Digital Signal Processing Reference
In-Depth Information
calibrating or learning the RT procedure. The full framework, including design time
and run time phases, is introduced next.
3.4 Proposed Control Framework
The ultimate CR is built on a RT-only framework and, hence, does not rely on any
DT model. As a result, modeling inaccuracies will not propagate to the control pro-
cedure at RT. Furthermore, a RT-only framework will be extremely flexible as each
scenario is learned through experience and interaction with the environment. Unfor-
tunately, RT algorithms to explore each situation and future evolutions in a holistic
way require enormous overhead. On the other hand, DT-only frameworks based on
worst-case operating conditions, are known to be severely suboptimal for dynamic
wireless networks.
In this topic, we present a framework that can exploit the flexibility, while main-
taining an acceptable complexity level. This framework relies on design control pro-
cedures at DT, that are efficiently executed at RT. This allows for an effective and
efficient RT operation. However, as the control procedures are built relying on DT
models, inaccuracies in these DT models may propagate to the control procedure.
Furthermore, creating DT models is becoming increasingly complex as wireless
terminals are facing very complex environments. In line with the CR paradigm, we
also embed a learning module to compensate these inaccuracies and to find suitable
operating points even in unforeseen situations.
This section is structured as follows. First, we detail the general design concepts
in Sect.
3.4.1
. In Sect.
3.4.2
, we describe the design flow. Finally, the RT operation
is detailed in Sect.
3.4.3
.
3.4.1 General Design Concepts
In this section, we describe the general design concepts of our framework. In
Fig.
3.1
, a condensed version of our proposed framework is shown. First, we
describe the impact of the increased flexibility provided by the SDR and OSA
paradigms on our control framework. Then, we discuss environment awareness and
why it is critical in the development of future CR systems. Finally, we discuss how
we can learn from extended interaction with the environment.
3.4.1.1 Control Dimensions
The quest for the ultimate CR is driven by new paradigms that increase the HW flex-
ibility (SDR) en the policy flexibility (OSA). As a result, the set of RT adjustable
control dimensions (also referred to as
knobs
) of a wireless terminal is steadily in-
creasing. It is important to rely on this knob set to reconfigure the system at RT. The
