Digital Signal Processing Reference
In-Depth Information
desired behavior that are both correct and representative of all the RT situations in
which the agent has to act. In uncharted territory, an agent must be able to learn
from experience. This makes these reinforcement techniques extremely suited for
CR systems.
This case study can be considered as part of a larger trend toward greater con-
tact between artificial intelligence and other engineering disciplines. Obviously, the
ultimate CR would be a prime example of such symbiosis.
3.5 Conclusions
In this chapter we have introduced the main challenges in the design of CR systems.
It has been shown that those challenges result in a rich set of new design opportu-
nities. We have also given an overview of the system design landscape and moti-
vated our selection for a hybrid DT/RT framework. Afterwards, we introduced our
cognitive framework for distributed optimization of best-effort systems. A detailed
design flow has then been specified to design systems according to this cognitive
framework.
In the remainder of this topic, focus will be on the RT steps of the proposed
framework. As already mentioned in the introduction, four case studies will be con-
sidered that focus on each of the steps: observe RT situation, map RT situation to
system scenario, execute RT procedure and finally learn and calibrate the RT proce-
dure.
