤
i
≤
=
(
K
i
|
t
k
→
K
i
|
t
k
+
1
)
. The actions are selected by the RT procedure and are used to
exploit and explore the environment.
3.4.2.3 Identify Dynamics
③
As discussed above, the dynamics of the wireless environment are the main drivers
to establish an elaborate hybrid DT/RT framework that allows adaptation to these
dynamics. The important dynamics are those that influence the performance of the
wireless terminal significantly. These should be identified and represented in the
environment model.
Dynamics are often difficult to model, especially those resulting from multi-user
interactions. In [46], we have illustrated this for large-scale IEEE 802.11 networks.
It is hence important to provide procedures flexible enough to adapt their behavior
at RT to account for modeling inaccuracies at DT.
