Contributions to Improve Cognitive Strategies with Respect to Wireless Coexistence

Cognitive radio (CR) can identify temporarily available opportunities in a shared radio environment to improve spectral efficiency and coexistence behavior of radio systems. It operates as a secondary user (SU) and accommodates itself in detected opportunities with an intention to avoid harmful collisions with coexisting primary user (PU) systems. Such opportunistic operation of a CR system requires efficient situational awareness and reliable decision making for radio resource allocation.

Situational awareness includes sensing the environment followed by a hypothesis testing for detection of available opportunities in the coexisting environment. This process is often known as spectral hole detection. Situational knowledge can be further enriched by forecasting the primary activities in the radio environment using predictive modeling based approaches. Improved knowledge about the coexisting environment essentially means better decision making for secondary resource allocation. This dissertation identifies limitations of existing predictive modeling and spectral hole detection based resource allocation strategies and suggest improvements.

Firstly, accurate and efficient estimation of statistical parameters of the radio environment is identified as a fundamental challenge to realize predictive modeling based cognitive approaches. Lots of useful training data which are essential to learn the system parameters are not available either because of environmental effects such as noise, interference and fading or because of limited system resources particularly sensor bandwidth. While handling environmental effects to improve signal reception in radio systems has already gained much attention, this dissertation addresses the problem of data losses caused by limited sensor bandwidth as it is totally ignored so far and presents bandwidth independent parameter estimation methods. Where, bandwidth independent means achieving the same level of estimation accuracy for any sensor bandwidth.

Secondly, this dissertation argues that the existing hole detection strategies are dumb because they provide very little information about the coexisting environment. Decision making for resource allocation based on this dumb hole detection approach cannot optimally exploit the opportunities available in the coexisting environment. As a solution, an intelligent hole detection scheme is proposed which suggests classifying the primary systems and using the documented knowledge of identified radio technologies to fully understand their coexistence behavior.

Finally, this dissertation presents a neuro-fuzzy signal classifier (NFSC) that uses bandwidth, operating frequency, pulse shape, hopping behavior and time behavior of signals as distinct features in order to xii identify the PU signals in coexisting environments. This classifier provides the foundation for bandwidth independent parameter estimation and intelligent hole detection. MATLAB/Simulink based simulations are used to support the arguments throughout in this dissertation. A proof-of-concept demonstrator using microcontroller and hardware defined radio (HDR) based transceiver is also presented at the end.

Vorschau

Zitieren

Zitierform:
Zitierform konnte nicht geladen werden.

Rechte

Nutzung und Vervielfältigung:
Alle Rechte vorbehalten